Optimizing CTR Prediction through Clustering-Based Feature Engineering

<!DOCTYPE html>     <html class="no-js" lang="en-US"> <head>  <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> <meta name="author" content="blogtaboola"> <meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0"> <meta name="theme-color" content="#009bc5" /> <meta name="msapplication-TileColor" content="#ffffff"> <meta name="msapplication-TileImage" content="https://blog.taboola.com/wp-content/themes/taboola-new/assets/images/favicon/ms-icon-144x144.png"> <meta name="google-site-verification" content="6HMSpm3CAekh6Wt58ZoQYP0GTtTr52LcNz94V7rp5XI" /> <meta name="category" content="Engineering">  <link rel="profile" href="https://gmpg.org/xfn/11"> <link rel="pingback" href="https://blog.taboola.com/xmlrpc.php"> <link rel="icon" type="image/png" href="https://blog.taboola.com/wp-content/themes/taboola-new/assets/images/favicon.png" /> <meta name='robots' content='index, follow, max-image-preview:large, max-snippet:-1, max-video-preview:-1' />  <title>Optimizing CTR Prediction through Clustering-Based Feature Engineering</title> <meta name="description" content="Learn how meaningful publisher features revolutionize recommendations. Explore data analysis, visualization, and CTR prediction. Witness improved models in action. Say hello to personalized content experiences!" /> <link rel="canonical" href="https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" /> <meta property="og:locale" content="en_US" /> <meta property="og:type" content="article" /> <meta property="og:title" content="Optimizing CTR Prediction through Clustering-Based Feature Engineering" /> <meta property="og:description" content="Learn how meaningful publisher features revolutionize recommendations. Explore data analysis, visualization, and CTR prediction. Witness improved models in action. Say hello to personalized content experiences!" /> <meta property="og:url" content="https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" /> <meta property="og:site_name" content="Taboola Blog" /> <meta property="article:publisher" content="https://www.facebook.com/taboola/" /> <meta property="article:published_time" content="2024-05-21T17:59:24+00:00" /> <meta property="article:modified_time" content="2024-05-27T12:27:47+00:00" /> <meta property="og:image" content="https://blog.taboola.com/wp-content/uploads/2024/05/shutterstock_1918977239.jpg" /> <meta property="og:image:width" content="2000" /> <meta property="og:image:height" content="1091" /> <meta property="og:image:type" content="image/jpeg" /> <meta name="author" content="Ofri Tirosh" /> <meta name="twitter:card" content="summary_large_image" /> <meta name="twitter:creator" content="@taboola" /> <meta name="twitter:site" content="@taboola" /> <meta name="twitter:label1" content="Written by" /> <meta name="twitter:data1" content="Ofri Tirosh" /> <meta name="twitter:label2" content="Est. reading time" /> <meta name="twitter:data2" content="12 minutes" />  <style id='co-authors-plus-coauthors-style-inline-css' type='text/css'> .wp-block-co-authors-plus-coauthors.is-layout-flow [class*=wp-block-co-authors-plus]{display:inline} </style> <style id='co-authors-plus-avatar-style-inline-css' type='text/css'> .wp-block-co-authors-plus-avatar :where(img){height:auto;max-width:100%;vertical-align:bottom}.wp-block-co-authors-plus-coauthors.is-layout-flow .wp-block-co-authors-plus-avatar :where(img){vertical-align:middle}.wp-block-co-authors-plus-avatar:is(.alignleft,.alignright){display:table}.wp-block-co-authors-plus-avatar.aligncenter{display:table;margin-inline:auto} </style> <style id='co-authors-plus-image-style-inline-css' type='text/css'> .wp-block-co-authors-plus-image{margin-bottom:0}.wp-block-co-authors-plus-image :where(img){height:auto;max-width:100%;vertical-align:bottom}.wp-block-co-authors-plus-coauthors.is-layout-flow .wp-block-co-authors-plus-image :where(img){vertical-align:middle}.wp-block-co-authors-plus-image:is(.alignfull,.alignwide) :where(img){width:100%}.wp-block-co-authors-plus-image:is(.alignleft,.alignright){display:table}.wp-block-co-authors-plus-image.aligncenter{display:table;margin-inline:auto} </style> <style id='classic-theme-styles-inline-css' type='text/css'> /*! This file is auto-generated */ .wp-block-button__link{color:#fff;background-color:#32373c;border-radius:9999px;box-shadow:none;text-decoration:none;padding:calc(.667em + 2px) calc(1.333em + 2px);font-size:1.125em}.wp-block-file__button{background:#32373c;color:#fff;text-decoration:none} </style> <style id='global-styles-inline-css' type='text/css'> :root{--wp--preset--aspect-ratio--square: 1;--wp--preset--aspect-ratio--4-3: 4/3;--wp--preset--aspect-ratio--3-4: 3/4;--wp--preset--aspect-ratio--3-2: 3/2;--wp--preset--aspect-ratio--2-3: 2/3;--wp--preset--aspect-ratio--16-9: 16/9;--wp--preset--aspect-ratio--9-16: 9/16;--wp--preset--color--black: #000000;--wp--preset--color--cyan-bluish-gray: #abb8c3;--wp--preset--color--white: #ffffff;--wp--preset--color--pale-pink: #f78da7;--wp--preset--color--vivid-red: #cf2e2e;--wp--preset--color--luminous-vivid-orange: #ff6900;--wp--preset--color--luminous-vivid-amber: #fcb900;--wp--preset--color--light-green-cyan: #7bdcb5;--wp--preset--color--vivid-green-cyan: #00d084;--wp--preset--color--pale-cyan-blue: #8ed1fc;--wp--preset--color--vivid-cyan-blue: #0693e3;--wp--preset--color--vivid-purple: #9b51e0;--wp--preset--gradient--vivid-cyan-blue-to-vivid-purple: linear-gradient(135deg,rgba(6,147,227,1) 0%,rgb(155,81,224) 100%);--wp--preset--gradient--light-green-cyan-to-vivid-green-cyan: linear-gradient(135deg,rgb(122,220,180) 0%,rgb(0,208,130) 100%);--wp--preset--gradient--luminous-vivid-amber-to-luminous-vivid-orange: linear-gradient(135deg,rgba(252,185,0,1) 0%,rgba(255,105,0,1) 100%);--wp--preset--gradient--luminous-vivid-orange-to-vivid-red: linear-gradient(135deg,rgba(255,105,0,1) 0%,rgb(207,46,46) 100%);--wp--preset--gradient--very-light-gray-to-cyan-bluish-gray: linear-gradient(135deg,rgb(238,238,238) 0%,rgb(169,184,195) 100%);--wp--preset--gradient--cool-to-warm-spectrum: linear-gradient(135deg,rgb(74,234,220) 0%,rgb(151,120,209) 20%,rgb(207,42,186) 40%,rgb(238,44,130) 60%,rgb(251,105,98) 80%,rgb(254,248,76) 100%);--wp--preset--gradient--blush-light-purple: linear-gradient(135deg,rgb(255,206,236) 0%,rgb(152,150,240) 100%);--wp--preset--gradient--blush-bordeaux: linear-gradient(135deg,rgb(254,205,165) 0%,rgb(254,45,45) 50%,rgb(107,0,62) 100%);--wp--preset--gradient--luminous-dusk: linear-gradient(135deg,rgb(255,203,112) 0%,rgb(199,81,192) 50%,rgb(65,88,208) 100%);--wp--preset--gradient--pale-ocean: linear-gradient(135deg,rgb(255,245,203) 0%,rgb(182,227,212) 50%,rgb(51,167,181) 100%);--wp--preset--gradient--electric-grass: linear-gradient(135deg,rgb(202,248,128) 0%,rgb(113,206,126) 100%);--wp--preset--gradient--midnight: linear-gradient(135deg,rgb(2,3,129) 0%,rgb(40,116,252) 100%);--wp--preset--font-size--small: 13px;--wp--preset--font-size--medium: 20px;--wp--preset--font-size--large: 36px;--wp--preset--font-size--x-large: 42px;--wp--preset--spacing--20: 0.44rem;--wp--preset--spacing--30: 0.67rem;--wp--preset--spacing--40: 1rem;--wp--preset--spacing--50: 1.5rem;--wp--preset--spacing--60: 2.25rem;--wp--preset--spacing--70: 3.38rem;--wp--preset--spacing--80: 5.06rem;--wp--preset--shadow--natural: 6px 6px 9px rgba(0, 0, 0, 0.2);--wp--preset--shadow--deep: 12px 12px 50px rgba(0, 0, 0, 0.4);--wp--preset--shadow--sharp: 6px 6px 0px rgba(0, 0, 0, 0.2);--wp--preset--shadow--outlined: 6px 6px 0px -3px rgba(255, 255, 255, 1), 6px 6px rgba(0, 0, 0, 1);--wp--preset--shadow--crisp: 6px 6px 0px rgba(0, 0, 0, 1);}:where(.is-layout-flex){gap: 0.5em;}:where(.is-layout-grid){gap: 0.5em;}body .is-layout-flex{display: flex;}.is-layout-flex{flex-wrap: wrap;align-items: center;}.is-layout-flex > :is(*, div){margin: 0;}body .is-layout-grid{display: grid;}.is-layout-grid > :is(*, div){margin: 0;}:where(.wp-block-columns.is-layout-flex){gap: 2em;}:where(.wp-block-columns.is-layout-grid){gap: 2em;}:where(.wp-block-post-template.is-layout-flex){gap: 1.25em;}:where(.wp-block-post-template.is-layout-grid){gap: 1.25em;}.has-black-color{color: var(--wp--preset--color--black) !important;}.has-cyan-bluish-gray-color{color: var(--wp--preset--color--cyan-bluish-gray) !important;}.has-white-color{color: var(--wp--preset--color--white) !important;}.has-pale-pink-color{color: var(--wp--preset--color--pale-pink) !important;}.has-vivid-red-color{color: var(--wp--preset--color--vivid-red) !important;}.has-luminous-vivid-orange-color{color: var(--wp--preset--color--luminous-vivid-orange) !important;}.has-luminous-vivid-amber-color{color: var(--wp--preset--color--luminous-vivid-amber) !important;}.has-light-green-cyan-color{color: var(--wp--preset--color--light-green-cyan) !important;}.has-vivid-green-cyan-color{color: var(--wp--preset--color--vivid-green-cyan) !important;}.has-pale-cyan-blue-color{color: var(--wp--preset--color--pale-cyan-blue) !important;}.has-vivid-cyan-blue-color{color: var(--wp--preset--color--vivid-cyan-blue) !important;}.has-vivid-purple-color{color: var(--wp--preset--color--vivid-purple) !important;}.has-black-background-color{background-color: var(--wp--preset--color--black) !important;}.has-cyan-bluish-gray-background-color{background-color: var(--wp--preset--color--cyan-bluish-gray) !important;}.has-white-background-color{background-color: var(--wp--preset--color--white) !important;}.has-pale-pink-background-color{background-color: var(--wp--preset--color--pale-pink) !important;}.has-vivid-red-background-color{background-color: var(--wp--preset--color--vivid-red) !important;}.has-luminous-vivid-orange-background-color{background-color: var(--wp--preset--color--luminous-vivid-orange) !important;}.has-luminous-vivid-amber-background-color{background-color: var(--wp--preset--color--luminous-vivid-amber) !important;}.has-light-green-cyan-background-color{background-color: var(--wp--preset--color--light-green-cyan) !important;}.has-vivid-green-cyan-background-color{background-color: var(--wp--preset--color--vivid-green-cyan) !important;}.has-pale-cyan-blue-background-color{background-color: var(--wp--preset--color--pale-cyan-blue) !important;}.has-vivid-cyan-blue-background-color{background-color: var(--wp--preset--color--vivid-cyan-blue) !important;}.has-vivid-purple-background-color{background-color: var(--wp--preset--color--vivid-purple) !important;}.has-black-border-color{border-color: var(--wp--preset--color--black) !important;}.has-cyan-bluish-gray-border-color{border-color: var(--wp--preset--color--cyan-bluish-gray) !important;}.has-white-border-color{border-color: var(--wp--preset--color--white) !important;}.has-pale-pink-border-color{border-color: var(--wp--preset--color--pale-pink) !important;}.has-vivid-red-border-color{border-color: var(--wp--preset--color--vivid-red) !important;}.has-luminous-vivid-orange-border-color{border-color: var(--wp--preset--color--luminous-vivid-orange) !important;}.has-luminous-vivid-amber-border-color{border-color: var(--wp--preset--color--luminous-vivid-amber) !important;}.has-light-green-cyan-border-color{border-color: var(--wp--preset--color--light-green-cyan) !important;}.has-vivid-green-cyan-border-color{border-color: var(--wp--preset--color--vivid-green-cyan) !important;}.has-pale-cyan-blue-border-color{border-color: var(--wp--preset--color--pale-cyan-blue) !important;}.has-vivid-cyan-blue-border-color{border-color: var(--wp--preset--color--vivid-cyan-blue) !important;}.has-vivid-purple-border-color{border-color: