http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/452034c1/demo/textclassification/index.html ---------------------------------------------------------------------- diff --git a/demo/textclassification/index.html b/demo/textclassification/index.html index 862af35..f4ee9ea 100644 --- a/demo/textclassification/index.html +++ b/demo/textclassification/index.html @@ -1,5 +1,5 @@ -<!DOCTYPE html><html><head><title>Text Classification Engine Tutorial</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="Text Classification Engine Tutorial"/><link rel="canonical" href="https://predictionio.incubator.apache.org/demo/textclassification/"/><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-eccfc6cb.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min.js"></script><s cript src="//cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}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://predictionio.incubator.apache.org/";><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a><span>â¢</span></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="/gallery/template-gallery">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-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>Text Classification Engine Tutorial</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="con tainer-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 cla ss="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="/batchpredict/"><spa n>Batch Predictions</span></a></li><li class="level-2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</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><li class="level-2"><a class="final" href="/deploy/plugin/"><span>Engine Server Plugin</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="/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><li class="level-2"><a class="final" href="/datacollection/plugin/"><span>Event Serve r Plugin</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>PredictionIO Official Templates</span></a><ul><li class="level-2"><a class="final" href="/templates/"><span>Intro</span></a></li><li class="level-2"><a class="expandible" href="#"><span>Recommendation</span></a><ul><li class="level-3"><a class="final" href="/templates/recommendation/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/t emplates/recommendation/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/evaluation/"><span>Evaluation Explained</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/reading-custom-events/"><span>Read Custom Events</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/customize-data-prep/"><span>Customize Data Preparator</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/customize-serving/"><span>Customize Serving</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/training-with-implicit-preference/"><span>Train with Implicit Preference</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/blacklist-items/"><span>Filter Recommended Items by Blacklist in Query</span></a></li ><li class="level-3"><a class="final" >href="/templates/recommendation/batch-evaluator/"><span>Batch Persistable >Evaluator</span></a></li></ul></li><li class="level-2"><a class="expandible" >href="#"><span>E-Commerce Recommendation</span></a><ul><li class="level-3"><a >class="final" >href="/templates/ecommercerecommendation/quickstart/"><span>Quick >Start</span></a></li><li class="level-3"><a class="final" >href="/templates/ecommercerecommendation/dase/"><span>DASE</span></a></li><li >class="level-3"><a class="final" >href="/templates/ecommercerecommendation/how-to/"><span>How-To</span></a></li><li > class="level-3"><a class="final" >href="/templates/ecommercerecommendation/train-with-rate-event/"><span>Train >with Rate Event</span></a></li><li class="level-3"><a class="final" >href="/templates/ecommercerecommendation/adjust-score/"><span>Adjust >Score</span></a></li></ul></li><li class="level-2"><a class="expandible" >href="#"><span>Similar Product</span></a><ul><li class="level-3"><a >class="fina l" href="/templates/similarproduct/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/multi-events-multi-algos/"><span>Multiple Events and Multiple Algorithms</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/return-item-properties/"><span>Returns Item Properties</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/train-with-rate-event/"><span>Train with Rate Event</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/rid-user-set-event/"><span>Get Rid of Events for Users</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/recommended-user/"><span>Recommend Users</span></a></li></ul></li><l i class="level-2"><a class="expandible" href="#"><span>Classification</span></a><ul><li class="level-3"><a class="final" href="/templates/classification/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/classification/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/classification/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/classification/add-algorithm/"><span>Use Alternative Algorithm</span></a></li><li class="level-3"><a class="final" href="/templates/classification/reading-custom-properties/"><span>Read Custom Properties</span></a></li></ul></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="/gallery/template-gallery/"><span>Browse</span></a></li><li class="level-2"><a class="final" href="/community/submit-template/"><span>Submit your Engine as a T emplate</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 active" 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>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="/cli/"><span>Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/resources/release/"><span>Release Cadence</span></a></li><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><li class="level-1"><a class="expandible" href="#"><span>Apache Software Foundation</span></a><ul><li class="level-2"><a class="final" href="https://www.apache.org/";><span>Apache