<!DOCTYPE html><html><head><title>DASE Components Explained (Recommendation)</title><meta charset="utf-8"/><meta content="IE=edge,chrome=1" http-equiv="X-UA-Compatible"/><meta name="viewport" content="width=device-width, initial-scale=1.0"/><meta class="swiftype" name="title" data-type="string" content="DASE Components Explained (Recommendation)"/><link rel="canonical" href="https://docs.prediction.io/templates/recommendation/dase/"/><link href="/images/favicon/normal-b330020a.png" rel="shortcut icon"/><link href="/images/favicon/apple-c0febcf2.png" rel="apple-touch-icon"/><link href="//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet"/><link href="//maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet"/><link href="/stylesheets/application-3598c7d7.css" rel="stylesheet" type="text/css"/><!--[if lt IE 9]><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min.js"></script><![endif]--><script src="//cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script>(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
})(window,document,'script','//www.google-analytics.com/analytics.js','ga');

ga('create', 'UA-38306178-1', 'auto');
ga('require', 'linkid', 'linkid.js');
ga('send', 'pageview');</script><script>!function(){var analytics=window.analytics=window.analytics||[];if(!analytics.initialize)if(analytics.invoked)window.console&&console.error&&console.error("Segment snippet included twice.");else{analytics.invoked=!0;analytics.methods=["trackSubmit","trackClick","trackLink","trackForm","pageview","identify","group","track","ready","alias","page","once","off","on"];analytics.factory=function(t){return function(){var e=Array.prototype.slice.call(arguments);e.unshift(t);analytics.push(e);return analytics}};for(var t=0;t<analytics.methods.length;t++){var e=analytics.methods[t];analytics[e]=analytics.factory(e)}analytics.load=function(t){var e=document.createElement("script");e.type="text/javascript";e.async=!0;e.src=("https:"===document.location.protocol?"https://":"http://")+"cdn.segment.com/analytics.js/v1/"+t+"/analytics.min.js";var n=document.getElementsByTagName("script")[0];n.parentNode.insertBefore(e,n)};analytics.SNIPPET_VERSION="3.0.1";
  analytics.load("YlF3updaI3DR96hnNgSGpR3PPBUGDzt8");
  analytics.page()
  }}();</script><script>RCX_CUSTOM_LIB="https://cdn.recontext.com/staging/rcx.min.js";
(function(b,d,a){b.RCX_OBJECT=a;a=b[a]||[];if(!a.snipV&&!a.libV){b.rcx=a;a.snipV="0.2.0";var g=function(a,b,c,d){a[b]=a[b]||function(){c.push([d].concat(Array.prototype.slice.call(arguments)))}};b="init page track identify link setUserProperty unsetUserProperty".split(" ");for(var f=0;f<b.length;f++){var e,c;e=b[f];c=e.split(".");2==c.length?(a[c[0]]=a[c[0]]||[],g(a[c[0]],c[1],a,e)):g(a,e,a,e)}a=d.createElement("script");a.type="text/javascript";a.async=!0;a.src="undefined"!==typeof RCX_CUSTOM_LIB?
RCX_CUSTOM_LIB:"https://cdn.recontext.com/rcx.min.js";d=d.getElementsByTagName("script")[0];d.parentNode.insertBefore(a,d)}})(window,document,"rcx");
rcx.init("kTxFcI3IWdXYfRsh6uuYuej4qYl8m8LVMePM2hdIkM9YjHqkAFC6mqdqO9fpp8p9");
rcx.page();</script><script>function t(e){analytics.identify(e); analytics.track("newsletter signup");
  rcx.track("newsletter signup", { '_email': e });}</script><script>!function(f,b,e,v,n,t,s){if(f.fbq)return;n=f.fbq=function(){n.callMethod?
n.callMethod.apply(n,arguments):n.queue.push(arguments)};if(!f._fbq)f._fbq=n;
n.push=n;n.loaded=!0;n.version='2.0';n.queue=[];t=b.createElement(e);t.async=!0;
t.src=v;s=b.getElementsByTagName(e)[0];s.parentNode.insertBefore(t,s)}(window,
document,'script','//connect.facebook.net/en_US/fbevents.js');