var(--wp--preset--color--vivid-purple) !important;}.has-vivid-cyan-blue-to-vivid-purple-gradient-background{background: var(--wp--preset--gradient--vivid-cyan-blue-to-vivid-purple) !important;}.has-light-green-cyan-to-vivid-green-cyan-gradient-background{background: var(--wp--preset--gradient--light-green-cyan-to-vivid-green-cyan) !important;}.has-luminous-vivid-amber-to-luminous-vivid-orange-gradient-background{background: var(--wp--preset--gradient--luminous-vivid-amber-to-luminous-vivid-orange) !important;}.has-luminous-vivid-orange-to-vivid-red-gradient-background{background: var(--wp--preset--gradient--luminous-vivid-orange-to-vivid-red) !important;}.has-very-light-gray-to-cyan-bluish-gray-gradient-background{background: var(--wp--preset--gradient--very-light-gray-to-cyan-bluish-gray) !important;}.has-cool-to-warm-spectrum-gradient-background{background: var(--wp--preset--gradient--cool-to-warm-spectrum) !important;}.has-blush-light-purple-gradient-background{background: var(--wp--preset--gradient--blush-light-purple) !important;}.has-blush-bordeaux-gradient-background{background: var(--wp--preset--gradient--blush-bordeaux) !important;}.has-luminous-dusk-gradient-background{background: var(--wp--preset--gradient--luminous-dusk) !important;}.has-pale-ocean-gradient-background{background: var(--wp--preset--gradient--pale-ocean) !important;}.has-electric-grass-gradient-background{background: var(--wp--preset--gradient--electric-grass) !important;}.has-midnight-gradient-background{background: var(--wp--preset--gradient--midnight) !important;}.has-small-font-size{font-size: var(--wp--preset--font-size--small) !important;}.has-medium-font-size{font-size: var(--wp--preset--font-size--medium) !important;}.has-large-font-size{font-size: var(--wp--preset--font-size--large) !important;}.has-x-large-font-size{font-size: var(--wp--preset--font-size--x-large) !important;} :where(.wp-block-post-template.is-layout-flex){gap: 1.25em;}:where(.wp-block-post-template.is-layout-grid){gap: 1.25em;} :where(.wp-block-columns.is-layout-flex){gap: 2em;}:where(.wp-block-columns.is-layout-grid){gap: 2em;} :root :where(.wp-block-pullquote){font-size: 1.5em;line-height: 1.6;} </style> <link rel='stylesheet' id='taboola-cache-manager-css' href='https://blog.taboola.com/wp-content/plugins/taboola-cache-manager/public/css/taboola-cache-manager-public.css?ver=1.3.2' type='text/css' media='all' /> <link rel="icon" href="https://blog.taboola.com/wp-content/uploads/2019/12/cropped-favicon-32x32.png" sizes="32x32" /> <link rel="icon" href="https://blog.taboola.com/wp-content/uploads/2019/12/cropped-favicon-192x192.png" sizes="192x192" /> <link rel="apple-touch-icon" href="https://blog.taboola.com/wp-content/uploads/2019/12/cropped-favicon-180x180.png" /> <meta name="msapplication-TileImage" content="https://blog.taboola.com/wp-content/uploads/2019/12/cropped-favicon-270x270.png" />  <link rel="preconnect" href="https://dev.visualwebsiteoptimizer.com" /> <script type='text/javascript' id='vwoCode'> window._vwo_code || (function() { var account_id = 750209, version = 2.0, settings_tolerance = 2000, hide_element = 'body', hide_element_style = 'opacity:0 !important;filter:alpha(opacity=0) !important;background:none !important', /* DO NOT EDIT BELOW THIS LINE */ f = false, w = window, d = document, v = d.querySelector('#vwoCode'), cK = '_vwo_' + account_id + '_settings', cc = {}; try { var c = JSON.parse(localStorage.getItem('_vwo_' + account_id + '_config')); cc = c && typeof c === 'object' ? c : {} } catch (e) {} var stT = cc.stT === 'session' ? w.sessionStorage : w.localStorage; code = { use_existing_jquery: function() { return typeof use_existing_jquery !== 'undefined' ? use_existing_jquery : undefined }, library_tolerance: function() { return typeof library_tolerance !== 'undefined' ? library_tolerance : undefined }, settings_tolerance: function() { return cc.sT || settings_tolerance }, hide_element_style: function() { return '{' + (cc.hES || hide_element_style) + '}' }, hide_element: function() { return typeof cc.hE === 'string' ? cc.hE : hide_element }, getVersion: function() { return version }, finish: function() { if (!f) { f = true; var e = d.getElementById('_vis_opt_path_hides'); if (e) e.parentNode.removeChild(e) } }, finished: function() { return f }, load: function(e) { var t = this.getSettings(), n = d.createElement('script'), i = this; if (t) { n.textContent = t; d.getElementsByTagName('head')[0].appendChild(n); if (!w.VWO || VWO.caE) { stT.removeItem(cK); i.load(e) } } else { n.fetchPriority = 'high'; n.src = e; n.type = 'text/javascript'; n.onerror = function() { _vwo_code.finish() }; d.getElementsByTagName('head')[0].appendChild(n) } }, getSettings: function() { try { var e = stT.getItem(cK); if (!e) { return } e = JSON.parse(e); if (Date.now() > e.e) { stT.removeItem(cK); return } return e.s } catch (e) { return } }, init: function() { if (d.URL.indexOf('__vwo_disable__') > -1) return; var e = this.settings_tolerance(); w._vwo_settings_timer = setTimeout(function() { _vwo_code.finish(); stT.removeItem(cK) }, e); var t = d.currentScript, n = d.createElement('style'), i = this.hide_element(), r = t && !t.async && i ? i + this.hide_element_style() : '', c = d.getElementsByTagName('head')[0]; n.setAttribute('id', '_vis_opt_path_hides'); v && n.setAttribute('nonce', v.nonce); n.setAttribute('type', 'text/css'); if (n.styleSheet) n.styleSheet.cssText = r; else n.appendChild(d.createTextNode(r)); c.appendChild(n); this.load('https://dev.visualwebsiteoptimizer.com/j.php?a=' + account_id + '&u=' + encodeURIComponent(d.URL) + '&vn=' + version) } }; w._vwo_code = code; code.init(); })(); </script>   <script> (function(w, d, s, l, i) { w[l] = w[l] || []; w[l].push({ 'gtm.start': new Date().getTime(), event: 'gtm.js' }); var f = d.getElementsByTagName(s)[0], j = d.createElement(s), dl = l != 'dataLayer' ? '&l=' + l : ''; j.async = true; j.src = 'https://www.googletagmanager.com/gtm.js?id=' + i + dl; f.parentNode.insertBefore(j, f); })(window, document, 'script', 'dataLayer', 'GTM-PC4JX7'); </script>  <script type="text/javascript"> window._taboola = window._taboola || []; _taboola.push({article:'auto'}); !function (e, f, u, i) { if (!document.getElementById(i)){ e.async = 1; e.src = u; e.id = i; f.parentNode.insertBefore(e, f); } }(document.createElement('script'), document.getElementsByTagName('script')[0], '//cdn.taboola.com/libtrc/taboola-blog/loader.js', 'tb_loader_script'); if(window.performance && typeof window.performance.mark == 'function') {window.performance.mark('tbl_ic');} </script> <style> label{display:inline-block;max-width:100%;margin-bottom:5px;font-weight:700}input[type=search]{box-sizing:border-box}input[type=radio]{margin:4px 0 0;line-height:normal}input[type=search]{-webkit-appearance:none}.btn{display:inline-block;padding:6px 12px;margin-bottom:0;font-size:14px;font-weight:400;line-height:1.42857143;text-align:center;white-space:nowrap;vertical-align:middle;-ms-touch-action:manipulation;touch-action:manipulation;background-image:none;border:1px solid transparent;border-radius:4px}.progress-bar{float:left;width:0;height:100%;font-size:12px;line-height:20px;color:#fff;text-align:center;background-color:#337ab7;box-shadow:inset 0 -1px 0 rgba(0,0,0,.15)}.container:after,.container:before,.row:after,.row:before{display:table;content:" "}.container:after,.row:after{clear:both}@-ms-viewport{width:100vw}html{font-family:sans-serif;-webkit-text-size-adjust:100%;-ms-text-size-adjust:100%}body{margin:0}aside,header{display:block}a{background-color:transparent}h1{margin:.67em 0;font-size:2em}svg:not(:root){overflow:hidden}input{margin:0;font:inherit;color:inherit}input[type=submit]{-webkit-appearance:button}input::-moz-focus-inner{padding:0;border:0}input{line-height:normal}input[type=radio]{box-sizing:border-box;padding:0}input[type=search]{box-sizing:content-box;-webkit-appearance:textfield}input[type=search]::-webkit-search-cancel-button,input[type=search]::-webkit-search-decoration{-webkit-appearance:none}*{box-sizing:border-box}:after,:before{box-sizing:border-box}html{font-size:10px}body{font-family:poppins,Helvetica,Arial,sans-serif;font-size:16px;line-height:1.5;color:#333;background-color:#fff}input{font-family:inherit;font-size:inherit;line-height:inherit}a{color:#337ab7;text-decoration:none}h1,h4{font-family:inherit;font-weight:500;line-height:1.1;color:inherit}h1{margin-top:20px;margin-bottom:10px}h4{margin-top:10px;margin-bottom:10px}h1{font-size:36px}h4{font-size:18px}p{margin:0 0 10px}ul{margin-top:0;margin-bottom:10px}ul ul{margin-bottom:0}.container{width:100%;padding-right:15px;padding-left:15px;margin-right:auto;margin-left:auto}.row{margin-right:-15px;margin-left:-15px}.col-md-11,.col-md-4,.col-md-8,.col-xs-12{position:relative;min-height:1px;padding-right:15px;padding-left:15px}.col-xs-12{float:left}.col-xs-12{width:100%}.pull-right{float:right!important}.pull-left{float:left!important}.hidden{display:none!important}html{font-family:sans-serif;-ms-text-size-adjust:100%;-webkit-text-size-adjust:100%}body{margin:0}aside,header{display:block}a{background-color:transparent}h1{font-size:2em;margin:.67em 0}svg:not(:root){overflow:hidden}input{color:inherit;font:inherit;margin:0}input[type=submit]{-webkit-appearance:button}input::-moz-focus-inner{border:0;padding:0}input{line-height:normal}input[type=radio]{box-sizing:border-box;padding:0}input[type=search]{-webkit-appearance:textfield;box-sizing:content-box}input[type=search]::-webkit-search-cancel-button,input[type=search]::-webkit-search-decoration{-webkit-appearance:none}html,input[type=search]{box-sizing:border-box}*,:after,:before{box-sizing:inherit}html{scroll-behavior:smooth}body{font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:300!important;max-width:100%;overflow-x:hidden}body #wrapper{background-color:#fff}.bg_turquoise{background-color:#00a0a0}.bg_turquoise{background-color:#0056f0}iframe{max-width:100%}.middot{display:inline-block;border-radius:50%;height:3px;width:3px;background-color:#777474;vertical-align:middle}@font-face{font-family:Poppins;font-weight:300;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Light.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Light.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Light.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Light.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Light.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:400;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Regular.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Regular.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Regular.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Regular.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Regular.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:500;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Medium.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Medium.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Medium.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Medium.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Medium.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:600;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-SemiBold.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-SemiBold.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-SemiBold.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-SemiBold.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-SemiBold.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:700;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Bold.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Bold.