Homepage</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/licenses/";><span>License</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/foundation/sponsorship.html";><span>Sponsorship</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/foundation/thanks.html";><span>Thanks</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/security/";><span>Security</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="breadcrumbs" class="hidden-sm hidden xs"><ul><li><a href="#">Demo Tutorials</a><span class="spacer">></span></li><li><span class="last">Text Classification Engine Tutorial</span></li ></ul></div><div id="page-title"><h1>Text Classification Engine >Tutorial</h1></div></div><div id="table-of-content-wrapper"><h5>On this >page</h5><aside id="table-of-contents"><ul> <li> <a >href="#introduction">Introduction</a> </li> <li> <a >href="#prerequisites">Prerequisites</a> </li> <li> <a >href="#engine-overview">Engine Overview</a> </li> <li> <a >href="#quick-start">Quick Start</a> </li> </ul> </li> <li> <a >href="#detailed-explanation-of-dase">Detailed Explanation of DASE</a> <ul> ><li> <a href="#importing-data">Importing Data</a> </li> <li> <a >href="#data-source-reading-event-data">Data Source: Reading Event Data</a> ></li> <li> <a href="#preparator-data-processing-with-dase">Preparator : Data >Processing With DASE</a> </li> <li> <a href="#algorithm-component">Algorithm >Component</a> </li> <li> <a >href="#serving-delivering-the-final-prediction">Serving: Delivering the Final >Prediction</a> </li> <li> <a >href="#evaluation-model-assessment-and-selection">Evaluation: Model >Assessment a nd Selection</a> </li> <li> <a href="#engine-deployment">Engine Deployment</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/tree/livedoc/docs/manual/source/demo/textclassification.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="breadcrumbs" class="hidden-sm hidden xs"><ul><li><a href="#">Demo Tutorials</a><span class="spacer">></span></li><li><span class="last">Text Classification Engine Tutorial</span></li></ul></div><div id="page-title"><h1>Text Classification Engine Tutorial</h1></div></div><div class="content"> <p>(Updated for Text Classification Template version 3.1)</p><h2 id='introduction' class='header-anchors'>Introduction</h2><p>In the real world, there are many applications that collect text as data. For example, spam detectors take email and header content to automatically determine what is or is not spam; applicati ons can gague the general sentiment in a geographical area by analyzing Twitter data; and news articles can be automatically categorized based solely on the text content.There are a wide array of machine learning models you can use to create, or train, a predictive model to assign an incoming article, or query, to an existing category. Before you can use these techniques you must first transform the text data (in this case the set of news articles) into numeric vectors, or feature vectors, that can be used to train your model.</p><p>The purpose of this tutorial is to illustrate how you can go about doing this using PredictionIO's platform. The advantages of using this platform include: a dynamic engine that responds to queries in real-time; <a href="http://en.wikipedia.org/wiki/Separation_of_concerns";>separation of concerns</a>, which offers code re-use and maintainability, and distributed computing capabilities for scalability and efficiency. Moreover, it is easy to incorporate non-trivial data modeling tasks into the DASE architecture allowing Data Scientists to focus on tasks related to modeling. This tutorial will exemplify some of these ideas by guiding you through PredictionIO's <a href="/gallery/template-gallery/#natural-language-processing">text classification template</a>.</p><h2 id='prerequisites' class='header-anchors'>Prerequisites</h2><p>Before getting started, please make sure that you have the latest version of Apache PredictionIO (incubating) <a href="http://predictionio.incubator.apache.org/install/";>installed</a>. We emphasize here that this is an engine template written in <strong>Scala</strong> and can be more generally thought of as an SBT project containing all the necessary components.</p><p>You should also download the engine template named Text Classification Engine that accompanies this tutorial by cloning the template repository:</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>git clone https://github.com/apache/incubator-predictionio-template-text-classifier.git < Your new engine directory > -</pre></td></tr></tbody></table> </div> <h2 id='engine-overview' class='header-anchors'>Engine Overview</h2><p>The engine follows the DASE architecture which we briefly review here. As a user, you are tasked with collecting data for your web or application, and importing it into PredictionIO's Event Server. Once the data is in the server, it can be read and processed by the engine via the Data Source and Preparation components, respectively. The Algorithm component uses the processed, or prepared, data to train a set of predictive models. Once you have trained these models, you are ready to deploy your engine and respond to real-time queries via the Serving component which combines the results from different fitted models. The Evaluation component is used to compute an appropriate metric to test the performance of a fitted model, as well as aid in the tuning of model hyper parameters.</p><p>This engine template is meant to handle text classification which means you will be worki ng with text data. This means that a query, or newly observed documents, will be of the form:</p><p><code>{text : String}</code>.</p><p>In the running example, a query would be an incoming news article. Once the engine is deployed it can process the query, and then return a Predicted Result of the form</p><p><code>{category : String, confidence : Double}</code>.