fbq('init', '1073028432707778');
fbq('track', "PageView");</script><script src="//use.typekit.net/mut4mjx.js"></script><script>try{Typekit.load();}catch(e){}</script></head><body><div id="global"><header><div class="container" id="header-wrapper"><div class="row"><div class="col-sm-12"><div id="logo-wrapper"><span id="drawer-toggle"></span><a href="#"></a><a href="http://prediction.io/"><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a></div><div id="menu-wrapper"><div id="header-nav-options-wrapper"><ul><li><a href="/">Install & Doc</a></li> <li><a href="/support">Support</a></li> </ul></div><div id="pill-wrapper"><a class="pill left" href="//templates.prediction.io/">TEMPLATES</a> <a class="pill right" href="//github.com/PredictionIO/PredictionIO/">OPEN SOURCE</a></div></div><img class="mobile-search-bar-toggler hidden-md hidden-lg" src="/images/icons/search-glass-704bd4ff.png"/></div></div></div></header><div id="search-bar-row-wrapper"><div class="container-fluid" id="search-bar-row"><div class="row"><div class="col-md-9 col-sm-11 col-xs-11"><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>DASE Components Explained (Recommendation)</h4></div><h4 class="hidden-sm hidden-xs">PredictionIO Docs</h4></div><div class="col-md-3 col-sm-1 col-xs-1 hidden-md hidden-lg"><img id="left-menu-indicator" src="/images/icons/down-arrow-dfe9f7fe.png"/></div><div class="col-md-3 col-sm-12 col-xs-12 swiftype-wrapper"><div class="swiftype"><form class="search-form"><img class="search-box-toggler hidden-xs hidden-sm" src="/images/icons/search-glass-704bd4ff.png"/><div class="search-box"><img src="/images/icons/search-glass-704bd4ff.png"/><input type="text" id="st-search-input" class="st-search-input" placeholder="Search Doc..."/></div><img class="swiftype-row-hider hidden-md hidden-lg" src="/images/icons/drawer-toggle-active-fcbef12a.png"/></form></div></div><div class="mobile-left-menu-toggler hidden-md hidden-lg"></div></div></div></div><div id="page" class="container-fluid"><div class="row"><div id="left-menu-wrapper" class="col-md-3"><nav id="nav-main"><ul><li class="level-1"><a class="expandible" href="/"><span>Apache PredictionIO (incubating) Documentation</span></a><ul><li class="level-2"><a class="final" href="/"><span>Welcome to Apache PredictionIO (incubating)</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Started</span></a><ul><li class="level-2"><a class="final" href="/start/"><span>A Quick Intro</span></a></li><li class="level-2"><a class="final" href="/install/"><span>Installing Apache PredictionIO (incubating)</span></a></li><li class="level-2"><a class="final" href="/start/download/"><span>Downloading an Engine Template</span></a></li><li class="level-2"><a class="final" href="/start/deploy/"><span>Deploying Your First Engine</span></a></li><li class="level-2"><a class="final" href="/start/customize/"><span>Customizing the Engine</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Integrating with Your App</span></a><ul><li class="level-2"><a class="final" href="/appintegration/"><span>App Integration Overview</span></a></li><li class="level-2"><a class="expandible" href="/sdk/"><span>List of SDKs</span></a><ul><li class="level-3"><a class="final" href="/sdk/java/"><span>Java & Android SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/php/"><span>PHP SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/python/"><span>Python SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/ruby/"><span>Ruby SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/community/"><span>Community Powered SDKs</span></a></li></ul></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Deploying an Engine</span></a><ul><li class="level-2"><a class="final" href="/deploy/"><span>Deploying as a Web Service</span></a></li><li class="level-2"><a class="final" href="/cli/#engine-commands"><span>Engine Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/deploy/engineparams/"><span>Setting Engine Parameters</span></a></li><li class="level-2"><a class="final" href="/deploy/enginevariants/"><span>Deploying Multiple Engine Variants</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Customizing an Engine</span></a><ul><li class="level-2"><a class="final" href="/customize/"><span>Learning DASE</span></a></li><li class="level-2"><a class="final" href="/customize/dase/"><span>Implement DASE</span></a></li><li class="level-2"><a class="final" href="/customize/troubleshooting/"><span>Troubleshooting Engine Development</span></a></li><li class="level-2"><a class="final" href="/api/current/#package"><span>Engine Scala APIs</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Collecting and Analyzing Data</span></a><ul><li class="level-2"><a class="final" href="/datacollection/"><span>Event Server Overview</span></a></li><li class="level-2"><a class="final" href="/cli/#event-server-commands"><span>Event Server Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventapi/"><span>Collecting Data with REST/SDKs</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventmodel/"><span>Events Modeling</span></a></li><li class="level-2"><a class="final" href="/datacollection/webhooks/"><span>Unifying Multichannel Data with Webhooks</span></a></li><li class="level-2"><a class="final" href="/datacollection/channel/"><span>Channel</span></a></li><li class="level-2"><a class="final" href="/datacollection/batchimport/"><span>Importing Data in Batch</span></a></li><li class="level-2"><a class="final" href="/datacollection/analytics/"><span>Using Analytics Tools</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Choosing an Algorithm(s)</span></a><ul><li class="level-2"><a class="final" href="/algorithm/"><span>Built-in Algorithm Libraries</span></a></li><li class="level-2"><a class="final" href="/algorithm/switch/"><span>Switching to Another Algorithm</span></a></li><li class="level-2"><a class="final" href="/algorithm/multiple/"><span>Combining Multiple Algorithms</span></a></li><li class="level-2"><a class="final" href="/algorithm/custom/"><span>Adding Your Own Algorithms</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>ML Tuning and Evaluation</span></a><ul><li class="level-2"><a class="final" href="/evaluation/"><span>Overview</span></a></li><li class="level-2"><a class="final" href="/evaluation/paramtuning/"><span>Hyperparameter Tuning</span></a></li><li class="level-2"><a class="final" href="/evaluation/evaluationdashboard/"><span>Evaluation Dashboard</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricchoose/"><span>Choosing Evaluation Metrics</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricbuild/"><span>Building Evaluation Metrics</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>System Architecture</span></a><ul><li class="level-2"><a class="final" href="/system/"><span>Architecture Overview</span></a></li><li class="level-2"><a class="final" href="/system/anotherdatastore/"><span>Using Another Data Store</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Engine Template Gallery</span></a><ul><li class="level-2"><a class="final" href="http://templates.prediction.io"><span>Browse</span></a></li><li class="level-2"><a class="final" href="/community/submit-template/"><span>Submit your Engine as a Template</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Demo Tutorials</span></a><ul><li class="level-2"><a class="final" href="/demo/tapster/"><span>Comics Recommendation Demo</span></a></li><li class="level-2"><a class="final" href="/demo/community/"><span>Community Contributed Demo</span></a></li><li class="level-2"><a class="final" href="/demo/textclassification/"><span>Text Classification Engine Tutorial</span></a></li></ul></li><li class="level-1"><a class="expandible" href="/community/"><span>Getting Involved</span></a><ul><li class="level-2"><a class="final" href="/community/contribute-code/"><span>Contribute Code</span></a></li><li class="level-2"><a class="final" href="/community/contribute-documentation/"><span>Contribute Documentation</span></a></li><li class="level-2"><a class="final" href="/community/contribute-sdk/"><span>Contribute a SDK</span></a></li><li class="level-2"><a class="final" href="/community/contribute-webhook/"><span>Contribute a Webhook</span></a></li><li class="level-2"><a class="final" href="/community/projects/"><span>Community Projects</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Help</span></a><ul><li class="level-2"><a class="final" href="/resources/faq/"><span>FAQs</span></a></li><li class="level-2"><a class="final" href="/support/"><span>Community Support</span></a></li><li class="level-2"><a class="final" href="/support/#enterprise-support"><span>Enterprise Support</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Resources</span></a><ul><li class="level-2"><a class="final" href="/resources/intellij/"><span>Developing Engines with IntelliJ IDEA</span></a></li><li class="level-2"><a class="final" href="/resources/upgrade/"><span>Upgrade Instructions</span></a></li><li class="level-2"><a class="final" href="/resources/glossary/"><span>Glossary</span></a></li></ul></li></ul></nav></div><div class="col-md-9 col-sm-12"><div class="content-header hidden-md hidden-lg"><div id="page-title"><h1>DASE Components Explained (Recommendation)</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#the-engine-design">The Engine Design</a> </li> <li> <a href="#data">Data</a> </li> <li> <a href="#algorithm">Algorithm</a> </li> <li> <a href="#serving">Serving</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/tree/livedoc/docs/manual/source/templates/recommendation/dase.html.md.erb"><img src="/images/icons/edit-pencil-d6c1bb3d.png"/>Edit this page</a></div><div class="content-header hidden-sm hidden-xs"><div id="page-title"><h1>DASE Components Explained (Recommendation)</h1></div></div><div class="content"><p>PredictionIO&#39;s DASE architecture brings the separation-of-concerns design principle to predictive engine development. DASE stands for the following components of an engine:</p> <ul> <li><strong>D</strong>ata - includes Data Source and Data Preparator</li> <li><strong>A</strong>lgorithm(s)</li> <li><strong>S</strong>erving</li> <li><strong>E</strong>valuator</li> </ul> <p><p>Let&#39;s look at the code and see how you can customize the engine you built from the Recommendation Engine Template.</p><div class="alert-message note"><p>Evaluator will not be covered in this tutorial. Please visit <a href="/templates/recommendation/evaluation/">evaluation explained</a> for using evaluation.</p></div></p><h2 id='the-engine-design' class='header-anchors'>The Engine Design</h2><p>As you can see from the Quick Start, <em>MyRecommendation</em> takes a JSON prediction query, e.g. <code>{ &quot;user&quot;: &quot;1&quot;, &quot;num&quot;: 4 }</code>, and return a JSON predicted result. In MyRecommendation/src/main/scala/<strong><em>Engine.scala</em></strong>, the <code>Query</code> case class defines the format of such <strong>query</strong>:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4</pre></td><td class="code"><pre><span class="k">case</span> <span class="k">class</span> <span class="nc">Query</span><span class="o">(</span>
  <span class="n">user</span><span class="k">:</span> <span class="kt">String</span><span class="o">,</span>
  <span class="n">num</span><span class="k">:</span> <span class="kt">Int</span>
<span class="o">)</span> <span class="k">extends</span> <span class="nc">Serializable</span>
</pre></td></tr></tbody></table> </div> <p>The <code>PredictedResult</code> case class defines the format of <strong>predicted result</strong>, such as</p><div class="highlight json"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6</pre></td><td class="code"><pre><span class="p">{</span><span class="s2">"itemScores"</span><span class="p">:[</span><span class="w">
  </span><span class="p">{</span><span class="s2">"item"</span><span class="p">:</span><span class="mi">22</span><span class="p">,</span><span class="s2">"score"</span><span class="p">:</span><span class="mf">4.07</span><span class="p">},</span><span class="w">
  </span><span class="p">{</span><span class="s2">"item"</span><span class="p">:</span><span class="mi">62</span><span class="p">,</span><span class="s2">"score"</span><span class="p">:</span><span class="mf">4.05</span><span class="p">},</span><span class="w">
  </span><span class="p">{</span><span class="s2">"item"</span><span class="p">:</span><span class="mi">75</span><span class="p">,</span><span class="s2">"score"</span><span class="p">:</span><span class="mf">4.04</span><span class="p">},</span><span class="w">
  </span><span class="p">{</span><span class="s2">"item"</span><span class="p">:</span><span class="mi">68</span><span class="p">,</span><span class="s2">"score"</span><span class="p">:</span><span class="mf">3.81</span><span class="p">}</span><span class="w">
</span><span class="p">]}</span><span class="w">
</span></pre></td></tr></tbody></table> </div> <p>with:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8</pre></td><td class="code"><pre><span class="k">case</span> <span class="k">class</span> <span class="nc">PredictedResult</span><span class="o">(</span>
  <span class="n">itemScores</span><span class="k">:</span> <span class="kt">Array</span><span class="o">[</span><span class="kt">ItemScore</span><span class="o">]</span>
<span class="o">)</span> <span class="k">extends</span> <span class="nc">Serializable</span>