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Bold.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Bold.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Bold.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:800;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-ExtraBold.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-ExtraBold.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-ExtraBold.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-ExtraBold.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-ExtraBold.ttf) format("truetype");font-display:swap}@font-face{font-family:Poppins;font-weight:900;font-style:normal;src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Black.eot);src:url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Black.eot?#iefix) format("embedded-opentype"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Black.woff2) format("woff2"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Black.woff) format("woff"),url(/wp-content/themes/taboola-new/assets/fonts/poppins/Poppins-Black.ttf) format("truetype");font-display:swap}.mb-20{margin:0 0 20px 0}.mt-20{margin:20px 0 0 0}body header#header{position:fixed;background-color:#fff;color:#0056f0;width:100%;top:0;height:74px;display:-ms-flexbox;display:flex;-ms-flex-align:center;align-items:center;left:0;z-index:1000;box-shadow:0 0 5px 0 rgba(0,0,0,.2)}body header#header a{color:#555;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:300!important;font-size:16px}body header#header #logo{width:140px;height:40px;display:inline-block;vertical-align:middle}body header#header #logo svg{width:100%}body header#header #desktop-menu{width:auto;display:none;padding:5px;margin-left:40px}body header#header #search{width:auto;display:none;position:relative;color:#555}body header#header #search *{color:#555}body header#header #search form.search-form{line-height:1;height:30px;border-radius:5px;color:#555;position:absolute;right:0;top:-14px}body header#header #search form.search-form label{width:0;overflow:hidden}body header#header #search form.search-form label span{display:none}body header#header #search form.search-form label input.search-field{border:none;border-bottom:1px solid #0056f0;color:#0056f0;margin:0;padding:5px 0;width:100%;overflow:hidden;opacity:0}body header#header #search form.search-form label input.search-field:-ms-input-placeholder{color:#0056f0;opacity:1}body header#header #search form.search-form label input.search-field:-ms-input-placeholder{color:#0056f0}body header#header #search form.search-form label input.search-field::-ms-input-placeholder{color:#0056f0}body header#header #search form.search-form input.search-submit{border:none;float:right;width:30px;height:30px;font-size:0;position:relative;padding:0;margin:0;color:#0056f0!important;z-index:-2}body header#header #search form.search-form i.fa{height:0;width:0;color:#0056f0!important;overflow:hidden;position:absolute;right:54px;left:auto;top:27px;background-image:url(/wp-content/themes/taboola-new/assets/images/close.svg);background-position:center;background-size:contain;background-repeat:no-repeat}body header#header #search form.search-form:after{content:"";width:20px;height:20px;position:absolute;z-index:1;right:0;top:4px;background-image:url(/wp-content/themes/taboola-new/assets/images/search-teal.svg);background-position:center;background-size:contain;background-repeat:no-repeat}body header#header .action-menu{display:-ms-flexbox;display:flex;-ms-flex-align:center;align-items:center;-ms-flex-pack:justify;justify-content:space-between;-ms-flex-direction:row-reverse;flex-direction:row-reverse;margin-top:7px}body header#header #mobile-menu{float:right;width:50px;margin-top:5px}body header#header #langs{position:relative;display:inline-block;vertical-align:middle}body header#header #langs .earth:before{content:"";background-image:url(/wp-content/themes/taboola-new/assets/images/globe.svg);background-position:center;background-size:contain;background-repeat:no-repeat;display:inline-block;width:27px;height:27px}body header#header #langs ul.lang_ul{position:absolute;z-index:1000;width:105px;list-style:none;font-size:14px;text-align:left;background-color:#fff;background-clip:padding-box;margin:0 -50px 0 0;right:50%;border:none;border-radius:5px;padding:5px 0;overflow:hidden;opacity:0;height:0;top:100%}body header#header #langs ul.lang_ul a{color:#555;text-decoration:blink!important;text-transform:capitalize;font-size:15px}body header#header #langs ul.lang_ul li{width:100%;background-color:transparent;color:#fff;padding:5px 15px}body header#header #langs ul.lang_ul li:last-child:after{display:none!important}body header#header .go_taboola{border:solid 1px transparent;text-align:center;border-radius:25px;text-decoration:blink;display:none;margin-top:2px;color:#0056f0;text-decoration:underline}body #wrapper{padding-top:74px}body #breadcrumbs{list-style:none;margin:0 0 20px;overflow:hidden;padding:0}body #breadcrumbs li{display:inline-block;vertical-align:middle;margin-right:5px}body #breadcrumbs li a{color:#e0e0e0}body #breadcrumbs li:after{content:"›";margin-left:5px;color:#e0e0e0}body #breadcrumbs li:last-of-type{color:#0056f0}body #breadcrumbs li:last-of-type:after{content:none;margin-left:0}#logo svg{fill:#0056f0!important;height:40px;width:140px;margin-top:5px}#desktop-menu .menu{margin:0;padding:0;list-style:none}#desktop-menu .menu li{position:relative}#desktop-menu .menu li.menu-item{display:inline-block;vertical-align:middle;margin-right:70px}#desktop-menu .menu li.menu-item .cj-dropdown-menu{display:none;position:absolute;opacity:0;top:102%;left:-20px;background:#fff;background:linear-gradient(180deg,rgba(255,255,255,0) 0,#fff 5%,#fff 100%);padding:30px 20px 20px;min-width:calc(100% + 40px);white-space:nowrap;border-radius:4px;box-shadow:0 20px 10px -12px rgba(0,0,0,.3);list-style:none}#desktop-menu .menu li.menu-item .cj-dropdown-menu li{margin:0 0 5px 0}#desktop-menu .menu li.menu-item .cj-dropdown-menu li a{font-size:14px}#desktop-menu .menu li.menu-item .cj-dropdown-menu li a.category-title{font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:500!important;color:#0056f0;margin-bottom:15px;display:block}#desktop-menu .menu li.menu-item .cj-dropdown-menu ul{list-style:none;padding:0}#desktop-menu .menu li.menu-item .cj-dropdown-menu .category-link{color:#555;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:300!important;display:block}#desktop-menu .menu li.menu-item .cj-dropdown-menu .showMoreCategories{margin-bottom:5px;display:block;font-size:14px}#desktop-menu .menu li.menu-item .cj-dropdown-menu .moreCategories{height:0;visibility:hidden;margin:0}#desktop-menu .menu li.menu-item .cj-dropdown-menu .menu-category-list{margin-right:50px;border-right:1px solid #efefef}#desktop-menu .menu li.menu-item .cj-dropdown-menu .menu-category-list:last-of-type{border-right:none;margin-right:0}#desktop-menu .menu li.menu-item .cj-dropdown-menu .menu-column{margin-right:50px}#desktop-menu .menu li.menu-item .cj-dropdown-menu .flex-row{display:-ms-flexbox;display:flex;-ms-flex-direction:row;flex-direction:row}#desktop-menu .menu li.menu-item .cj-dropdown-menu .flex-col{display:-ms-flexbox;display:flex;-ms-flex-direction:column;flex-direction:column}#desktop-menu .menu li .parent-link{position:relative}#desktop-menu .menu li .parent-link:before{height:1px;content:"";background-color:#0056f0;width:0;bottom:-4px;position:absolute;left:0}#mobile-menu .menu_slide{transform:translateX(100%);background-color:#0056f0;width:280px;height:calc(100% - 73px);line-height:1;overflow:hidden;position:fixed;bottom:0;right:0;border-top:1px solid #fff}#mobile-menu .menu_slide .menu-scroll-container{overflow-y:scroll;height:calc(100% - 48px)}#mobile-menu .menu_slide #mobile-categories-list{margin:0;padding:0;list-style:none}#mobile-menu .menu_slide #mobile-categories-list>li{color:#fff}#mobile-menu .menu_slide #mobile-categories-list input{display:none}#mobile-menu .menu_slide #mobile-categories-list label{display:-ms-flexbox;display:flex;width:100%;margin:0;padding:22px 30px 16px;font-size:18px;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:300!important;border-bottom:1px solid #0c6af2;position:relative}#mobile-menu .menu_slide #mobile-categories-list label .toggle-arrow{background-image:url(/wp-content/themes/taboola-new/assets/images/arrow.svg);background-position:center;background-size:contain;background-repeat:no-repeat;width:16px;height:16px;margin-top:-8px;transform:rotate(-90deg);position:absolute;right:25px;top:50%}#mobile-menu .menu_slide #mobile-categories-list .content{margin:0;padding:0;list-style:none;max-height:0;overflow:hidden;background:#0c6af2;border-bottom:1px solid #0056f0}#mobile-menu .menu_slide #mobile-categories-list .content li{line-height:1.3;padding:10px 0 10px 45px}#mobile-menu .menu_slide #mobile-categories-list .content li a{color:#fff;font-size:14px}#mobile-menu .menu_slide #mobile-categories-list .content .showMoreCategories{display:block}#mobile-menu .menu_slide #mobile-categories-list .content .moreCategories{display:none}#mobile-menu .menu_slide .taboola-com-btn{position:absolute;left:0;width:100%;bottom:0;right:0;padding:18px 20px 14px 30px;background:#0c6af2;color:#fff}#mobile-menu .menu_slide .search{width:100%;margin:0;display:block;padding:22px 30px 20px 30px;border-bottom:1px solid #0c6af2}#mobile-menu .menu_slide .search form.search-form{position:relative;line-height:1}#mobile-menu .menu_slide .search form.search-form label{width:100%;overflow:hidden;margin:0}#mobile-menu .menu_slide .search form.search-form label span{display:none}#mobile-menu .menu_slide .search form.search-form label input.search-field{border:none;border-bottom:1px solid #fff;margin:0;width:100%;overflow:hidden;-webkit-appearance:none;padding:5px 25px 5px 0}#mobile-menu .menu_slide .search form.search-form input.search-submit{border:none;float:right;width:30px;height:30px;font-size:0;padding:0;margin:0;z-index:2;position:absolute;top:0;right:0}#mobile-menu .menu_slide .search form.search-form:after{content:"";background-image:url(/wp-content/themes/taboola-new/assets/images/search.svg);background-position:center;background-size:contain;background-repeat:no-repeat;width:16px;height:16px;position:absolute;z-index:1;right:0;top:5px;text-align:center}#mobile-menu .triger_menu{width:30px;background-color:transparent;padding:5px 0;float:right;display:block}#mobile-menu .triger_menu span{width:100%;height:3px;margin-bottom:6px;border-radius:3px;display:block;background-color:#0056f0}#mobile-menu .triger_menu span:last-child{margin-bottom:0}.menu-block{display:none;width:100%;height:130vh;position:fixed;background:rgba(162,162,162,.8);top:0;left:0;z-index:1}body #featured{background-position:center;background-size:cover;min-height:auto!important;max-height:none!important;color:#fff;text-align:center;margin-bottom:45px;display:block;width:100%}footer#footer .widget_footer ul li{margin:0 0 15px}footer#footer .widget_footer ul li a{color:#676767;font-size:14px}form.search-form label{font-size:0;line-height:1}form.search-form input{font-size:16px;color:#fff;background-color:transparent;font-weight:100;width:100%;margin:8px 0;border-radius:1px;padding:15px}#header .search-result{display:none;position:absolute;background:#fff;top:60px;width:100%;text-align:left;box-shadow:0 20px 10px -12px rgba(0,0,0,.3);padding:0 2em 1em 2em}#sidebar{max-width:305px;padding:0}#sidebar aside{width:100%;padding:0;margin:0 0 30px;color:#fff;font-size:17px;text-align:left;color:#333;font-size:14px}#sidebar aside.bg_turquoise{border:none;background-color:transparent;margin:15px 