</p><p>Here category is the model's class assignment for this new text document (i.e. the best guess for this article's categorization), and confidence, a value between 0 and 1 representing your confidence in the category prediction (0 meaning you have no confidence in the prediction). The Actual Result is of the form</p><p><code>{category : String}</code>.</p><p>This is used in the evaluation stage when estimating the performance of your predictive model (how well does the model predict categories). Please refer to the <a href="https://predictionio.incubator.apache.org/customize/";>following tutorial</a> for a more de tailed explanation of how your engine will interact with your web application, as well as an in depth-overview of DASE.</p><h2 id='quick-start' class='header-anchors'>Quick Start</h2><p>This is a quick start guide in case you want to start using the engine right away. Sample email data for spam classification will be used. For more detailed information, read the subsequent sections.</p><h3 id='1.-create-a-new-application.' class='header-anchors'>1. Create a new application.</h3><p>After the application is created, you will be given an access key and application ID for the application.</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><span class="gp">$ </span>pio app new MyTextApp +<!DOCTYPE html><html><head><title>Text Classification Engine Tutorial</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="Text Classification Engine Tutorial"/><link rel="canonical" href="https://predictionio.apache.org/demo/textclassification/"/><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-eccfc6cb.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min.js"></script><script src= "//cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}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://predictionio.apache.org/";><img alt="Apache PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a><span>â¢</span></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="/gallery/template-gallery">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-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="c ol-md-9 col-sm-11 col-xs-11"><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>Text Classification Engine Tutorial</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⢠Documentation</span></a><ul><li >class="level-2"><a class="final" href="/"><span>Welcome to Apache >PredictionIO</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</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="/batchpredict/"><span>Batch Predictions</span></a></li><li class="level- 2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</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><li class="level-2"><a class="final" href="/deploy/plugin/"><span>Engine Server Plugin</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="expandib le" 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="/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><li class="level-2"><a class="final" href="/datacollection/plugin/"><span>Event Server Plugin</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="/eva luation/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>PredictionIO Official Templates</span></a><ul><li class="level-2"><a class="final" href="/templates/"><span>Intro</span></a></li><li class="level-2"><a class="expandible" href="#"><span>Recommendation</span></a><ul><li class="level-3"><a class="final" href="/templates/recommendation/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/dase/"><span>DASE</span></a> </li><li class="level-3"><a class="final" href="/templates/recommendation/evaluation/"><span>Evaluation Explained</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/reading-custom-events/"><span>Read Custom Events</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/customize-data-prep/"><span>Customize Data Preparator</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/customize-serving/"><span>Customize Serving</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/training-with-implicit-preference/"><span>Train with Implicit Preference</span></a></li><li class="level-3"><a class="final" href="/templates/recommendation/blacklist-items/"><span>Filter Recommended Items by Blacklist in Query</span></a></li><li class="level-3"><a class="final" href="/templat es/recommendation/batch-evaluator/"><span>Batch Persistable Evaluator</span></a></li></ul></li><li class="level-2"><a class="expandible" href="#"><span>E-Commerce Recommendation</span></a><ul><li class="level-3"><a class="final" href="/templates/ecommercerecommendation/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/ecommercerecommendation/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/ecommercerecommendation/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/ecommercerecommendation/train-with-rate-event/"><span>Train with Rate Event</span></a></li><li class="level-3"><a class="final" href="/templates/ecommercerecommendation/adjust-score/"><span>Adjust Score</span></a></li></ul></li><li class="level-2"><a class="expandible" href="#"><span>Similar Product</span></a><ul><li class="level-3"><a class="final" href="/templates/similarproduct/quickstart/"><spa n>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/multi-events-multi-algos/"><span>Multiple Events and Multiple Algorithms</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/return-item-properties/"><span>Returns Item Properties</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/train-with-rate-event/"><span>Train with Rate Event</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/rid-user-set-event/"><span>Get Rid of Events for Users</span></a></li><li class="level-3"><a class="final" href="/templates/similarproduct/recommended-user/"><span>Recommend Users</span></a></li></ul></li><li class="level-2"><a class="expandible" href="#"><sp an>Classification</span></a><ul><li class="level-3"><a class="final" href="/templates/classification/quickstart/"><span>Quick Start</span></a></li><li class="level-3"><a