<span class="k">case</span> <span class="k">class</span> <span class="nc">ItemScore</span><span class="o">(</span>
  <span class="n">item</span><span class="k">:</span> <span class="kt">String</span><span class="o">,</span>
  <span class="n">score</span><span class="k">:</span> <span class="kt">Double</span>
<span class="o">)</span> <span class="k">extends</span> <span class="nc">Serializable</span>
</pre></td></tr></tbody></table> </div> <p>Finally, <code>RecommendationEngine</code> is the <em>Engine Factory</em> that defines the components this engine will use: Data Source, Data Preparator, Algorithm(s) and Serving components.</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10</pre></td><td class="code"><pre><span class="k">object</span> <span class="nc">RecommendationEngine</span> <span class="k">extends</span> <span class="nc">IEngineFactory</span> <span class="o">{</span>
  <span class="k">def</span> <span class="n">apply</span><span class="o">()</span> <span class="k">=</span> <span class="o">{</span>
    <span class="k">new</span> <span class="nc">Engine</span><span class="o">(</span>
      <span class="n">classOf</span><span class="o">[</span><span class="kt">DataSource</span><span class="o">],</span>
      <span class="n">classOf</span><span class="o">[</span><span class="kt">Preparator</span><span class="o">],</span>
      <span class="nc">Map</span><span class="o">(</span><span class="s">"als"</span> <span class="o">-&gt;</span> <span class="n">classOf</span><span class="o">[</span><span class="kt">ALSAlgorithm</span><span class="o">]),</span>
      <span class="n">classOf</span><span class="o">[</span><span class="kt">Serving</span><span class="o">])</span>
  <span class="o">}</span>
  <span class="o">...</span>
<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <h3 id='spark-mllib' class='header-anchors'>Spark MLlib</h3><p>Spark&#39;s MLlib ALS algorithm takes training data of RDD type, i.e. <code>RDD[Rating]</code> and train a model, which is a <code>MatrixFactorizationModel</code> object.</p><p>PredictionIO Recommendation Engine Template, which <em>MyRecommendation</em> bases on, integrates this algorithm under the DASE architecture. We will take a closer look at the DASE code below.</p><div class="alert-message info"><p><a href="https://spark.apache.org/docs/latest/mllib-collaborative-filtering.html">Check this out</a> to learn more about MLlib&#39;s ALS collaborative filtering algorithm.</p></div><h2 id='data' class='header-anchors'>Data</h2><p>In the DASE architecture, data is prepared by 2 components sequentially: <em>Data Source</em> and <em>Data Preparator</em>. <em>Data Source</em> and <em>Data Preparator</em> takes data from the data store and prepares <code>RDD[Rating]</code> for the ALS algorithm.</p><h3 id='data-source' class='header-anchors'>Data Source</h3><p>In MyRecommendation/src/main/scala/<strong><em>DataSource.scala</em></strong>, the <code>readTraining</code> method of class <code>DataSource</code> reads, and selects, data from the <em>Event Store</em> (data store of the <em>Event Server</em>) and returns <code>TrainingData</code>.</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46</pre></td><td class="code"><pre><span class="k">case</span> <span class="k">class</span> <span class="nc">DataSourceParams</span><span class="o">(</span><span class="n">appName</span><span class="k">:</span> <span class="kt">String</span><span class="o">)</span> <span class="k">extends</span> <span class="nc">Params</span>

<span class="k">class</span> <span class="nc">DataSource</span><span class="o">(</span><span class="k">val</span> <span class="n">dsp</span><span class="k">:</span> <span class="kt">DataSourceParams</span><span class="o">)</span>
  <span class="k">extends</span> <span class="nc">PDataSource</span><span class="o">[</span><span class="kt">TrainingData</span>,
      <span class="kt">EmptyEvaluationInfo</span>, <span class="kt">Query</span>, <span class="kt">EmptyActualResult</span><span class="o">]</span> <span class="o">{</span>