0 30px 0;background-color:#0056f0}#sidebar aside.bg_turquoise .campaign-cta-wrapper{width:100%;padding:40px 40px 40px 48px}#sidebar aside.bg_turquoise .campaign-cta-wrapper h4{color:#fff;margin:10px 0;width:100%;font-size:27px;line-height:1.2em}#sidebar aside.bg_turquoise .btn{font-size:16px;color:#fff;background-color:#0056f0;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:400!important;border-radius:30px;display:inline-block;padding:9px 30px 9px;border:1px solid #fff;margin:5px 0 0 0;text-transform:capitalize;min-width:auto;max-width:100%;white-space:normal}#strip_campaign .btn.campaign{font-size:16px;color:#fff;background-color:#0056f0;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:400!important;border-radius:30px;display:inline-block;padding:9px 30px 9px;border:1px solid #fff;margin:5px 0 0 0;text-transform:capitalize;min-width:auto;max-width:100%;white-space:normal}body.single #content .sticky-mobile-cta{background-color:#0056f0;position:fixed;bottom:0;left:0;right:0;width:100%;height:60px;bottom:-60px;z-index:999;padding:20px 0;text-align:center}body.single #content .sticky-mobile-cta a{color:#fff}.campaign-cta-wrapper a{color:#0056f0!important}@media (min-width:768px){.container{width:750px}body header#header #search form.search-form label{margin-bottom:-1px}}@media screen and (min-width:768px){body header#header .action-menu{width:150px}}@media only screen and (min-width:990px){body.single #content .sticky-mobile-cta{display:none}}@media (min-width:992px){.container{width:970px}.col-md-11,.col-md-4,.col-md-8{float:left}.col-md-11{width:91.66666667%}.col-md-8{width:66.66666667%}.col-md-4{width:33.33333333%}}@media only screen and (min-width:992px){footer#footer .widget_footer ul li{margin:0 0 23px}#sidebar aside{padding:0 0;margin:0 0 60px!important}}@media screen and (min-width:1024px){.col-centered{float:none;margin:0 auto;padding:0}}@media (min-width:1200px){.container{width:1170px}}@media only screen and (min-width:1200px){body header#header .container{padding-top:6px}body header#header #desktop-menu{display:block;margin-top:4px}body header#header #search{display:inline-block;vertical-align:middle}body header#header #langs .earth:before{width:20px;height:20px;margin-top:5px}body header#header .go_taboola{display:inline-block}#mobile-menu{display:none}}@media screen and (min-height:789px){#sidebar aside{font-size:16px}}@media screen and (min-height:960px){#sidebar aside{margin:0 0 60px!important}}@media only screen and (max-width:992px){#sidebar{display:none}}@media screen and (max-width:460px){body #breadcrumbs li{font-size:10px}}.hs-inline-form{background-color:#f2f5f7;padding:60px 80px;margin:30px 0}.hs-inline-form h4{margin-bottom:30px}input:-ms-input-placeholder{color:#555}input::-webkit-input-placeholder{color:#555}input::-moz-placeholder{color:#555}input:-ms-input-placeholder{color:#555}input:-moz-placeholder{color:#555}.single-post-share-buttons{margin:20px 0}.single-post-share-buttons .share-button{width:35px;height:35px;display:inline-block;vertical-align:middle;margin-right:15px;background-position:center;background-size:contain;background-repeat:no-repeat}.single-post-share-buttons .facebook{background-image:url(/wp-content/themes/taboola-new/assets/images/facebook_gray.svg)}.single-post-share-buttons .twitter{background-image:url(/wp-content/themes/taboola-new/assets/images/twitter_gray.svg)}.single-post-share-buttons .linkedin{background-image:url(/wp-content/themes/taboola-new/assets/images/linkedin_gray.svg)}body.single header.post-info{background-color:#fff;color:#000;margin-bottom:0;padding:0;display:inline-block}body.single header.post-info .post-title h1{font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:500!important;color:#333;text-align:left;font-size:30px;line-height:1.2}body.single header.post-info .post-by,body.single header.post-info .post-info-meta{color:#777474;font-size:14px;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:400!important;text-align:left;margin-bottom:0}body.single header.post-info .post-by{font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:400!important}body.single header.post-info .post-by a{color:#777474!important}body.single header.post-info .middot{margin-left:5px;margin-right:5px}body.single #featured{height:435px;margin-bottom:25px}body.single #featured.image-post-desktop{display:none}body.single #featured.image-post-mobile{display:block}body.single #content{padding-bottom:45px}body.single #content p{color:#5d5d5d;line-height:1.7;margin-bottom:20px;text-align:left}body.single #content #breadcrumbs{padding-top:30px;margin:0 0 25px;font-size:14px}body.single #content #breadcrumbs a,body.single #content #breadcrumbs li:last-of-type{color:#777474}body.single #content #breadcrumbs .post-category a{font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:400!important;color:#0056f0;font-size:14px}body.single #content a:not(.btn){color:#0056f0}body.single #content i,body.single #content p{font-size:16px;overflow-wrap:anywhere;word-break:normal}body.single .right-sidebar-wrapper .sidebar-item{margin-bottom:20px}body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{padding:15px 10px;text-align:left;width:213px}body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper a{background-color:#0c6af2;color:#fff!important;padding:10px 30px;border-radius:25px;min-width:200px;border:none;display:inline-block;font-family:Poppins,Helvetica,Arial,sans-serif;font-weight:500!important}.row-no-padding [class*=col-]{padding-left:0!important;padding-right:0!important}.col-centered{float:none;margin:0 auto}#sidebar{position:sticky;top:110px}.progress-container{position:fixed;top:0;z-index:900;width:100%;background-color:#f1f1f1;width:100%;height:8px;background:#ccc}.progress-bar{height:8px;background:#0c6af2;width:0%}@media only screen and (min-width:768px){body.single #featured.image-post-desktop{display:block}body.single #featured.image-post-mobile{display:none}}@media only screen and (min-width:992px){body.single header.post-info .post-title h1{font-size:33px}body.single #content p{line-height:1.7;margin-bottom:25px}}@media screen and (min-width:1024px){.single-post-share-buttons{position:absolute;margin:0 0 0 -60px;top:0}.single-post-share-buttons .share-button{display:block;margin-bottom:10px;margin-right:0}body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{display:block}}@media screen and (min-height:667px){body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{padding:20px}}@media screen and (min-height:789px){body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{padding:30px}}@media screen and (min-height:960px){body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{padding:45px 30px}#sidebar{top:160px}}@media screen and (max-width:1200px){.progress-container{display:none}}@media screen and (max-width:767px){body.single .right-sidebar-wrapper .sidebar-item .campaign-cta-wrapper{width:100%}}@media only screen and (max-width:412px){body.single #featured{margin-bottom:0}body.single #content #breadcrumbs{padding-top:5px;list-style:1}} </style> <script type='application/ld+json'> { "@context": "http://schema.org/", "@type": "BlogPosting", "headline": "Optimizing CTR Prediction through Clustering-Based Feature Engineering", "image": { "@type": "ImageObject", "url": "https://blog.taboola.com/wp-content/uploads/2024/05/shutterstock_1918977239.jpg" }, "datePublished": "2024-05-21", "dateModified": "2024-05-27 ", "author": { "@type": "Person", "name": "Ofri Tirosh" }, "publisher": { "@type": "Organization", "name": "Taboola", "url": "https://www.taboola.com", "logo": { "@type": "ImageObject", "url": "https://blog.taboola.com/wp-content/themes/taboola/assets/images/logo.jpg" }, "mainEntityOfPage": "https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" } } </script> </head> <body class="single animated loaded"> <div class="menu-block" style=""></div>  <noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-PC4JX7" height="0" width="0" style="display:none;visibility:hidden"></iframe></noscript>  <div id="wrapper"> <header id="header"> <div class="container"> <a id="logo" class="pull-left" href="https://blog.taboola.com" title="Taboola Blog"> <svg viewBox="0 0 1003.26 190.69" fill="#0056f0" stock-width="0px"> <title>Taboola Blog</title> <g> <g> <g> <path d="M37.06,40.59H0V6.77h114.79v33.82h-37.1l.02,99.12h-40.63l-.02-99.12Z" /> <path d="M148.6,96.28c-4.46,2.2-9.48,3.47-14.49,4.55-8.18,1.81-12.45,3.82-12.49,10.31-.02,4.46,4.77,8.94,11.44,8.98,8.34.04,14.86-4.92,15.48-14.94l.05-8.91ZM183.73,118.19c-.05,7.23.28,15.4,4.32,21.54l-37.47-.22c-.91-2.6-1.45-6.5-1.24-9.09h-.38c-7.83,8.85-19.16,11.38-31.04,11.32-18.54-.11-33.88-9.1-33.78-29.32.18-30.42,35.79-29.1,54.54-32.51,5.01-.9,10.03-2.36,10.06-8.48.04-6.49-6.06-8.95-11.99-8.98-11.31-.05-13.58,5.68-13.78,9.75l-34.13-.19c1.26-27.07,26.88-31.38,49.7-31.25,45.99.27,45.52,19.38,45.4,37.92l-.22,39.51Z" /> <path d="M240.33,68.07c-13.16-.07-16.2,12.34-16.27,23.64-.07,11.5,2.83,23.95,15.99,24.03,13.17.07,16.21-12.34,16.28-23.84.07-11.31-2.83-23.76-16-23.83M189.13,6.56h36.72l-.27,47.7h.37c5.61-8.13,15.46-12.9,26.41-12.84,32.26.19,40.82,27.69,40.69,50.32-.14,24.11-13.47,50.92-40.72,50.77-17.99-.11-23.52-7-28.12-13.16h-.37l-.07,10.36h-34.64V6.56Z" /> <path d="M344.59,33.5c-13.92-.08-15.85,13.45-15.91,23.64-.05,10.2,1.72,23.94,15.63,24.03,13.91.07,15.85-13.63,15.91-23.84.06-10.2-1.72-23.75-15.63-23.83M344.15,107.87c-36.17-.21-52.36-24.23-52.19-50.94.16-26.71,16.61-50.35,52.78-50.13,36.17.2,52.36,24.03,52.2,50.75-.16,26.7-16.61,50.54-52.78,50.33" /> <path d="M435.72,33.5c-13.91-.08-15.84,13.45-15.9,23.65-.07,10.2,1.71,23.93,15.62,24.02,13.9.08,15.84-13.63,15.9-23.83.06-10.2-1.72-23.76-15.63-23.83M435.29,107.88c-36.16-.22-52.34-24.24-52.18-50.95.15-26.7,16.61-50.35,52.77-50.14,36.17.21,52.35,24.04,52.2,50.75-.16,26.71-16.62,50.54-52.79,50.34" /> <path d="M492.04,6.8h35.94v132.92h-35.94V6.8Z" /> <path d="M594.92,95.96c-4.46,2.19-9.48,3.46-14.48,4.54-8.19,1.81-12.45,3.83-12.49,10.32-.03,4.45,4.77,8.93,11.44,8.97,8.35.04,14.87-4.92,15.48-14.94l.05-8.89ZM630.26,117.85c-.03,7.23-.44,15.72,3.62,21.86h-36.91c-.9-2.6-1.53-7.04-1.32-9.63h-.37c-7.84,8.85-19.16,11.38-31.03,11.32-18.55-.12-33.89-9.11-33.78-29.33.18-30.42,35.78-29.09,54.53-32.5,5.01-.9,10.03-2.36,10.07-8.48.04-6.49-6.06-8.94-12-8.98-11.32-.06-13.57,5.67-13.77,9.75l-34.15-.19c1.27-27.08,26.9-31.39,49.71-31.24,46,.26,45.52,19.36,45.41,37.91v39.51Z" /> <path d="M305.73,107.88c26.08,11.65,53.13,16.82,82.61,16.99,30.97.18,54.46-5.67,84.19-16.99l-.17,32.08c-26.61,13.2-55.47,20.31-84.23,20.15-31.34-.19-53.93-6.62-82.6-20.15l.19-32.08Z" /> </g> <g> <path d="M745.94,82.84c5.24,6.19,7.87,13.46,7.87,21.8,0,6.82-1.61,12.92-4.83,18.29-3.22,5.37-8.03,9.6-14.41,12.7-6.38,3.1-14.12,4.64-23.22,4.64h-42.27V7.2h41.51c13.27,0,23.28,3.07,30.04,9.19,6.76,6.13,10.14,14.06,10.14,23.79,0,8.47-2.27,15.32-6.82,20.57-4.55,5.25-10.36,8.81-17.44,10.71,7.71,1.39,14.18,5.18,19.43,11.37ZM677.6,67.67h33.36c9.98,0,17.66-2.37,23.03-7.11,5.37-4.74,8.06-11.28,8.06-19.62s-2.65-14.47-7.96-19.15c-5.31-4.67-13.33-7.01-24.07-7.01h-32.41v52.89ZM736.08,125.2c6.13-5.12,9.19-12.16,9.19-21.14s-3.13-15.89-9.38-21.14c-6.26-5.24-14.82-7.87-25.69-7.87h-32.6v57.81h33.17c10.74,0,19.17-2.56,25.31-7.68Z" /> <path d="M775.98,0v140.27h-8.53V0h8.53Z" /> <path