class="final" href="/templates/classification/dase/"><span>DASE</span></a></li><li class="level-3"><a class="final" href="/templates/classification/how-to/"><span>How-To</span></a></li><li class="level-3"><a class="final" href="/templates/classification/add-algorithm/"><span>Use Alternative Algorithm</span></a></li><li class="level-3"><a class="final" href="/templates/classification/reading-custom-properties/"><span>Read Custom Properties</span></a></li></ul></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="/gallery/template-gallery/"><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 active" >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>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="/cli/"><span>Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/resources/release/"><span>Release Cadence</span></a></li><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><li class="level-1 "><a class="expandible" href="#"><span>Apache Software Foundation</span></a><ul><li class="level-2"><a class="final" href="https://www.apache.org/";><span>Apache Homepage</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/licenses/";><span>License</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/foundation/sponsorship.html";><span>Sponsorship</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/foundation/thanks.html";><span>Thanks</span></a></li><li class="level-2"><a class="final" href="https://www.apache.org/security/";><span>Security</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="breadcrumbs" class="hidden-sm hidden xs"><ul><li><a href="#">Demo Tutorials</a><span class="spacer">></span></li><li><span class="last">Text Classification Engine Tutorial</span></li></ul></div><div id="page-title"><h1>Text Classifica tion Engine Tutorial</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#introduction">Introduction</a> </li> <li> <a href="#prerequisites">Prerequisites</a> </li> <li> <a href="#engine-overview">Engine Overview</a> </li> <li> <a href="#quick-start">Quick Start</a> </li> </ul> </li> <li> <a href="#detailed-explanation-of-dase">Detailed Explanation of DASE</a> <ul> <li> <a href="#importing-data">Importing Data</a> </li> <li> <a href="#data-source-reading-event-data">Data Source: Reading Event Data</a> </li> <li> <a href="#preparator-data-processing-with-dase">Preparator : Data Processing With DASE</a> </li> <li> <a href="#algorithm-component">Algorithm Component</a> </li> <li> <a href="#serving-delivering-the-final-prediction">Serving: Delivering the Final Prediction</a> </li> <li> <a href="#evaluation-model-assessment-and-selection">Evaluation: Model Assessment and Selection</a> </li> <li> <a href="#engine-deploym ent">Engine Deployment</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/tree/livedoc/docs/manual/source/demo/textclassification.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="breadcrumbs" class="hidden-sm hidden xs"><ul><li><a href="#">Demo Tutorials</a><span class="spacer">></span></li><li><span class="last">Text Classification Engine Tutorial</span></li></ul></div><div id="page-title"><h1>Text Classification Engine Tutorial</h1></div></div><div class="content"> <p>(Updated for Text Classification Template version 3.1)</p><h2 id='introduction' class='header-anchors'>Introduction</h2><p>In the real world, there are many applications that collect text as data. For example, spam detectors take email and header content to automatically determine what is or is not spam; applications can gague the general sentiment in a geographica l area by analyzing Twitter data; and news articles can be automatically categorized based solely on the text content.There are a wide array of machine learning models you can use to create, or train, a predictive model to assign an incoming article, or query, to an existing category. Before you can use these techniques you must first transform the text data (in this case the set of news articles) into numeric vectors, or feature vectors, that can be used to train your model.</p><p>The purpose of this tutorial is to illustrate how you can go about doing this using PredictionIO's platform. The advantages of using this platform include: a dynamic engine that responds to queries in real-time; <a href="http://en.wikipedia.org/wiki/Separation_of_concerns";>separation of concerns</a>, which offers code re-use and maintainability, and distributed computing capabilities for scalability and efficiency. Moreover, it is easy to incorporate non-trivial data modeling tasks into the DASE archi tecture allowing Data Scientists to focus on tasks related to modeling. This tutorial will exemplify some of these ideas by guiding you through PredictionIO's <a href="/gallery/template-gallery/#natural-language-processing">text classification template</a>.</p><h2 id='prerequisites' class='header-anchors'>Prerequisites</h2><p>Before getting started, please make sure that you have the latest version of Apache PredictionIO <a href="http://predictionio.apache.org/install/";>installed</a>. We emphasize here that this is an engine template written in <strong>Scala</strong> and can be more generally thought of as an SBT project containing all the necessary components.</p><p>You should also download the engine template named Text Classification Engine that accompanies this tutorial by cloning the template repository:</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>git clone >https://github.com/apache/incubator-predictionio-template-text-classifier.git >< Your new engine directory > +</pre></td></tr></tbody></table> </div> <h2 id='engine-overview' class='header-anchors'>Engine Overview</h2><p>The engine follows the DASE architecture which we briefly review here. As a user, you are tasked with collecting data for your web or application, and importing it into PredictionIO's Event Server. Once the data is in the server, it can be read and processed by the engine via the Data Source and Preparation components, respectively. The Algorithm component uses the processed, or prepared, data to train a set of predictive models. Once you have trained these models, you are ready to deploy your engine and respond to real-time queries via the Serving component which combines the results from different fitted models. The Evaluation component is used to compute an appropriate metric to test the performance of a fitted model, as well as aid in the tuning of model hyper parameters.