  <span class="nd">@transient</span> <span class="k">lazy</span> <span class="k">val</span> <span class="n">logger</span> <span class="k">=</span> <span class="nc">Logger</span><span class="o">[</span><span class="kt">this.</span><span class="k">type</span><span class="o">]</span>

  <span class="k">def</span> <span class="n">getRatings</span><span class="o">(</span><span class="n">sc</span><span class="k">:</span> <span class="kt">SparkContext</span><span class="o">)</span><span class="k">:</span> <span class="kt">RDD</span><span class="o">[</span><span class="kt">Rating</span><span class="o">]</span> <span class="k">=</span> <span class="o">{</span>

    <span class="k">val</span> <span class="n">eventsRDD</span><span class="k">:</span> <span class="kt">RDD</span><span class="o">[</span><span class="kt">Event</span><span class="o">]</span> <span class="k">=</span> <span class="nc">PEventStore</span><span class="o">.</span><span class="n">find</span><span class="o">(</span>
      <span class="n">appName</span> <span class="k">=</span> <span class="n">dsp</span><span class="o">.</span><span class="n">appName</span><span class="o">,</span>
      <span class="n">entityType</span> <span class="k">=</span> <span class="nc">Some</span><span class="o">(</span><span class="s">"user"</span><span class="o">),</span>
      <span class="n">eventNames</span> <span class="k">=</span> <span class="nc">Some</span><span class="o">(</span><span class="nc">List</span><span class="o">(</span><span class="s">"rate"</span><span class="o">,</span> <span class="s">"buy"</span><span class="o">)),</span> <span class="c1">// read "rate" and "buy" event
</span>      <span class="c1">// targetEntityType is optional field of an event.
</span>      <span class="n">targetEntityType</span> <span class="k">=</span> <span class="nc">Some</span><span class="o">(</span><span class="nc">Some</span><span class="o">(</span><span class="s">"item"</span><span class="o">)))(</span><span class="n">sc</span><span class="o">)</span>

    <span class="k">val</span> <span class="n">ratingsRDD</span><span class="k">:</span> <span class="kt">RDD</span><span class="o">[</span><span class="kt">Rating</span><span class="o">]</span> <span class="k">=</span> <span class="n">eventsRDD</span><span class="o">.</span><span class="n">map</span> <span class="o">{</span> <span class="n">event</span> <span class="k">=&gt;</span>
      <span class="k">val</span> <span class="n">rating</span> <span class="k">=</span> <span class="k">try</span> <span class="o">{</span>
        <span class="k">val</span> <span class="n">ratingValue</span><span class="k">:</span> <span class="kt">Double</span> <span class="o">=</span> <span class="n">event</span><span class="o">.</span><span class="n">event</span> <span class="k">match</span> <span class="o">{</span>
          <span class="k">case</span> <span class="s">"rate"</span> <span class="k">=&gt;</span> <span class="n">event</span><span class="o">.</span><span class="n">properties</span><span class="o">.</span><span class="n">get</span><span class="o">[</span><span class="kt">Double</span><span class="o">](</span><span class="s">"rating"</span><span class="o">)</span>
          <span class="k">case</span> <span class="s">"buy"</span> <span class="k">=&gt;</span> <span class="mf">4.0</span> <span class="c1">// map buy event to rating value of 4
</span>          <span class="k">case</span> <span class="k">_</span> <span class="k">=&gt;</span> <span class="k">throw</span> <span class="k">new</span> <span class="nc">Exception</span><span class="o">(</span><span class="n">s</span><span class="s">"Unexpected event ${event} is read."</span><span class="o">)</span>
        <span class="o">}</span>
        <span class="c1">// entityId and targetEntityId is String
</span>        <span class="nc">Rating</span><span class="o">(</span><span class="n">event</span><span class="o">.</span><span class="n">entityId</span><span class="o">,</span>
          <span class="n">event</span><span class="o">.</span><span class="n">targetEntityId</span><span class="o">.</span><span class="n">get</span><span class="o">,</span>
          <span class="n">ratingValue</span><span class="o">)</span>
      <span class="o">}</span> <span class="k">catch</span> <span class="o">{</span>
        <span class="k">case</span> <span class="n">e</span><span class="k">:</span> <span class="kt">Exception</span> <span class="o">=&gt;</span> <span class="o">{</span>
          <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="o">(</span><span class="n">s</span><span class="s">"Cannot convert ${event} to Rating. Exception: ${e}."</span><span class="o">)</span>
          <span class="k">throw</span> <span class="n">e</span>
        <span class="o">}</span>
      <span class="o">}</span>
      <span class="n">rating</span>
    <span class="o">}.</span><span class="n">cache</span><span class="o">()</span>

    <span class="n">ratingsRDD</span>
  <span class="o">}</span>

  <span class="k">override</span>
  <span class="k">def</span> <span class="n">readTraining</span><span class="o">(</span><span class="n">sc</span><span class="k">:</span> <span class="kt">SparkContext</span><span class="o">)</span><span class="k">:</span> <span class="kt">TrainingData</span> <span class="o">=</span> <span class="o">{</span>
    <span class="k">new</span> <span class="nc">TrainingData</span><span class="o">(</span><span class="n">getRatings</span><span class="o">(</span><span class="n">sc</span><span class="o">))</span>
  <span class="o">}</span>