d="M815.22,135.34c-7.71-4.29-13.74-10.46-18.1-18.48-4.36-8.02-6.54-17.41-6.54-28.15s2.21-19.93,6.63-27.96c4.42-8.02,10.52-14.15,18.29-18.39,7.77-4.23,16.46-6.35,26.06-6.35s18.26,2.12,25.97,6.35c7.71,4.24,13.74,10.33,18.1,18.29,4.36,7.96,6.54,17.31,6.54,28.05s-2.21,20.13-6.63,28.15c-4.42,8.03-10.52,14.19-18.29,18.48-7.77,4.3-16.46,6.44-26.06,6.44s-18.26-2.15-25.97-6.44ZM862.13,129.18c6.51-3.35,11.72-8.44,15.64-15.26,3.92-6.82,5.88-15.23,5.88-25.21s-1.96-18.2-5.88-25.02c-3.92-6.82-9.1-11.91-15.54-15.26-6.44-3.35-13.4-5.02-20.85-5.02s-14.38,1.68-20.76,5.02c-6.38,3.35-11.53,8.44-15.45,15.26-3.92,6.82-5.88,15.16-5.88,25.02s1.92,18.39,5.78,25.21c3.85,6.82,8.97,11.91,15.35,15.26,6.38,3.35,13.3,5.02,20.76,5.02s14.44-1.67,20.95-5.02Z" /> <path d="M978.81,43.88c7.58,5.25,12.89,11.97,15.92,20.19v-26.73h8.53v106.15c0,9.48-2.08,17.79-6.25,24.93-4.17,7.14-9.79,12.64-16.87,16.49-7.08,3.85-14.97,5.78-23.69,5.78-12.89,0-23.57-3.1-32.04-9.29-8.47-6.19-14.03-14.53-16.68-25.02h8.53c2.53,8.21,7.27,14.75,14.22,19.62,6.95,4.86,15.61,7.3,25.97,7.3,7.2,0,13.71-1.61,19.52-4.83,5.81-3.22,10.39-7.84,13.74-13.84,3.35-6,5.02-13.05,5.02-21.14v-30.14c-3.03,8.22-8.37,15.01-16.02,20.38-7.65,5.37-16.84,8.06-27.58,8.06-9.6,0-18.17-2.18-25.68-6.54-7.52-4.36-13.4-10.55-17.63-18.58-4.24-8.02-6.35-17.34-6.35-27.96s2.11-19.9,6.35-27.86c4.23-7.96,10.11-14.09,17.63-18.39,7.52-4.29,16.08-6.44,25.68-6.44,10.87,0,20.09,2.62,27.68,7.87ZM989.14,64.92c-3.73-6.89-8.81-12.19-15.26-15.92-6.45-3.73-13.65-5.59-21.61-5.59s-15.67,1.8-21.99,5.4c-6.32,3.6-11.25,8.81-14.79,15.64-3.54,6.82-5.31,14.91-5.31,24.26s1.77,17.44,5.31,24.26c3.54,6.82,8.5,12.07,14.88,15.73,6.38,3.67,13.68,5.5,21.89,5.5s15.19-1.86,21.7-5.59c6.51-3.73,11.59-9.03,15.26-15.92,3.66-6.89,5.5-14.88,5.5-23.98s-1.87-16.9-5.59-23.79Z" /> </g> </g> </g> </svg> </a> <div id="desktop-menu" class="pull-left"> <ul id="menu-main-menu" class="menu"><li class="menu-item"> <a class="parent-link" href=" https://blog.taboola.com/category/advertisers-and-marketers/">Advertisers & Marketers</a> <ul class="cj-dropdown-menu mega"><li class="flex-col menu-category-list"><a href="https://blog.taboola.com/category/articles/" class="category-title">Articles</a><div class='flex-row'><ul class='menu-column'><li><a href="https://blog.taboola.com/category/ecommerce/" class="category-link">eCommerce</a></li><li><a href="https://blog.taboola.com/category/affiliate-marketing/" class="category-link">Affiliate Marketing</a></li><li><a href="https://blog.taboola.com/category/landing-pages/" class="category-link">Landing Pages</a></li><li><a href="https://blog.taboola.com/category/lead-generation/" class="category-link">Lead Generation</a></li><li><a href="https://blog.taboola.com/category/marketing-attribution/" class="category-link">Marketing Attribution</a></li><li><a href="https://blog.taboola.com/category/native-advertising/" class="category-link">Native Advertising</a></li><li><a href="https://blog.taboola.com/category/brand-awareness/" class="category-link">Brand Awareness</a></li></ul><ul class='menu-column'><li><a href="https://blog.taboola.com/category/content-marketing/" class="category-link">Content Marketing</a></li><li><a href="https://blog.taboola.com/category/performance-marketing/" class="category-link">Performance Marketing</a></li><li><a href="https://blog.taboola.com/category/taboola-best-practices/" class="category-link">Taboola Best Practices</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/programmatic-marketing/" class="category-link">Programmatic Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/video-marketing/" class="category-link">Video Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/driving-traffic/" class="category-link">Driving Traffic</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/agencies/" class="category-link">Agencies</a></li><li><span class="showMoreCategories">More..</span></li></ul></div></li><li class="flex-col menu-category-list"><a href="https://blog.taboola.com/category/resources/" class="category-title">Resources</a><div class='flex-row'><ul class='menu-column'><li><a href="https://blog.taboola.com/category/research-and-reports/" class="category-link">Research & Reports</a></li><li><a href="https://blog.taboola.com/create-ads-effortlessly/" class="category-link">GenAI Ad Maker</a></li><li><a href="https://trends.taboola.com/" class="category-link">Advertising Trends & Consumer Insights from Taboola Trends</a></li><li><a href="https://trends.taboola.com/headline-analyzer" class="category-link">Headline Title Analyzer</a></li></ul><ul class='menu-column'><li><a href="https://trends.taboola.com/trending-topics" class="category-link">Trending Topics</a></li><li class="moreCategories"><a href="https://trends.taboola.com/image-trends" class="category-link">Trending Images</a></li><li><span class="showMoreCategories">More..</span></li></ul></div></li></ul></li><li class="menu-item"> <a class="parent-link" href=" https://blog.taboola.com/category/publishers/">Publishers</a> <ul class="cj-dropdown-menu "><li><a href="https://blog.taboola.com/category/case-studies/" class="category-link">Case Studies</a></li><li><a href="https://blog.taboola.com/category/data-and-trends/" class="category-link">Data & Trends</a></li><li><a href="https://blog.taboola.com/category/revenue-optimization/" class="category-link">Revenue Optimization</a></li></ul></li><li class="menu-item"> <a class="parent-link" href=" #">Company</a> <ul class="cj-dropdown-menu mega"><li class="flex-col menu-category-list"><a href="https://blog.taboola.com/category/taboola-updates/" class="category-title">Taboola Updates</a><div class='flex-row'><ul class='menu-column'><li><a href="https://blog.taboola.com/category/product-news/" class="category-link">Product News</a></li><li><a href="https://blog.taboola.com/category/taboola-life/" class="category-link">Taboola Life</a></li></ul><ul class='menu-column'><li class="moreCategories"><a href="https://blog.taboola.com/category/taboola-news/" class="category-link">Company News</a></li><li><span class="showMoreCategories">More..</span></li></ul></div></li></ul></li><li class="menu-item"> <a class="parent-link" href=" https://blog.taboola.com/category/engineering/">Engineering</a> <ul class="cj-dropdown-menu "><li><a href="https://blog.taboola.com/category/big-data/" class="category-link">Big Data</a></li><li><a href="https://blog.taboola.com/category/culture/" class="category-link">Culture</a></li><li><a href="https://blog.taboola.com/category/data-science/" class="category-link">Data Science</a></li><li><a href="https://blog.taboola.com/category/java/" class="category-link">Java</a></li><li><a href="https://blog.taboola.com/category/javascript/" class="category-link">Javascript</a></li><li><a href="https://blog.taboola.com/category/machine-learning/" class="category-link">Machine Learning</a></li><li><a href="https://blog.taboola.com/category/system/" class="category-link">System</a></li><li><a href="https://blog.taboola.com/category/tips-and-tricks/" class="category-link">Tips and Tricks</a></li></ul></li></ul> </div> <div id="mobile-menu"> <div class="triger_menu"> <span></span> <span></span> <span></span> </div> <div class="menu_slide"> <div class="menu-scroll-container"> <div class="search"> <form role="search" method="get" class="search-form" action="https://blog.taboola.com/" autocomplete="off"> <label> <span class="screen-reader-text">Search for:</span> <input type="search" class="search-field" placeholder="Search …" value="" name="s"> </label> <input type="submit" class="search-submit" value="Search"> <i class="fa fa-times closed" aria-hidden="true"></i> <div class="search-result" data-result></div> </form> </div> <ul id="mobile-categories-list"><li> <input type="radio" id="tab-0" name="categories" value="0"> <label class="toggleMobileCategoryList" for="tab-0">Advertisers & Marketers<span class="toggle-arrow"></span> </label><ul class="content"><li><a href="https://blog.taboola.com/category/ecommerce/">eCommerce</a></li><li><a href="https://blog.taboola.com/category/affiliate-marketing/">Affiliate Marketing</a></li><li><a href="https://blog.taboola.com/category/landing-pages/">Landing Pages</a></li><li><a href="https://blog.taboola.com/category/lead-generation/">Lead Generation</a></li><li><a href="https://blog.taboola.com/category/marketing-attribution/">Marketing Attribution</a></li><li><a href="https://blog.taboola.com/category/native-advertising/">Native Advertising</a></li><li><a href="https://blog.taboola.com/category/brand-awareness/">Brand Awareness</a></li><li><a href="https://blog.taboola.com/category/content-marketing/">Content Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/performance-marketing/">Performance Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/taboola-best-practices/">Taboola Best Practices</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/programmatic-marketing/">Programmatic Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/video-marketing/">Video Marketing</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/driving-traffic/">Driving Traffic</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/agencies/">Agencies</a></li><li class="moreCategories"><a href="https://blog.taboola.com/category/research-and-reports/">Research & Reports</a></li><li class="moreCategories"><a href="https://blog.taboola.com/create-ads-effortlessly/">GenAI Ad Maker</a></li><li class="moreCategories"><a href="https://trends.taboola.com/">Advertising Trends & Consumer Insights from Taboola Trends</a></li><li class="moreCategories"><a href="https://trends.taboola.com/headline-analyzer">Headline Title Analyzer</a></li><li class="moreCategories"><a href="https://trends.taboola.com/trending-topics">Trending Topics</a></li><li class="moreCategories"><a href="https://trends.taboola.com/image-trends">Trending Images</a></li><li class="showMoreCategories">More..</li></ul></li><li> <input type="radio" id="tab-1" name="categories" value="1"> <label class="toggleMobileCategoryList" for="tab-1">Publishers<span class="toggle-arrow"></span> </label><ul class="content"><li><a href="https://blog.taboola.com/category/case-studies/">Case Studies</a></li><li><a href="https://blog.taboola.com/category/data-and-trends/">Data & Trends</a></li><li><a href="https://blog.taboola.com/category/revenue-optimization/">Revenue Optimization</a></li></ul></li><li> <input type="radio" id="tab-2" name="categories" value="2"> <label class="toggleMobileCategoryList" for="tab-2">Company<span class="toggle-arrow"></span> </label><ul class="content"><li><a href="https://blog.taboola.com/category/product-news/">Product News</a></li><li><a href="https://blog.taboola.com/category/taboola-life/">Taboola Life</a></li><li><a href="https://blog.taboola.com/category/taboola-news/">Company News</a></li></ul></li><li> <input type="radio" id="tab-3" name="categories" value="3"> <label class="toggleMobileCategoryList" for="tab-3">Engineering<span class="toggle-arrow"></span> </label><ul class="content"><li><a href="https://blog.taboola.com/category/big-data/">Big Data</a></li><li><a href="https://blog.taboola.com/category/culture/">Culture</a></li><li><a href="https://blog.taboola.com/category/data-science/">Data Science</a></li><li><a href="https://blog.taboola.com/category/java/">Java</a></li><li><a href="https://blog.taboola.com/category/javascript/">Javascript</a></li><li><a href="https://blog.taboola.com/category/machine-learning/">Machine Learning</a></li><li><a href="https://blog.taboola.com/category/system/">System</a></li><li><a href="https://blog.taboola.com/category/tips-and-tricks/">Tips and Tricks</a></li></ul></li></ul> <a class="taboola-com-btn" href="https://www.taboola.com?int_source=blog_taboola-header-button_A_{Optimizing CTR Prediction through Clustering-Based Feature Engineering}_{taboola.com}" target="_blank" title="Go to Taboola.com !"