</p><p>This engine template is meant to handle text classification which means you will be worki ng with text data. This means that a query, or newly observed documents, will be of the form:</p><p><code>{text : String}</code>.</p><p>In the running example, a query would be an incoming news article. Once the engine is deployed it can process the query, and then return a Predicted Result of the form</p><p><code>{category : String, confidence : Double}</code>.</p><p>Here category is the model's class assignment for this new text document (i.e. the best guess for this article's categorization), and confidence, a value between 0 and 1 representing your confidence in the category prediction (0 meaning you have no confidence in the prediction). The Actual Result is of the form</p><p><code>{category : String}</code>.</p><p>This is used in the evaluation stage when estimating the performance of your predictive model (how well does the model predict categories). Please refer to the <a href="https://predictionio.apache.org/customize/";>following tutorial</a> for a more detailed exp lanation of how your engine will interact with your web application, as well as an in depth-overview of DASE.</p><h2 id='quick-start' class='header-anchors'>Quick Start</h2><p>This is a quick start guide in case you want to start using the engine right away. Sample email data for spam classification will be used. For more detailed information, read the subsequent sections.</p><h3 id='1.-create-a-new-application.' class='header-anchors'>1. Create a new application.</h3><p>After the application is created, you will be given an access key and application ID for the application.</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><span class="gp">$ </span>pio app new MyTextApp </pre></td></tr></tbody></table> </div> <h3 id='2.-import-the-tutorial-data.' class='header-anchors'>2. Import the tutorial data.</h3><p>There are three different data sets available, each giving a different use case for this engine. Please refer to the <strong>Data Source: Reading Event Data</strong> section to see how to appropriate modify the <code>DataSource</code> class for use with each respective data set. The default data set is an e-mail spam data set.</p><p>These data sets have already been processed and are ready for <a href="/datacollection/batchimport/">batch import</a>. Replace <code>***</code> with your actual application ID.</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</pre></td><td class="code"><pre><span class="gp">$ </span>pio import --appid <span class="k">***</span> --input data/stopwords.json @@ -72,7 +72,7 @@ </pre></td></tr></tbody></table> </div> <p>you should see following outputs returned by the engine:</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><span class="o">{</span><span class="s2">"category"</span>:<span class="s2">"spam"</span>,<span class="s2">"confidence"</span>:0.5268770133242983<span class="o">}</span> </pre></td></tr></tbody></table> </div> <h3 id='5.b.evaluate-your-training-model-and-tune-parameters.' class='header-anchors'>5.b.Evaluate your training model and tune parameters.</h3><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><span class="gp">$ </span>pio <span class="nb">eval </span>org.template.textclassification.AccuracyEvaluation org.template.textclassification.EngineParamsList </pre></td></tr></tbody></table> </div> <p><strong>Note:</strong> Training and evaluation stages are generally different stages of engine development. Evaluation is there to help you choose the best <a href="/evaluation/paramtuning/">algorithm parameters</a> to use for training an engine that is to be deployed as a web service.</p><p>Depending on your needs, in steps (5.x.) above, you can configure your Spark settings by typing a command of the form:</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><span class="gp">$ </span>pio <span class="nb">command </span>command_parameters -- --master url --driver-memory <span class="o">{</span>0<span class="o">}</span>G --executor-memory <span class="o">{</span>1<span class="o">}</span>G --conf spark.akka.framesize<span class="o">={</span>2<span class="o">}</span> --total_executor_cores <span class="o">{</span>3<span class="o">}</span> -</pre></td></tr></tbody></table> </div> <p>Only the latter commands are listed as these are some of the more commonly modified values. See the <a href="https://spark.apache.org/docs/latest/spark-standalone.html";>Spark documentation</a> and the <a href="http://predictionio.incubator.apache.org/resources/faq/";>PredictionIO FAQ's</a> for more information.</p><p><strong>Note:</strong> We recommend you set your driver memory to <code>1G</code> or <code>2G</code> as the data size when dealing with text can be very large.</p><h1 id='detailed-explanation-of-dase' class='header-anchors'>Detailed Explanation of DASE</h1><h2 id='importing-data' class='header-anchors'>Importing Data</h2><p>In the quick start, email spam classification is used. This template can easily be modified for other types text classification.</p><p>If you want to import different sets of data, follow the Quick Start instructions to import data from different files. Make sure that the Data Source is modified according ly to match the <code>event</code>, <code>entityType</code>, and <code>properties</code> fields set for the specific dataset. The following section explains this in more detail.