<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p><code>PEventStore</code> is an object which provides function to access data that is collected by PredictionIO <em>Event Server</em>. <code>PEventStore.find(...)</code> specifies the events that you want to read. PredictionIO automatically loads the parameters of <em>datasource</em> specified in MyRecommendation/<strong><em>engine.json</em></strong>, including <em>appName</em>, to <code>dsp</code>.</p><p>In <strong><em>engine.json</em></strong>:</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9</pre></td><td class="code"><pre><span class="o">{</span>
  ...
  <span class="s2">"datasource"</span>: <span class="o">{</span>
    <span class="s2">"params"</span> : <span class="o">{</span>
      <span class="s2">"appName"</span>: <span class="s2">"MyApp1"</span>
    <span class="o">}</span>
  <span class="o">}</span>,
  ...
<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p>Each <em>rate</em> and <em>buy</em> user event data is read as <code>Rating</code>. For flexibility, this Recommendation engine template is designed to support user ID and item ID in <code>String</code>. Since Spark MLlib&#39;s <code>Rating</code> class assumes <code>Int</code>-only user ID and item ID, you have to define a new <code>Rating</code> class:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5</pre></td><td class="code"><pre><span class="k">case</span> <span class="k">class</span> <span class="nc">Rating</span><span class="o">(</span>
  <span class="n">user</span><span class="k">:</span> <span class="kt">String</span><span class="o">,</span>
  <span class="n">item</span><span class="k">:</span> <span class="kt">String</span><span class="o">,</span>
  <span class="n">rating</span><span class="k">:</span> <span class="kt">Double</span>
<span class="o">)</span>
</pre></td></tr></tbody></table> </div> <p><code>TrainingData</code> contains an RDD of all these <code>Rating</code> events. The class definition of <code>TrainingData</code> is:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3</pre></td><td class="code"><pre><span class="k">class</span> <span class="nc">TrainingData</span><span class="o">(</span>
  <span class="k">val</span> <span class="n">ratings</span><span class="k">:</span> <span class="kt">RDD</span><span class="o">[</span><span class="kt">Rating</span><span class="o">]</span>
<span class="o">)</span> <span class="k">extends</span> <span class="nc">Serializable</span> <span class="o">{...}</span>
</pre></td></tr></tbody></table> </div> <p>and PredictionIO passes the returned <code>TrainingData</code> object to <em>Data Preparator</em>.</p> <div class="alert-message info"><p>You could <a href="/templates/recommendation/reading-custom-events/">modify the DataSource to read custom events</a> other than the default <strong>rate</strong> and <strong>buy</strong>.</p></div><h3 id='data-preparator' class='header-anchors'>Data Preparator</h3><p>In MyRecommendation/src/main/scala/<strong><em>Preparator.scala</em></strong>, the <code>prepare</code> method of class <code>Preparator</code> takes <code>TrainingData</code> as its input and performs any necessary feature selection and data processing tasks. At the end, it returns <code>PreparedData</code> which should contain the data <em>Algorithm</em> needs. For MLlib ALS, it is <code>RDD[Rating]</code>.</p><p>By default, <code>prepare</code> simply copies the unprocessed <code>TrainingData</code> data to <code>PreparedData</code>:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10
11</pre></td><td class="code"><pre><span class="k">class</span> <span class="nc">Preparator</span>
  <span class="k">extends</span> <span class="nc">PPreparator</span><span class="o">[</span><span class="kt">TrainingData</span>, <span class="kt">PreparedData</span><span class="o">]</span> <span class="o">{</span>

  <span class="k">def</span> <span class="n">prepare</span><span class="o">(</span><span class="n">sc</span><span class="k">:</span> <span class="kt">SparkContext</span><span class="o">,</span> <span class="n">trainingData</span><span class="k">:</span> <span class="kt">TrainingData</span><span class="o">)</span><span class="k">:</span> <span class="kt">PreparedData</span> <span class="o">=</span> <span class="o">{</span>
    <span class="k">new</span> <span class="nc">PreparedData</span><span class="o">(</span><span class="n">ratings</span> <span class="k">=</span> <span class="n">trainingData</span><span class="o">.</span><span class="n">ratings</span><span class="o">)</span>
  <span class="o">}</span>
<span class="o">}</span>

<span class="k">class</span> <span class="nc">PreparedData</span><span class="o">(</span>
  <span class="k">val</span> <span class="n">ratings</span><span class="k">:</span> <span class="kt">RDD</span><span class="o">[</span><span class="kt">Rating</span><span class="o">]</span>
<span class="o">)</span> <span class="k">extends</span> <span class="nc">Serializable</span>
</pre></td></tr></tbody></table> </div> <p>PredictionIO passes the returned <code>PreparedData</code> object to Algorithm&#39;s <code>train</code> function.</p> <h2 id='algorithm' class='header-anchors'>Algorithm</h2><p>In MyRecommendation/src/main/scala/<strong><em>ALSAlgorithm.scala</em></strong>, the two methods of the algorithm class are <code>train</code> and <code>predict</code>. <code>train</code> is responsible for training a predictive model. PredictionIO will store this model and <code>predict</code> is responsible for using this model to make prediction.</p><h3 id='train(...)' class='header-anchors'>train(...)</h3><p><code>train</code> is called when you run <strong>pio train</strong>. This is where MLlib ALS algorithm, i.e. <code>ALS.train</code>, is used to train a predictive model.</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39</pre></td><td class="code"><pre>  <span class="k">def</span> <span class="n">train</span><span class="o">(</span><span class="n">sc</span><span class="k">:</span> <span class="kt">SparkContext</span><span class="o">,</span> <span class="n">data</span><span class="k">:</span> <span class="kt">PreparedData</span><span class="o">)</span><span class="k">:</span> <span class="kt">ALSModel</span> <span class="o">=</span> <span class="o">{</span>
    <span class="o">...</span>
    <span class="c1">// Convert user and item String IDs to Int index for MLlib
</span>    <span class="k">val</span> <span class="n">userStringIntMap</span> <span class="k">=</span> <span class="nc">BiMap</span><span class="o">.</span><span class="n">stringInt</span><span class="o">(</span><span class="n">data</span><span class="o">.</span><span class="n">ratings</span><span class="o">.</span><span class="n">map</span><span class="o">(</span><span class="k">_</span><span class="o">.</span><span class="n">user</span><span class="o">))</span>
    <span class="k">val</span> <span class="n">itemStringIntMap</span> <span class="k">=</span> <span class="nc">BiMap</span><span class="o">.</span><span class="n">stringInt</span><span class="o">(</span><span class="n">data</span><span class="o">.</span><span class="n">ratings</span><span class="o">.</span><span class="n">map</span><span class="o">(</span><span class="k">_</span><span class="o">.</span><span class="n">item</span><span class="o">))</span>
    <span class="k">val</span> <span class="n">mllibRatings</span> <span class="k">=</span> <span class="n">data</span><span class="o">.</span><span class="n">ratings</span><span class="o">.</span><span class="n">map</span><span class="o">(</span> <span class="n">r</span> <span class="k">=&gt;</span>
      <span class="c1">// MLlibRating requires integer index for user and item
</span>      <span class="nc">MLlibRating</span><span class="o">(</span><span class="n">userStringIntMap</span><span class="o">(</span><span class="n">r</span><span class="o">.</span><span class="n">user</span><span class="o">),</span> <span class="n">itemStringIntMap</span><span class="o">(</span><span class="n">r</span><span class="o">.</span><span class="n">item</span><span class="o">),</span> <span class="n">r</span><span class="o">.</span><span class="n">rating</span><span class="o">)</span>
    <span class="o">)</span>

    <span class="c1">// seed for MLlib ALS
</span>    <span class="k">val</span> <span class="n">seed</span> <span class="k">=</span> <span class="n">ap</span><span class="o">.</span><span class="n">seed</span><span class="o">.</span><span class="n">getOrElse</span><span class="o">(</span><span class="nc">System</span><span class="o">.</span><span class="n">nanoTime</span><span class="o">)</span>