> www.taboola.com </a> </div> </div> </div> <div class="action-menu pull-right"> <a href="https://www.taboola.com?int_source=blog_taboola-header-button_A_{Optimizing CTR Prediction through Clustering-Based Feature Engineering}_{taboola.com}" title="Go to Taboola.com !" target="_blank" class="go_taboola borderiko">taboola.com</a> <div id="langs"> <div class="earth"> <ul class="lang_ul"> <li><a href="https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" title="English"> English</a></li> <li><a href="https://blog.taboola.com/de/" title="Deutsch"> Deutsch</a></li> <li><a href="https://blog.taboola.com/fr/" title="Français"> Français</a></li> <li><a href="https://blog.taboola.com/ja/" title="日本語"> 日本語</a></li> <li><a href="https://blog.taboola.com/pt/" title="Português"> Português</a></li> <li><a href="https://blog.taboola.com/es/" title="Español"> Español</a></li> <li><a href="https://blog.taboola.com/zh/" title="中文 (台灣)"> 中文 (台灣)</a></li> </ul> </div> </div> <div id="search"> <form role="search" method="get" class="search-form" action="https://blog.taboola.com/" autocomplete="off"> <label> <span class="screen-reader-text">Search for:</span> <input type="search" class="search-field" placeholder="Search …" value="" name="s"> </label> <input type="submit" class="search-submit" value="Search"> <i class="fa fa-times closed" aria-hidden="true"></i> <div class="search-result" data-result></div> </form> </div> </div> </div>  </header>  <script type="text/javascript"> var waitForGlobal = function(key, callback) { if (window[key]) { callback(); } else { setTimeout(function() { waitForGlobal(key, callback); }, 100); } }; </script> <div class="progress-container"> <div class="progress-bar"></div> </div> <div id="content" class="container"> <div class="row "> <div class="col-xs-12 col-md-11 col-centered"> <div class="row row-no-padding"> <div class="col-xs-12 mobile-featured"> <div id="featured" class="image-post-mobile" style="background-image: url(https://blog.taboola.com/wp-content/uploads/2024/05/shutterstock_1918977239.jpg)"></div> </div> </div> <div class="row"> <div class="col-xs-12 col-md-8"> <div class="breadcrumbs"> <ul id="breadcrumbs"><li class="post-category"><a href="https://blog.taboola.com/category/engineering/">Engineering</a></li></ul> </div> </div> </div> <div class="row"> <div class="col-xs-12 col-md-8 "> <header class="post-info"> <div class="post-title"> <div class="col-xs-12"> <div class="row"> <h1 class="mb-20">Optimizing CTR Prediction through Clustering-Based Feature Engineering</h1> </div> </div> </div> <div class="post-info-meta"> <span class="post-by"> Posted by <a href="https://blog.taboola.com/author/ofri-tirosh/" title="Posts by Ofri Tirosh" class="author url fn" rel="author">Ofri Tirosh</a> </span> <i class="middot">   </i> <span class="posted-on"> <time class="entry-date published" datetime="May 21">May 21</time></span> <i class="middot">   </i> 8 Minutes read </div> </header> </div> <div class="col-xs-12 col-md-8 mt-20"> <div class="article-content-wrapper"> <div id="featured" class="image-post-desktop" style="background-image: url(https://blog.taboola.com/wp-content/uploads/2024/05/shutterstock_1918977239.jpg)"></div><p data-pm-slice="1 1 []">With more than 500,000 recommendations per second, Taboola is the world’s leading native advertising platform in the world. When it comes to recommending content, choosing the right features can make a huge impact.</p> <p data-pm-slice="1 1 []">Taboola is working with many publishers, e.g., news, shopping, sports websites, etc. The most basic feature to represent a publisher could be its identifier. Using basic features like publisher-id doesn’t tell us much about what are the user‘s preferences. Furthermore, the sparsity of a feature such as publisher-id requires a lot of data to achieve good results.</p> <p data-pm-slice="1 1 []">Lastly, identifier based features are prone to the cold-start phenomena, in which newly added publishers do not appear in the dataset, and are not familiar to the model yet, and more importantly, they don’t have enough traffic yet for the model to extract other meaningful features.</p> <p>A better approach is to use features that characterize publishers by their articles (i.e., news items) and the users that visit them. Helping the model to better understand the publisher and select a more personalized content effectively, and enhancing user engagement and satisfaction, while improving our models’ performance and generalization.</p> <p>In this blog I will describe the process of data analysis for extraction of meaningful publisher-related features for better CTR prediction.</p> <h3>Getting Cozy with Your Data</h3> <p>This journey begins with a main dataset containing information about items recommended by Taboola with information about user and publisher, over the first week of 2024.</p> <p><img fetchpriority="high" decoding="async" class="aligncenter" src="https://lh7-us.googleusercontent.com/9ybrj05ir3ESoAeMeWP0hNN5ctsOkMvem92oksTRIvMY2TXF2M_4QwCtFZEfZ1MGjPTUqRzrlXOCXvy17zMvUnTUgLV856-XHqeV3HLeTncp3w0KnjQUua6k9YMawC0-azVOvOkzmeQB6xRSESI4IFk" width="602" height="216" /></p> <p><img decoding="async" class="aligncenter" src="https://lh7-us.googleusercontent.com/lZS-BbP4Zn7LNLJKQUKIKFooLk7oNKCmvkRQHPP5wsDp9KGBvz26aPRNSCY2bOHWxKvya1vvOpJqmmCrKaGWVnbSlzoBc3Uzg3q41GC_cye4d-dGT3zEfTXdeZ99fmnGY5LSzL03xYrM_wjQDRELUd0" width="602" height="201" /></p> <p><strong>Table 1:</strong> Example of our main dataset from the 1st day of 2024.</p> <p>The current dataset contains information such as:</p> <ul> <li>Placement: The name of the section where the recommended item appeared within the page.</li> <li>Session Referrer: The domain of the website that has referred the user to the publisher in the current session.</li> <li>Organic Clicks: Amount of user historical organic clicks (i.e., articles clicks).</li> <li>Sponsored Clicks: Amount of user historical sponsored clicks (i.e., ads clicks).</li> <li>Pvs: Amount of page views of the current itemId.</li> <li>Session Depth- number of pages the user passed through until he/she got to the current page view.</li> <li>Clicks: 0/1 for user click/no-click on the item.</li> <li>Source/ Target Upper/Alchemy Taxonomy – The category the article belongs to (i.e., Sports). Source is the current article, Target is the recommended article by taboola.</li> </ul> <p>In addition, I used another complementary dataset for “publisher_common_features”. This dataset provides information about each publisher’s top five common user geolocations, such as country, city, region, and DMA code (for US publishers), and the five most common categories / taxonomies read by the users in that publisher (i.e., sports, entertainment, etc.).</p> <p><img decoding="async" src="https://lh7-us.googleusercontent.com/NACvtm-JqdsVZcln41den-NtqLEgHGx6yjDJlwoz_yc5LH-sCMifUcN4-e-ec0nPjR8Pt2vLboNPvZM11m4WrRHp-IQnKCleyWCK5bSM44cxvK3WL02gJfOh_LTsiaYm2mGbBrR_uAVhBS9kGCbeXas" width="602" height="240" /></p> <p><strong>Table 2:</strong> Example of our complementary data set.</p> <h3>Sampling Snacks from the Dataset Buffet</h3> <p>First, I filtered the main dataset to retain only samples from users that appeared in more than one of the days in the week, and had rc/sp clicks values. On top of this, I sampled only 250K samples from every week day.</p> <p>Once we have a data frame ready we can start handling missing values and remove outliers.</p> <h3>Borrowing Data from the Neighbors’ Yard</h3> <p><script src="https://gist.github.com/ofrizzz/0dd52df37eb05abedfc1a0455466c7b0.js"></script></p> <p>As you can see in the snippet above, there are some missing values in the Country and Region columns, and a significant number of missing values in the Session Referrer column, we will handle these two differently.</p> <p>For the Country and Region, which are both user features, we can infer some of the values by finding the userId in another sample of the dataset and set the Country and Region from that sample to the missing values one. After applying this technique, we had a very small number of missing values, so we could drop those samples.</p> <p>With the Session Referrer feature, the case is different. The missing value in this column, points out that there is no information about the referral of this session, and we can simply set the value in those samples to “OTHER”.</p> <h3>Good Bye Long Tail!</h3> <p>When working with a dataset that is collected based on user sessions online, it is important to remove outliers, for these outliers are mostly bots with no intention of clicking suggested content.</p> <div> <table> <tbody> <tr> <td></td> <td>Sponsored Clicks</td> <td>Organic Clicks</td> </tr> <tr> <td>Before Removal</td> <td><img decoding="async" src="https://lh7-us.googleusercontent.com/mcdY-tQHReC76VB7vD5jrMoFdrwzWh6c0_6tTDHxs_SiS7X4_HWGO2VJ0703KK4PVcctvbINNZyql_wqyaNyUWnBsxww_OLbpET6v46p1uBjM4c6jlz38uz78LmXQ5sQRR5dMM3EARSsr4bA-sK6_Ho" width="343.91519434628975" height="215" /></td> <td><img decoding="async" src="https://lh7-us.googleusercontent.com/ARVgKYQF1_ZxF_FcdV8NIUzlO5pNx-TQquJuPnjZdv1Eb-xCpQY8aEewSufa8BFrl-NSWTbdU87qCp_BCKR_mj7b0qoGrZ2MYm6-1WPO-x0puk6VkbHSCVYguxzW_nh5hGwm0KTNfiIasRnTUEaKxkE" width="341.69879518072287" height="213" /></td> </tr> <tr> <td>After Removal</td> <td><img decoding="async" src="https://lh7-us.googleusercontent.com/3fDrn8TpH0Y9jJfpc8ARsMMJvUAjubZybcnR0C9dM8aP1_GhRf34pNU8xRh8XKlZSVJgnx8yU0ggFRlS5TUAHDz3QHZ2N1jDlo9xufGyjQfXWSEXMgmuqOu0hCTOHzncOaN0yE90mE4OI8aclxvE6Xk" width="338.71232876712327" height="213" /></td> <td> <h1><img decoding="async" src="https://lh7-us.googleusercontent.com/wW-AsCUhvbJlSLLf64lz29yNrJCuk7kcWAm6xSqbcGQuidYMnim1UCFMoqW8p5Qk7sEVRDFaCI-9ZOKU_MoiwutWpiNLtfiaaNKcWRv4rTSkHiBzr9Q67T1jBL1HqhaIK8flha9iSMz2otgUyHu4E9w" width="349.4754098360656" height="219" /></h1> </td> </tr> </tbody> </table> </div> <p><strong>Table 3:</strong> Distribution of rcClicks and spClicks before and after the removal of outliers.</p> <p>In table 3, you can see examples of anomaly removals. In the distribution of rcClicks and spClicks there are long tails that can be removed. By using <a href="https://www.investopedia.com/terms/z/zscore.asp" target="_blank" rel="noopener">z score</a> of more than 20, we got this nice and gradual log-normal distributed curve.</p> <h2>Unearthing Hidden Gems in the Data Mine</h2> <p>As mentioned before, we want to extract some meaningful features describing our publishers.</p> <p>We start by analyzing which properties of an item recommendation are likely to have influence on the behavior of the user (i.e., clicks).</p> <p>One hypothesis was that some impact can come from the following features:</p> <ul> <li>is_weekend*: Binary feature for weekend sessions.</li> </ul> <ul> <li>upper_taxonomy_match*: 0/1 for match between the source and target upper taxonomy**.</li> <li>alchemy_taxonomy_match*: 0/1 for the match between the source and target alchemy taxonomy.</li> <li>in_market: Calculated as 0/1 for a match between the user’s location such as: country, region, city and DMA code</li> </ul> <ul> <li>is_sp_lover/is_rc_lover: if the user had more than 40% rcClicks and spClicks he was considered as sp/rc lovers (i.e., sponsored or organic clicker).</li> <li>is_sp/rc_lovers_publisher: 1 for publishers with more than 50% of the users being sp/rc lovers, 0 otherwise.</li> <li>is_TAXONOMY_publisher: For each TAXONOMY (i.e., article category) we added a binary feature if the TAXONOMY is one of the top 5 common upper taxonomies of the source page</li> <li>common_target_taxonomy: If the traffic that was explained by the users for this taxonomy (i.e., sports) of this publisher was more that 70%, the publisher was considered as this “taxonomy” publisher. otherwise- “None”.