</p><h2 id='data-source:-reading-event-data' class='header-anchors'>Data Source: Reading Event Data</h2><p>Now that the data has been imported into PredictionIO's Event Server, it needs to be read from storage to be used by the engine. This is precisely what the DataSource engine component is for, which is implemented in the template script <code>DataSource.scala</code>. The class <code>Observation</code> serves as a wrapper for storing the information about a news document needed to train a model. The attribute label refers to the label of the category a document belongs to, and text, stores the actual document content as a string. The class TrainingData is used to store an RDD of Observation objects along with the set of stop words.</p><p>The class <code>DataSourceParams</code> is used to specify the pa rameters needed to read and prepare the data for processing. This class is initialized with two parameters <code>appName</code> and <code>evalK</code>. The first parameter specifies your application name (i.e. MyTextApp), which is needed so that the DataSource component knows where to pull the event data from. The second parameter is used for model evaluation and specifies the number of folds to use in <a href="http://en.wikipedia.org/wiki/Cross-validation_%28statistics%29";>cross-validation</a> when estimating a model performance metric.</p><p>The final and most important ingredient is the DataSource class. This is initialized with its corresponding parameter class, and extends <code>PDataSource</code>. This <strong>must</strong> implement the method <code>readTraining</code> which returns an instance of type TrainingData. This method completely relies on the defined private methods readEventData and readStopWords. Both of these functions read data observations as Event instances, c reate an RDD containing these events and finally transforms the RDD of events into an object of the appropriate type as seen below:</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1 +</pre></td></tr></tbody></table> </div> <p>Only the latter commands are listed as these are some of the more commonly modified values. See the <a href="https://spark.apache.org/docs/latest/spark-standalone.html";>Spark documentation</a> and the <a href="http://predictionio.apache.org/resources/faq/";>PredictionIO FAQ's</a> for more information.</p><p><strong>Note:</strong> We recommend you set your driver memory to <code>1G</code> or <code>2G</code> as the data size when dealing with text can be very large.</p><h1 id='detailed-explanation-of-dase' class='header-anchors'>Detailed Explanation of DASE</h1><h2 id='importing-data' class='header-anchors'>Importing Data</h2><p>In the quick start, email spam classification is used. This template can easily be modified for other types text classification.</p><p>If you want to import different sets of data, follow the Quick Start instructions to import data from different files. Make sure that the Data Source is modified accordingly to matc h the <code>event</code>, <code>entityType</code>, and <code>properties</code> fields set for the specific dataset. The following section explains this in more detail.</p><h2 id='data-source:-reading-event-data' class='header-anchors'>Data Source: Reading Event Data</h2><p>Now that the data has been imported into PredictionIO's Event Server, it needs to be read from storage to be used by the engine. This is precisely what the DataSource engine component is for, which is implemented in the template script <code>DataSource.scala</code>. The class <code>Observation</code> serves as a wrapper for storing the information about a news document needed to train a model. The attribute label refers to the label of the category a document belongs to, and text, stores the actual document content as a string. The class TrainingData is used to store an RDD of Observation objects along with the set of stop words.</p><p>The class <code>DataSourceParams</code> is used to specify the parameters n eeded to read and prepare the data for processing. This class is initialized with two parameters <code>appName</code> and <code>evalK</code>. The first parameter specifies your application name (i.e. MyTextApp), which is needed so that the DataSource component knows where to pull the event data from. The second parameter is used for model evaluation and specifies the number of folds to use in <a href="http://en.wikipedia.org/wiki/Cross-validation_%28statistics%29";>cross-validation</a> when estimating a model performance metric.</p><p>The final and most important ingredient is the DataSource class. This is initialized with its corresponding parameter class, and extends <code>PDataSource</code>. This <strong>must</strong> implement the method <code>readTraining</code> which returns an instance of type TrainingData. This method completely relies on the defined private methods readEventData and readStopWords. Both of these functions read data observations as Event instances, create an R DD containing these events and finally transforms the RDD of events into an object of the appropriate type as seen below:</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 @@ -386,7 +386,7 @@ </pre></td></tr></tbody></table> </div> <p>The last and final object implemented in this class simply creates a Map with keys being class labels and values, the corresponding category.</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</pre></td><td class="code"><pre> <span class="c1">// 5. Finally extract category map, associating label to category. </span> <span class="k">val</span> <span class="n">categoryMap</span> <span class="k">=</span> <span class="n">td</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">map</span><span class="o">(</span><span class="n">e</span> <span class="k">=></span> <span class="o">(</span><span class="n">e</span><span class="o">.</span><span class="n">label</span><span class="o">,</span> <span class="n">e</span><span class="o">.</span><span class="n">category</span><span class="o">)).