    <span class="c1">// If you only have one type of implicit event (Eg. "view" event only),
</span>    <span class="c1">// replace ALS.train(...) with
</span>    <span class="c1">//val m = ALS.trainImplicit(
</span>      <span class="c1">//ratings = mllibRatings,
</span>      <span class="c1">//rank = ap.rank,
</span>      <span class="c1">//iterations = ap.numIterations,
</span>      <span class="c1">//lambda = ap.lambda,
</span>      <span class="c1">//blocks = -1,
</span>      <span class="c1">//alpha = 1.0,
</span>      <span class="c1">//seed = seed)
</span>
    <span class="k">val</span> <span class="n">m</span> <span class="k">=</span> <span class="nc">ALS</span><span class="o">.</span><span class="n">train</span><span class="o">(</span>
      <span class="n">ratings</span> <span class="k">=</span> <span class="n">mllibRatings</span><span class="o">,</span>
      <span class="n">rank</span> <span class="k">=</span> <span class="n">ap</span><span class="o">.</span><span class="n">rank</span><span class="o">,</span>
      <span class="n">iterations</span> <span class="k">=</span> <span class="n">ap</span><span class="o">.</span><span class="n">numIterations</span><span class="o">,</span>
      <span class="n">lambda</span> <span class="k">=</span> <span class="n">ap</span><span class="o">.</span><span class="n">lambda</span><span class="o">,</span>
      <span class="n">blocks</span> <span class="k">=</span> <span class="o">-</span><span class="mi">1</span><span class="o">,</span>
      <span class="n">seed</span> <span class="k">=</span> <span class="n">seed</span><span class="o">)</span>