</li> </ul> <p>*For our purpose these features will be used as calibration/ context features</p> <p>**Taxonomy is a way of classifying items into groups with common characteristics. In our case, it will be the item/ publisher’s main topics (e.g. news, sports, etc…). Upper taxonomy refers to high level categories, and alchemy refers to more detailed and specific categories.</p> <h2>How Publishers Found Their Soulmates</h2> <p>In order to get some knowledge about how meaningful our features are, we will first use visualizations of our data to see which features show separation in the publishers’ data.</p> <p>After performing one-hot encoding to the categorical features, and scaling the numerical features with Min-Max scaler, we used 2D and 3D PCA for visualizing the distribution of publishers.</p> <p><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/Gc1Cmz_60lLGgEa0JG6OCSMJhtryVykL12RHGsMBvg-FZ8mv13hWqE0Tez3pf9M_3kNYaebyz-aZdwKVEIGfuez3Vjts51tg7GKOajELPqGfe-OTrX4Vkfz2j9aBoQ67ExTSEoqcq61qSvuQpjSfnDg" width="325" height="197" /><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/INr2dL7RoBHpITXvM2w2u_Zf0iDWm5KKwY324oC5m5MyEUqpq7ChBUcWtKV5iOjFqcWRSogpoEalMloJqIsB6FccXMHdPpUK-c_fIidAAe4L_7c9Ty7NvlRtWOUO6ppCrVZ8jMMPgOJnILAOrsXl8w8" width="325" height="203" /></p> <p><strong>Figure 1:</strong> Distribution of publishers representation in 2 and 3 dimensions using PCA.</p> <div></div> <p>Here we can already see some separation with accordance to the publisher taxonomy and</p> <p>whether it is a sponsored lovers publisher.</p> <p>Since the existing publishers’ features do not necessarily define which publishers are similar in the context of CTR prediction, I decided to try representing the publishers using different methods that may better fit our target.</p> <h3>Sharing Users Like Secret Handshakes</h3> <p>As part of our efforts to partition our publishers effectively, we decided to represent our data as a network graph. In this graph, nodes represent publishers, and weighted edges between two nodes represent the number of shared users between the two respective publishers (see example in Figure 2).</p> <p>Through this approach, we aim to find patterns and insights of why certain publishers are grouped together within the network.</p> <p><script src="https://gist.github.com/ofrizzz/06dbf9265f13826ad72a88cc0e70f9c4.js"></script></p> <p>There are some very good techniques to get a partition of nodes in a network. We used the <a href="https://en.wikipedia.org/wiki/Louvain_method" target="_blank" rel="noopener">Louvain-method</a> for community detection.</p> <p>In order to avoid isolated or disconnected nodes and ensure a cohesive network structure, we removed publishers with low connectivity to the rest of the network.</p> <p>After analyzing the partition, it turned out that the feature that defined this clustering is the publisher language, which is an existing feature in our dataset.</p> <div><img loading="lazy" decoding="async" class="aligncenter size-large wp-image-29069" src="https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300-1024x427.png" alt="" width="1024" height="427" srcset="https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300-1024x427.png 1024w, https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300-688x287.png 688w, https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300-206x86.png 206w, https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300-768x320.png 768w, https://blog.taboola.com/wp-content/uploads/2024/05/Screenshot-2024-05-23-070300.png 1182w" sizes="(max-width: 1024px) 100vw, 1024px" /></div> <h3>Aiming at the relevant task – CTR Prediction</h3> <p>I used one of the known ctr prediction models, DeepFM (Guo, H., Tang, R., et al., 2017) from <a href="https://github.com/shenweichen/DeepCTR" target="_blank" rel="noopener">DeepCTR </a>package. This model utilizes factorization machines, as well as deep neural networks to predict item click/non-click based on sparse and dense features.</p> <p>During this process, every categorical feature value is mapped to an embedding space of predefined size.</p> <p>After training the model with our data, I extracted the embeddings of our items and examined their clustering via <a href="https://en.wikipedia.org/wiki/T-distributed_stochastic_neighbor_embedding" target="_blank" rel="noopener">t-SNE</a>.</p> <p>With this embedded space representation of our publishers, I expected that the new publisher features will:</p> <ol> <li>Cluster the items in a different manner.</li> <li>Improve CTR prediction.</li> </ol> <p>Here is an example of how the publishers are distributed in the newly trained embedding space:</p> <p><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/SqSMgVz1cxEGfqlwLQLSRLQUDGOpVaz7AuJJjlG9VhhNuP6gt6RBg1VyuIBBPnEX7IOFarJhGXm0e2_GUOQe2H48HuCgMC3EMUf6FffNCt7x-bfY9AIGyhBTHAe6Rc3SZUSMA_bdauSpysoHSnovz3o" width="350" height="218" /><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/Y_qKDyXtD1ObGkDj3MDsJfoiMULVR7cOs9Ab6ovYjvkWziJNeLWAZiWDuO0mmQtXYy1a4fySWabkVEtReTbg3w-6u86JrKK97y8j_hXpsJSnbFYuSpClToaIBFPSMbGEVWJkXpsA0xmmGlBRjEpIzKU" width="350" height="219" /></p> <p><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/sCvWQY4qfjLU0LAsrNL-uPlQtfRknDKwaCyAI5tGhk4XzcptkyLs63PwH-boc6rN22JoPlP5iWnxBZ9jBgEx4E3DBmA1JtnOeDUH1o06CaD5R8uiXKbnhmzXvvBlKM9pfy7LKG0YSaiB7ydMAuikMwo" width="350" height="218" /><img loading="lazy" decoding="async" src="https://lh7-us.googleusercontent.com/hDDU_ykaBCx_3Qo8erkp-2BqtjguIcd9PIa2xLzw533ILslY7zJdYxLru28umhiqaTnAqZFDxjWhq85AuEDZWZhFl_wbyi86458ThF7GtCNrPRjBnVPKU6cc9WS3afOPqwooOej838gO0yqemCkBYEU" width="350" height="218" /></p> <p><strong>Figure 4:</strong> Illustrating the distribution of publishers in the embedding space. Each point represents a publisher, with its position determined by t-SNE. The plot is color-coded to highlight different features of publishers, such as taxonomy like news or sports and whether large portion their users are sponsored content clickers.</p> <p>As demonstrated in the visualization provided in Figure 4, we observe a clear separation among the data points, in some features (e.g., Sports, News, SP lovers) more than others (e.g., Entertainment), indicating the effectiveness of the meaningful features incorporated into our representation. This separation is a testament to the significance of the features we have included, which capture essential aspects of the data and contribute to its distinct clustering.</p> <h3>The Moment of Truth</h3> <p>To validate my hypothesis, I’ve conducted a comparative analysis between the outcomes of the original model, trained on the original dataset, and those of a modified model trained exclusively on data associated with a specific cluster. Ensuring fairness, all models were trained and tested on an equal number of samples.</p> <p><img loading="lazy" decoding="async" class="aligncenter size-full wp-image-29079" src="https://blog.taboola.com/wp-content/uploads/2024/05/chart.png" alt="" width="955" height="715" srcset="https://blog.taboola.com/wp-content/uploads/2024/05/chart.png 955w, https://blog.taboola.com/wp-content/uploads/2024/05/chart-688x515.png 688w, https://blog.taboola.com/wp-content/uploads/2024/05/chart-206x154.png 206w, https://blog.taboola.com/wp-content/uploads/2024/05/chart-768x575.png 768w" sizes="(max-width: 955px) 100vw, 955px" /></p> <div><strong>Table 4:</strong> Comparison between different models that were trained on specific clusters of publishers.</div> <p>As we expected, the models that were trained and tested with only samples from a well defined cluster, performed significantly better than the one that was trained on the unclustered data.</p> <h3>Separation with Style</h3> <p>One valid suggestion that can be derived from the results in table 4 is to separate the training dataset into dedicated groups of publishers per common category. This can benefit the model, and will eventually lead to better tailored recommendations.</p> <h2>Wrapping It Up: The Part Where We Say: “Bye and Thanks for All the Data!”</h2> <p>In my journey to improve content recommendations, I inspected a data-driven clustering approach, in order to find meaningful publisher features in Taboola’s network. I analyzed user interactions, locations, and content types to find subtle clues that could boost recommendations and keep our users engaged. We discovered interesting patterns among publishers. This opens up exciting possibilities for generating more personalized recommendations.</p> <p>To this end, we obtained an actual improvement in our models’ performance, with notable enhancements observed in key metrics such as AUC and log-loss. These positive outcomes validate the efficacy of our approach and reinforce our confidence in the potential of data-driven clustering to improve content recommendations.</p> <h2>Acknowledgments</h2> <p>This blogpost was created by Ofri Tirosh, and mentored by Taboola mentors<a href="https://www.linkedin.com/in/galikatz/" target="_blank" rel="noopener"> Gali Katz</a> and <a href="http://dorianyitshaki.medium.com" target="_blank" rel="noopener">Dorian Yitzhach</a> as part of the <a href="https://starshipinternship.wixsite.com/site" target="_blank" rel="noopener">Starship Internship Program </a>together with the <a href="https://in.bgu.ac.il/en/engn/sise/Pages/default.aspx" target="_blank" rel="noopener">Software and Information System Engineering Department at Ben-Gurion University</a>.</p> <h2>References</h2> <p>Guo, H., Tang, R., Ye, Y., Li, Z., & He, X. (2017). DeepFM: a factorization-machine based neural network for CTR prediction. <em>arXiv preprint arXiv:1703.04247</em>.</p> <div class="single-post-share-buttons"> <a href="https://www.facebook.com/sharer/sharer.php?u=https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/&t=Optimizing CTR Prediction through Clustering-Based Feature Engineering" title="Facebook Share Button" rel="nofollow noopener" target="_blank" class="share-button facebook"> </a> <a href="https://www.linkedin.com/sharing/share-offsite/?url=https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" title="Linkedin Share Button" rel="nofollow noopener" target="_blank" class="share-button linkedin"> </a> <a href="https://twitter.com/intent/tweet?text=Optimizing CTR Prediction through Clustering-Based Feature Engineering&url=https://blog.taboola.com/optimizing-ctr-prediction-through-clustering-based-feature-engineering/" title="Twitter Share Button" rel="nofollow noopener" target="_blank" class="share-button twitter"> </a> </div> </div> <footer class="post-info "> <span class="posted-on">Originally Published: <time class="entry-date published" datetime="21 May 2024">21 May 2024</time></span> </footer> <div id="strip_campaign"> <div class="contact_form"> <div class="row"> <div class="col-xs-12"> <div class="bg_turquoise strip_campaign_content"> <h4>Start Your Taboola Career Today!</h4> <a class='btn campaign' href=' https://careers.taboola.com?lan=en&int_source=blogpost_mobile-bottom-button_Apply-Today_Optimizing+CTR+Prediction+through+Clustering-Based+Feature+Engineering' target='_blank'>Apply Today</a> </div> </div> </div> </div> </div>  <div class="row feed-taboola"> <div class="col-xs-12"> <script src="//cdn.taboola.com/libtrc/taboolablog-sc/tfa.js"></script> <div id="taboola-below-article-thumbnails"></div> <script type="text/javascript"> window._taboola = window._taboola || []; _taboola.push({ mode: 'organic-thumbnails-a', container: 'taboola-below-article-thumbnails', placement: 'Below Article Thumbnails', target_type: 'mix' }); </script> </div> </div>  </div>  <div id="sidebar" class="col-xs-12 col-md-4 pull-right mt-20"> <div class="right-sidebar-wrapper"> <aside class="widget bg_turquoise"> <div class="sidebar-item"> <div class="campaign-cta-wrapper"> <h4>Interested in a Taboola Career?