</span><span class="n">collectAsMap</span> -</pre></td></tr></tbody></table> </div> <h2 id='algorithm-component' class='header-anchors'>Algorithm Component</h2><p>The algorithm components in this engine, <code>NBAlgorithm</code> and <code>LRAlgorithm</code>, actually follows a very general form. Firstly, a parameter class must again be initialized to feed in the corresponding Algorithm model parameters. For example, NBAlgorithm incorporates NBAlgorithmParams which holds the appropriate additive smoothing parameter lambda for the Naive Bayes model.</p><p>The main class of interest in this component is the class that extends <a href="https://predictionio.incubator.apache.org/api/current/#org.apache.predictionio.controller.P2LAlgorithm";>P2LAlgorithm</a>. This class must implement a method named train which will output your predictive model (as a concrete object, this will be implemented via a Scala class). It must also implement a predict method that transforms a query to an appropriate feature vector, and uses this to predict w ith the fitted model. The vectorization function is implemented by a PreparedData object, and the categorization (prediction) is handled by an instance of the NBModel implementation. Again, this demonstrates the facility with which different models can be incorporated into PredictionIO's DASE architecture.</p><p>The model class itself will be discussed in the following section, however, turn your attention to the TextManipulationEngine object defined in the script <code>Engine.scala</code>. You can see here that the engine is initialized by specifying the DataSource, Preparator, and Serving classes, as well as a Map of algorithm names to Algorithm classes. This tells the engine which algorithms to run. In practice, you can have as many statistical learning models as you'd like, you simply have to implement a new algorithm component to do this. However, this general design form will persist, and the main meat of the work should be in the implementation of your model class.</p ><p>The following subsection will go over our Naive Bayes implementation in >NBModel.</p><h3 id='naive-bayes-classification' class='header-anchors'>Naive >Bayes Classification</h3><p>This Training Model class only uses the >Multinomial Naive Bayes <a >href="https://spark.apache.org/docs/latest/mllib-naive-bayes.html";>implementation</a> > found in the Spark MLLib library. However, recall that the predicted results >required in the specifications listed in the overview are of the >form:</p><p><code>{category: String, confidence: Double}</code>.</p><p>The >confidence value should really be interpreted as the probability that a >document belongs to a category given its vectorized form. Note that >MLLib's Naive Bayes model has the class members pi (\(\pi\)), and theta >(\(\theta\)). \(\pi\) is a vector of log prior class probabilities, which >shows your prior beliefs regarding the probability that an arbitrary document >belongs in a category. \(\theta\) is a C \(\times\) D matrix, where C is the n umber of classes, and D, the number of features, giving the log probabilities that parametrize the Multinomial likelihood model assumed for each class. The multinomial model is easiest to think about as a problem of randomly throwing balls into bins, where the ball lands in each bin with a certain probability. The model treats each n-gram as a bin, and the corresponding t.f.-i.d.f. value as the number of balls thrown into it. The likelihood is the probability of observing a (vectorized) document given that it comes from a particular class.</p><p>Now, letting \(\mathbf{x}\) be a vectorized text document, then it can be shown that the vector</p><p>$$ \frac{\exp\left(\pi + \theta\mathbf{x}\right)}{\left|\left|\exp\left(\pi + \theta\mathbf{x}\right)\right|\right|} $$</p><p>is a vector with C components that represent the posterior class membership probabilities for the document given \(\mathbf{x}\). That is, the update belief regarding what category this document belongs to after observ ing its vectorized form. This is the motivation behind defining the class NBModel which uses Spark MLLib's NaiveBayesModel, but implements a separate prediction method.</p><p>The private methods innerProduct and getScores are implemented to do the matrix computation above.</p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1 +</pre></td></tr></tbody></table> </div> <h2 id='algorithm-component' class='header-anchors'>Algorithm Component</h2><p>The algorithm components in this engine, <code>NBAlgorithm</code> and <code>LRAlgorithm</code>, actually follows a very general form. Firstly, a parameter class must again be initialized to feed in the corresponding Algorithm model parameters. For example, NBAlgorithm incorporates NBAlgorithmParams which holds the appropriate additive smoothing parameter lambda for the Naive Bayes model.</p><p>The main class of interest in this component is the class that extends <a href="https://predictionio.apache.org/api/current/#org.apache.predictionio.controller.P2LAlgorithm";>P2LAlgorithm</a>. This class must implement a method named train which will output your predictive model (as a concrete object, this will be implemented via a Scala class). It must also implement a predict method that transforms a query to an appropriate feature vector, and uses this to predict with the fi tted model. The vectorization function is implemented by a PreparedData object, and the categorization (prediction) is handled by an instance of the NBModel implementation. Again, this demonstrates the facility with which different models can be incorporated into PredictionIO's DASE architecture.</p><p>The model class itself will be discussed in the following section, however, turn your attention to the TextManipulationEngine object defined in the script <code>Engine.scala</code>. You can see here that the engine is initialized by specifying the DataSource, Preparator, and Serving classes, as well as a Map of algorithm names to Algorithm classes. This tells the engine which algorithms to run. In practice, you can have as many statistical learning models as you'd like, you simply have to implement a new algorithm component to do this. However, this general design form will persist, and the main meat of the work should be in the implementation of your model class.