    <span class="k">new</span> <span class="nc">ALSModel</span><span class="o">(</span>
      <span class="n">rank</span> <span class="k">=</span> <span class="n">m</span><span class="o">.</span><span class="n">rank</span><span class="o">,</span>
      <span class="n">userFeatures</span> <span class="k">=</span> <span class="n">m</span><span class="o">.</span><span class="n">userFeatures</span><span class="o">,</span>
      <span class="n">productFeatures</span> <span class="k">=</span> <span class="n">m</span><span class="o">.</span><span class="n">productFeatures</span><span class="o">,</span>
      <span class="n">userStringIntMap</span> <span class="k">=</span> <span class="n">userStringIntMap</span><span class="o">,</span>
      <span class="n">itemStringIntMap</span> <span class="k">=</span> <span class="n">itemStringIntMap</span><span class="o">)</span>
  <span class="o">}</span>
</pre></td></tr></tbody></table> </div> <h4 id='working-with-spark-mllib&#39;s-als.train(....)' class='header-anchors'>Working with Spark MLlib&#39;s ALS.train(....)</h4><p>As mentioned above, MLlib&#39;s <code>Rating</code> does not support <code>String</code> user ID and item ID. Its <code>ALS.train</code> thus also assumes <code>Int</code>-only <code>Rating</code>.</p><p>Here you need to map your String-supported <code>Rating</code> to MLlib&#39;s Integer-only <code>Rating</code>. First, you can rename MLlib&#39;s Integer-only <code>Rating</code> to <code>MLlibRating</code> for clarity:</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>import org.apache.spark.mllib.recommendation.<span class="o">{</span>Rating <span class="o">=</span>&gt; MLlibRating<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p>You then create a bi-directional map with <code>BiMap.stringInt</code> which maps each String record to an Integer index.</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2</pre></td><td class="code"><pre>val userStringIntMap <span class="o">=</span> BiMap.stringInt<span class="o">(</span>data.ratings.map<span class="o">(</span>_.user<span class="o">))</span>
val itemStringIntMap <span class="o">=</span> BiMap.stringInt<span class="o">(</span>data.ratings.map<span class="o">(</span>_.item<span class="o">))</span>
</pre></td></tr></tbody></table> </div> <p>Finally, you re-create each <code>Rating</code> event as <code>MLlibRating</code>:</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>MLlibRating<span class="o">(</span>userStringIntMap<span class="o">(</span>r.user<span class="o">)</span>, itemStringIntMap<span class="o">(</span>r.item<span class="o">)</span>, r.rating<span class="o">)</span>
</pre></td></tr></tbody></table> </div> <p>In addition to <code>RDD[MLlibRating]</code>, <code>ALS.train</code> takes the following parameters: <em>rank</em>, <em>iterations</em>, <em>lambda</em> and <em>seed</em>.</p><p>The values of these parameters are specified in <em>algorithms</em> of MyRecommendation/<strong><em>engine.json</em></strong>:</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10
11
12
13
14
15</pre></td><td class="code"><pre><span class="o">{</span>
  ...
  <span class="s2">"algorithms"</span>: <span class="o">[</span>
    <span class="o">{</span>
      <span class="s2">"name"</span>: <span class="s2">"als"</span>,
      <span class="s2">"params"</span>: <span class="o">{</span>
        <span class="s2">"rank"</span>: 10,
        <span class="s2">"numIterations"</span>: 20,
        <span class="s2">"lambda"</span>: 0.01,
        <span class="s2">"seed"</span>: 3
      <span class="o">}</span>
    <span class="o">}</span>
  <span class="o">]</span>
  ...
<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p>PredictionIO will automatically loads these values into the constructor <code>ap</code>, which has a corresponding case case <code>ALSAlgorithmParams</code>:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5</pre></td><td class="code"><pre><span class="k">case</span> <span class="k">class</span> <span class="nc">ALSAlgorithmParams</span><span class="o">(</span>
  <span class="n">rank</span><span class="k">:</span> <span class="kt">Int</span><span class="o">,</span>
  <span class="n">numIterations</span><span class="k">:</span> <span class="kt">Int</span><span class="o">,</span>
  <span class="n">lambda</span><span class="k">:</span> <span class="kt">Double</span><span class="o">,</span>
  <span class="n">seed</span><span class="k">:</span> <span class="kt">Option</span><span class="o">[</span><span class="kt">Long</span><span class="o">])</span> <span class="k">extends</span> <span class="nc">Params</span>
</pre></td></tr></tbody></table> </div> <p>The <code>seed</code> parameter is an optional parameter, which is used by MLlib ALS algorithm internally to generate random values. If the <code>seed</code> is not specified, current system time would be used and hence each train may produce different reuslts. Specify a fixed value for the <code>seed</code> if you want to have deterministic result (For example, when you are testing).</p><p><code>ALS.train</code> then returns a <code>MatrixFactorizationModel</code> model which contains RDD data. RDD is a distributed collection of items which <em>does not</em> persist. To store the model, you convert the model to <code>ALSModel</code> class at the end. <code>ALSModel</code> is a persistable class that extends <code>MatrixFactorizationModel</code>.</p> <blockquote> <p>The detailed implementation can be found at MyRecommendation/src/main/scala/<strong><em>ALSModel.scala</em></strong></p></blockquote> <p>PredictionIO will automatically store the returned model, i.e. <code>ALSModel</code> in this case.</p><h3 id='predict(...)' class='header-anchors'>predict(...)</h3><p><code>predict</code> is called when you send a JSON query to <a href="http://localhost:8000/queries.json">http://localhost:8000/queries.json</a>. PredictionIO converts the query, such as <code>{ &quot;user&quot;: &quot;1&quot;, &quot;num&quot;: 4 }</code> to the <code>Query</code> class you defined previously.</p><p>The predictive model <code>MatrixFactorizationModel</code> of MLlib ALS, which is now extended as <code>ALSModel</code>, offers a method called <code>recommendProducts</code>. <code>recommendProducts</code> takes two parameters: user id (i.e. the <code>Int</code> index of <code>query.user</code>) and the number of items to be returned (i.e. <code>query.num</code>). It predicts the top <em>num</em> of items a user will like.</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9
10
11
12
13
14
15</pre></td><td class="code"><pre>  <span class="k">def</span> <span class="n">predict</span><span class="o">(</span><span class="n">model</span><span class="k">:</span> <span class="kt">ALSModel</span><span class="o">,</span> <span class="n">query</span><span class="k">:</span> <span class="kt">Query</span><span class="o">)</span><span class="k">:</span> <span class="kt">PredictedResult</span> <span class="o">=</span> <span class="o">{</span>
    <span class="c1">// Convert String ID to Int index for Mllib
</span>    <span class="n">model</span><span class="o">.</span><span class="n">userStringIntMap</span><span class="o">.</span><span class="n">get</span><span class="o">(</span><span class="n">query</span><span class="o">.</span><span class="n">user</span><span class="o">).</span><span class="n">map</span> <span class="o">{</span> <span class="n">userInt</span> <span class="k">=&gt;</span>
      <span class="c1">// create inverse view of itemStringIntMap
</span>      <span class="k">val</span> <span class="n">itemIntStringMap</span> <span class="k">=</span> <span class="n">model</span><span class="o">.</span><span class="n">itemStringIntMap</span><span class="o">.</span><span class="n">inverse</span>
      <span class="c1">// recommendProducts() returns Array[MLlibRating], which uses item Int
</span>      <span class="c1">// index. Convert it to String ID for returning PredictedResult
</span>      <span class="k">val</span> <span class="n">itemScores</span> <span class="k">=</span> <span class="n">model</span><span class="o">.</span><span class="n">recommendProducts</span><span class="o">(</span><span class="n">userInt</span><span class="o">,</span> <span class="n">query</span><span class="o">.</span><span class="n">num</span><span class="o">)</span>
        <span class="o">.</span><span class="n">map</span> <span class="o">(</span><span class="n">r</span> <span class="k">=&gt;</span> <span class="nc">ItemScore</span><span class="o">(</span><span class="n">itemIntStringMap</span><span class="o">(</span><span class="n">r</span><span class="o">.</span><span class="n">product</span><span class="o">),</span> <span class="n">r</span><span class="o">.</span><span class="n">rating</span><span class="o">))</span>
      <span class="k">new</span> <span class="nc">PredictedResult</span><span class="o">(</span><span class="n">itemScores</span><span class="o">)</span>
    <span class="o">}.</span><span class="n">getOrElse</span><span class="o">{</span>
      <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="o">(</span><span class="n">s</span><span class="s">"No prediction for unknown user ${query.user}."</span><span class="o">)</span>
      <span class="k">new</span> <span class="nc">PredictedResult</span><span class="o">(</span><span class="nc">Array</span><span class="o">.</span><span class="n">empty</span><span class="o">)</span>
    <span class="o">}</span>
  <span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p>Note that <code>recommendProducts</code> returns the <code>Int</code> indices of items. You map them back to <code>String</code> with <code>itemIntStringMap</code> before they are returned.</p> <blockquote> <p>You have defined the class <code>PredictedResult</code> earlier.</p></blockquote> <p>PredictionIO passes the returned <code>PredictedResult</code> object to <em>Serving</em>.</p><h2 id='serving' class='header-anchors'>Serving</h2><p>The <code>serve</code> method of class <code>Serving</code> processes predicted result. It is also responsible for combining multiple predicted results into one if you have more than one predictive model. <em>Serving</em> then returns the final predicted result. PredictionIO will convert it to a JSON response automatically.</p><p>In MyRecommendation/src/main/scala/<strong><em>Serving.scala</em></strong>,</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1
2
3
4
5
6
7
8
9</pre></td><td class="code"><pre><span class="k">class</span> <span class="nc">Serving</span>
  <span class="k">extends</span> <span class="nc">LServing</span><span class="o">[</span><span class="kt">Query</span>, <span class="kt">PredictedResult</span><span class="o">]</span> <span class="o">{</span>