</h4> <div class=''><a class='btn' href=' https://careers.taboola.com?lan=en&int_source=blogpost_right-rail-button_Apply-Today_Optimizing+CTR+Prediction+through+Clustering-Based+Feature+Engineering' target='_blank'>Apply Today</a></div> </div> </div> </aside> </div> </div> <div class="sticky-mobile-cta"> <div class=''><a class='' href=' https://careers.taboola.com?lan=en&int_source=blogpost_mobile-bottom-button_Apply-Today_Optimizing+CTR+Prediction+through+Clustering-Based+Feature+Engineering' target='_blank'>Interested in a Taboola Career?</a></div></div> </div> </div> </div> </div> <script type="text/javascript"> window.ctaRef = "?int_source=blog_inline-link_A_Optimizing CTR Prediction through Clustering-Based Feature Engineering_"; </script> </div>  <footer id="footer"> <div class="container"> <div class="row"> <style> .widget_footer { width: 50%; } @media only screen and ( max-width: 767px ) { .widget_footer { padding: 0 15px; } } @media only screen and ( min-width: 768px ) { .widget_footer { width: 19% !important; margin-right: 1% !important; } } </style> <div id="fw1" class="widget_footer"> <aside id="nav_menu-2" class="widget widget_nav_menu"><h3 class="widget-title">About Us</h3><div class="menu-about-us-container"><ul id="menu-about-us" class="menu"><li id="menu-item-4242" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-4242"><a target="_blank" rel="noopener" href="https://www.taboola.com/">Taboola Advertising Platform</a></li> <li id="menu-item-28758" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-28758"><a href="https://www.taboola.com/advertise">Advertise</a></li> <li id="menu-item-4243" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-4243"><a target="_blank" rel="noopener" href="https://engineering.taboola.com">Engineering Blog</a></li> <li id="menu-item-4245" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-4245"><a target="_blank" rel="noopener" href="https://www.taboola.com/terms-of-use">Terms of Use</a></li> <li id="menu-item-4246" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-4246"><a target="_blank" rel="noopener" href="https://www.taboola.com/cookie-policy">Cookie Policy</a></li> <li id="menu-item-4247" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-4247"><a target="_blank" rel="noopener" href="https://www.taboola.com/privacy-policy">Privacy Policy</a></li> <li id="menu-item-11635" class="california-only-link hidden menu-item menu-item-type-custom menu-item-object-custom menu-item-11635"><a href="https://ccparequest.taboola.com/">CCPA Notices</a></li> </ul></div></aside> </div> <div id="fw2" class="widget_footer"> <aside id="nav_menu-3" class="widget widget_nav_menu"><h3 class="widget-title">Categories</h3><div class="menu-categories-container"><ul id="menu-categories" class="menu"><li id="menu-item-11316" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11316"><a href="https://blog.taboola.com/category/advertisers-and-marketers/">Advertisers & Marketers</a></li> <li id="menu-item-11318" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11318"><a href="https://blog.taboola.com/category/publishers/">Publishers</a></li> <li id="menu-item-11319" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11319"><a href="https://blog.taboola.com/category/taboola-updates/">Taboola Updates</a></li> <li id="menu-item-25952" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-25952"><a href="https://blog.taboola.com/category/company-news/">Company News</a></li> <li id="menu-item-25953" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-25953"><a href="https://blog.taboola.com/category/audience-engagement/">Audience Engagement</a></li> </ul></div></aside> </div> <div id="fw3" class="widget_footer"> <aside id="nav_menu-4" class="widget widget_nav_menu"><h3 class="widget-title">Topics</h3><div class="menu-topics-container"><ul id="menu-topics" class="menu"><li id="menu-item-16214" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-16214"><a href="https://blog.taboola.com/category/ecommerce/">eCommerce</a></li> <li id="menu-item-16213" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-16213"><a href="https://blog.taboola.com/category/affiliate-marketing/">Affiliate Marketing</a></li> <li id="menu-item-28764" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-28764"><a href="https://blog.taboola.com/category/marketing-attribution/">Marketing Attribution</a></li> <li id="menu-item-11325" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11325"><a href="https://blog.taboola.com/category/native-advertising/">Native Advertising</a></li> <li id="menu-item-11324" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11324"><a href="https://blog.taboola.com/category/lead-generation/">Lead Generation</a></li> <li id="menu-item-11326" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11326"><a href="https://blog.taboola.com/category/performance-marketing/">Performance Marketing</a></li> <li id="menu-item-11320" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11320"><a href="https://blog.taboola.com/category/brand-awareness/">Brand Awareness</a></li> <li id="menu-item-11321" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11321"><a href="https://blog.taboola.com/category/content-marketing/">Content Marketing</a></li> <li id="menu-item-11322" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11322"><a href="https://blog.taboola.com/category/driving-traffic/">Driving Traffic</a></li> <li id="menu-item-11323" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11323"><a href="https://blog.taboola.com/category/landing-pages/">Landing Pages</a></li> <li id="menu-item-11327" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11327"><a href="https://blog.taboola.com/category/taboola-best-practices/">Taboola Best Practices</a></li> <li id="menu-item-11328" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11328"><a href="https://blog.taboola.com/category/video-marketing/">Video Marketing</a></li> <li id="menu-item-11329" class="menu-item menu-item-type-taxonomy menu-item-object-category menu-item-11329"><a href="https://blog.taboola.com/category/revenue-optimization/">Revenue Optimization</a></li> </ul></div></aside> </div> <div id="fw4" class="widget_footer"> <aside id="nav_menu-25" class="widget widget_nav_menu"><h3 class="widget-title">Resources</h3><div class="menu-resources-en-container"><ul id="menu-resources-en" class="menu"><li id="menu-item-26870" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26870"><a href="https://www.taboola.com/native-advertising">Native Advertising</a></li> <li id="menu-item-26871" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26871"><a href="https://blog.taboola.com/effective-native-advertising-examples/">Native Advertising Examples</a></li> <li id="menu-item-27120" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-27120"><a href="https://blog.taboola.com/ecommerce-marketing-and-promotion/">Best eCommerce Marketing Strategies</a></li> <li id="menu-item-28726" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-28726"><a href="https://blog.taboola.com/how-to-grow-your-ecommerce-business/">How to Grow Your eCommerce Business</a></li> <li id="menu-item-28727" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-28727"><a href="https://blog.taboola.com/boost-ecommerce-roi/">Boost Your eCommerce ROI</a></li> <li id="menu-item-25938" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-25938"><a href="https://blog.taboola.com/introductory-guide-performance-marketing/">Guide to Performance Marketing</a></li> <li id="menu-item-26874" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26874"><a href="https://blog.taboola.com/lead-generation-campaign/">Launch a Lead Generation Campaign</a></li> <li id="menu-item-26877" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26877"><a href="https://blog.taboola.com/affiliate-marketing/">How to Start Affiliate Marketing</a></li> <li id="menu-item-27094" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-27094"><a href="https://blog.taboola.com/affiliate-marketing-promotion-methods/">Affiliate Marketing Strategies</a></li> <li id="menu-item-26879" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26879"><a href="https://blog.taboola.com/affiliate-marketing-tools-for-beginners/">Tools for Affiliate Marketing</a></li> <li id="menu-item-26876" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26876"><a href="https://blog.taboola.com/affiliate-marketing-products/">Top Selling Affiliate Products</a></li> <li id="menu-item-25945" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-25945"><a href="https://blog.taboola.com/high-ticket-affiliate-programs-products/">High Ticket Affiliate Marketing</a></li> <li id="menu-item-26949" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26949"><a href="/conversion-rate-optimization-beginners-guide/">Conversion Rate Optimization</a></li> <li id="menu-item-26883" class="menu-item menu-item-type-custom menu-item-object-custom menu-item-26883"><a href="https://blog.taboola.com/google-ads-alternative/">Google Ads Alternatives</a></li> </ul></div></aside> </div> <div id="fw6" class="widget_footer"> <aside id="text-2" class="widget widget_text"><h3 class="widget-title">Let’s Connect</h3> <div class="textwidget"></div> </aside> <aside id="text-4" class="widget widget_text"> <h3 class="widget-title custom"></h3> <div class="textwidget"> <ul class="socials_link"> <li> <a href="https://www.linkedin.com/company/taboola" target="_blank" title="Go to Linkedin page"> <i class="svg-icon svg-icon-linkedin"></i> </a> </li> <li> <a href="https://twitter.com/taboola" target="_blank" title="Go to Twitter page"> <i class="svg-icon svg-icon-twitter"></i> </a> </li> <li> <a href="https://www.facebook.com/taboola/" target="_blank" title="Go to Facebook page"> <i class="svg-icon svg-icon-facebook"></i> </a> </li> </ul> </div> </aside> </div> </div> </div> </footer>  <link rel='stylesheet' id='main-styles-css' href='https://blog.taboola.com/wp-content/themes/taboola-new/dist/css/main.min.css?ver=6.6.2' type='text/css' media='all' /> <link rel='stylesheet' id='page-styles-css' href='https://blog.taboola.com/wp-content/themes/taboola-new/dist/css/single.min.css?ver=6.6.2' type='text/css' media='all' /> <script type="text/javascript" src="https://blog.taboola.com/wp-content/themes/taboola-new/dist/js/main.min.js" id="taboola-main-js-js"></script> <script type="text/javascript" src="https://blog.taboola.com/wp-content/themes/taboola-new/dist/js/single.min.js" id="taboola-page-scripts-js"></script> <script type="text/javascript"> geoFunctions(function() { if (isCountryCodeInList([ "US", "CA", "GB", "AT", "FR", "ZA", "DE", "CH", "NZ", "ES", "AU", "IE", "BE", "LU", "MC", "MX", "PL", "NL", "BR", "JP", "IS", "NO", "SE", "FI", "DK", "IN", "IL" ])) { // Show in case of segmentation $(".segmented-row .nonsegmented-cta").removeClass("hidden"); } else { $(".segmented-row .segmented-cta").removeClass("hidden"); } // California if (getCountryCode() == "US" && getState() == "CA") { $(".california-only-link").removeClass("hidden"); } }); $(".nonsegmented-row .nonsegmented-cta").removeClass("hidden"); </script> </body> </html>

CINXE.COM

Optimizing CTR Prediction through Clustering-Based Feature Engineering