  <span class="k">override</span>
  <span class="k">def</span> <span class="n">serve</span><span class="o">(</span><span class="n">query</span><span class="k">:</span> <span class="kt">Query</span><span class="o">,</span>
    <span class="n">predictedResults</span><span class="k">:</span> <span class="kt">Seq</span><span class="o">[</span><span class="kt">PredictedResult</span><span class="o">])</span><span class="k">:</span> <span class="kt">PredictedResult</span> <span class="o">=</span> <span class="o">{</span>
    <span class="n">predictedResults</span><span class="o">.</span><span class="n">head</span>
  <span class="o">}</span>
<span class="o">}</span>
</pre></td></tr></tbody></table> </div> <p>When you send a JSON query to <a href="http://localhost:8000/queries.json">http://localhost:8000/queries.json</a>, <code>PredictedResult</code> from all models will be passed to <code>serve</code> as a sequence, i.e. <code>Seq[PredictedResult]</code>.</p> <blockquote> <p>An engine can train multiple models if you specify more than one Algorithm component in <code>object RecommendationEngine</code> inside <strong><em>Engine.scala</em></strong>. Since only one <code>ALSAlgorithm</code> is implemented by default, this <code>Seq</code> contains one element.</p></blockquote> <p>Now you should have a good understanding of the DASE model. We will show you an example of customizing the Data Preparator to exclude certain items from your training set.</p><h4 id='<a-href="/templates/recommendation/reading-custom-events/">next:-reading-custom-events</a>' class='header-anchors' ><a href="/templates/recommendation/reading-custom-events/">Next: Reading Custom Events</a></h4></div></div></div></div><footer><div class="container"><div class="seperator"></div><div class="row"><div class="col-md-4 col-md-push-8 col-xs-12"><div class="subscription-form-wrapper"><h4>Subscribe to our Newsletter</h4><form class="ajax-form" id="subscribe-form" method="POST" action="https://script.google.com/macros/s/AKfycbwhzeKCQJjQ52eVAqNT_vcklH07OITUO7wzOMDXvK6EGAWgaZgF/exec"><input class="required underlined-input" type="email" placeholder="Your email address" name="subscription_email" id="subscription_email"/><input class="pill-button" value="SUBSCRIBE" type="submit" data-state-normal="SUBSCRIBE" data-state-sucess="SUBSCRIBED!" data-state-loading="SENDING..." onclick="t($('#subscription_email').val());"/><p class="result"></p></form></div></div><div class="col-md-2 col-md-pull-4 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Community</h4><ul><li><a href="//docs.prediction.io/install/" target="blank">Download</a></li><li><a href="//docs.prediction.io/" target="blank">Docs</a></li><li><a href="//github.com/PredictionIO/PredictionIO" target="blank">GitHub</a></li><li><a href="//groups.google.com/forum/#!forum/predictionio-user" target="blank">Support Forum</a></li><li><a href="//stackoverflow.com/questions/tagged/predictionio" target="blank">Stackoverflow</a></li><li><a href="mailto:&#x73;&#x75;&#x70;&#x70;&#x6F;&#x72;&#x74;&#x40;&#x70;&#x72;&#x65;&#x64;&#x69;&#x63;&#x74;&#x69;&#x6F;&#x6E;&#x2E;&#x69;&#x6F;" target="blank">Contact Us</a></li></ul></div></div><div class="col-md-2 col-md-pull-4 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Contribute</h4><ul><li><a href="//docs.prediction.io/community/contribute-code/" target="blank">Contribute</a></li><li><a href="//github.com/PredictionIO/PredictionIO" target="blank">Source Code</a></li><li><a href="//predictionio.atlassian.net/secure/Dashboard.jspa" target="blank">Bug Tracker</a></li><li><a href="//groups.google.com/forum/#!forum/predictionio-dev" target="blank">Contributors&#146; Forum</a></li><li><a href="//prediction.io/cla">Contributor Agreement</a></li><li><a href="//predictionio.uservoice.com/forums/219398-general/filters/top">Request Features</a></li></ul></div></div><div class="col-md-2 col-md-pull-4 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Enterprise</h4><ul><li><a href="//docs.prediction.io/support/" target="blank">Support</a></li><li><a href="//prediction.io/enterprise">Enterprise</a></li><li><a href="//prediction.io/products/predictionio-enterprise">Services</a></li></ul></div><div class="footer-link-column-row"><h4>Connect</h4><ul><li><a href="//blog.prediction.io/" target="blank">Blog</a></li><li><a href="//predictionio.theresumator.com/" target="blank">Careers</a></li></ul></div></div><div class="col-md-2 col-md-pull-4 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Partnership</h4><ul><li><a href="//prediction.io/partners/program">Partner Program</a></li></ul></div></div></div></div><div id="footer-bottom"><div class="container"><div class="row"><div class="col-md-12"><div id="footer-logo-wrapper"><img alt="PredictionIO" src="/images/logos/logo-white-d1e9c6e6.png"/></div><div id="social-icons-wrapper"><a class="github-button" href="https://github.com/PredictionIO/PredictionIO" data-style="mega" data-count-href="/PredictionIO/PredictionIO/stargazers" data-count-api="/repos/PredictionIO/PredictionIO#stargazers_count" data-count-aria-label="# stargazers on GitHub" aria-label="Star PredictionIO/PredictionIO on GitHub">Star</a> <a class="github-button" href="https://github.com/PredictionIO/PredictionIO/fork" data-icon="octicon-git-branch" data-style="mega" data-count-href="/PredictionIO/PredictionIO/network" data-count-api="/repos/PredictionIO/PredictionIO#forks_count" data-count-aria-label="# forks on GitHub" aria-label="Fork PredictionIO/PredictionIO on GitHub">Fork</a> <script id="github-bjs" async="" defer="" src="https://buttons.github.io/buttons.js"></script><a href="//www.facebook.com/predictionio" target="blank"><img alt="PredictionIO on Twitter" src="/images/icons/twitter-ea9dc152.png"/></a> <a href="//twitter.com/predictionio" target="blank"><img alt="PredictionIO on Facebook" src="/images/icons/facebook-5c57939c.png"/></a> </div></div></div></div></div></footer></div><script>(function(w,d,t,u,n,s,e){w['SwiftypeObject']=n;w[n]=w[n]||function(){
(w[n].q=w[n].q||[]).push(arguments);};s=d.createElement(t);
e=d.getElementsByTagName(t)[0];s.async=1;s.src=u;e.parentNode.insertBefore(s,e);
})(window,document,'script','//s.swiftypecdn.com/install/v1/st.js','_st');

_st('install','HaUfpXXV87xoB_zzCQ45');</script><script>var _qevents = _qevents || [];
(function() {
var elem = document.createElement('script');
elem.src = (document.location.protocol == "https:" ? "https://secure" : "http://edge") + ".quantserve.com/quant.js";
elem.async = true;
elem.type = "text/javascript";
var scpt = document.getElementsByTagName('script')[0];
scpt.parentNode.insertBefore(elem, scpt);
})();
_qevents.push({
qacct:"p-stVMxuw8H5EPX"
});</script><noscript><div style="display:none;"><img src="//pixel.quantserve.com/pixel/p-stVMxuw8H5EPX.gif" border="0" height="1" width="1" alt="Quantcast"/></div></noscript><script>adroll_adv_id = "CPSSMJFFZ5DDHITC2STA54";
adroll_pix_id = "UWX4N2WIMJADVHJGOFTM44";
(function () {
    var _onload = function(){
        if (document.readyState && !/loaded|complete/.test(document.readyState)){setTimeout(_onload, 10);return}
        if (!window.__adroll_loaded){__adroll_loaded=true;setTimeout(_onload, 50);return}
        var scr = document.createElement("script");
        var host = (("https:" == document.location.protocol) ? "https://s.adroll.com" : "http://a.adroll.com");
        scr.setAttribute('async', 'true');
        scr.type = "text/javascript";
        scr.src = host + "/j/roundtrip.js";
        ((document.getElementsByTagName('head') || [null])[0] ||
            document.getElementsByTagName('script')[0].parentNode).appendChild(scr);
    };
    if (window.addEventListener) {window.addEventListener('load', _onload, false);}
    else {window.attachEvent('onload', _onload)}
}());</script><script src="/javascripts/application-5a24945b.js"></script></body></html>