</p><p>The fo llowing subsection will go over our Naive Bayes implementation in NBModel.</p><h3 id='naive-bayes-classification' class='header-anchors'>Naive Bayes Classification</h3><p>This Training Model class only uses the Multinomial Naive Bayes <a href="https://spark.apache.org/docs/latest/mllib-naive-bayes.html";>implementation</a> found in the Spark MLLib library. However, recall that the predicted results required in the specifications listed in the overview are of the form:</p><p><code>{category: String, confidence: Double}</code>.</p><p>The confidence value should really be interpreted as the probability that a document belongs to a category given its vectorized form. Note that MLLib's Naive Bayes model has the class members pi (\(\pi\)), and theta (\(\theta\)). \(\pi\) is a vector of log prior class probabilities, which shows your prior beliefs regarding the probability that an arbitrary document belongs in a category. \(\theta\) is a C \(\times\) D matrix, where C is the number of c lasses, and D, the number of features, giving the log probabilities that parametrize the Multinomial likelihood model assumed for each class. The multinomial model is easiest to think about as a problem of randomly throwing balls into bins, where the ball lands in each bin with a certain probability. The model treats each n-gram as a bin, and the corresponding t.f.-i.d.f. value as the number of balls thrown into it. The likelihood is the probability of observing a (vectorized) document given that it comes from a particular class.</p><p>Now, letting \(\mathbf{x}\) be a vectorized text document, then it can be shown that the vector</p><p>$$ \frac{\exp\left(\pi + \theta\mathbf{x}\right)}{\left|\left|\exp\left(\pi + \theta\mathbf{x}\right)\right|\right|} $$</p><p>is a vector with C components that represent the posterior class membership probabilities for the document given \(\mathbf{x}\). That is, the update belief regarding what category this document belongs to after observing its ve ctorized form. This is the motivation behind defining the class NBModel which uses Spark MLLib's NaiveBayesModel, but implements a separate prediction method.</p><p>The private methods innerProduct and getScores are implemented to do the matrix computation above.</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 @@ -513,7 +513,7 @@ </span><span class="p">}</span><span class="w"> </span><span class="p">]</span><span class="w"> </span><span class="p">}</span><span class="w"> -</span></pre></td></tr></tbody></table> </div> <h2 id='serving:-delivering-the-final-prediction' class='header-anchors'>Serving: Delivering the Final Prediction</h2><p>The serving component is the final stage in the engine, and in a sense, the most important. This is the final stage in which you combine the results obtained from the different models you choose to run. The Serving class extends the <a href="https://predictionio.incubator.apache.org/api/current/#org.apache.predictionio.controller.LServing";>LServing</a> class which must implement a method called serve. This takes a query and an associated sequence of predicted results, which contains the predicted results from the different algorithms that are implemented in your engine, and combines the results to yield a final prediction. It is this final prediction that you will receive after sending a query.</p><p>For example, you could choose to slightly modify the implementation to return class probabilities coming from a mixture of model estimates for class probabilities, or any other technique you could conceive for combining your results. The default engine setting has this set to yield the label from the model predicting with greater confidence.</p><h2 id='evaluation:-model-assessment-and-selection' class='header-anchors'>Evaluation: Model Assessment and Selection</h2><p> A predictive model needs to be evaluated to see how it will generalize to future observations. PredictionIO uses cross-validation to perform model performance metric estimates needed to assess your particular choice of model. The script <code>Evaluation.scala</code> available with the engine template exemplifies what a usual evaluator setup will look like. First, you must define an appropriate metric. In the engine template, since the topic is text classification, the default metric implemented is category accuracy.</p><p> Second you must define an evaluation object (i.e. extends the class <a href="https://predictionio.incubator.apache .org/api/current/#org.apache.predictionio.controller.Evaluation">Evaluation</a>). Here, you must specify the actual engine and metric components that are to be used for the evaluation. In the engine template, the specified engine is the TextManipulationEngine object, and metric, Accuracy. Lastly, you must specify the parameter values that you want to test in the cross validation. You see in the following block of 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 +</span></pre></td></tr></tbody></table> </div> <h2 id='serving:-delivering-the-final-prediction' class='header-anchors'>Serving: Delivering the Final Prediction</h2><p>The serving component is the final stage in the engine, and in a sense, the most important. This is the final stage in which you combine the results obtained from the different models you choose to run. The Serving class extends the <a href="https://predictionio.apache.org/api/current/#org.apache.predictionio.controller.LServing";>LServing</a> class which must implement a method called serve. This takes a query and an associated sequence of predicted results, which contains the predicted results from the different algorithms that are implemented in your engine, and combines the results to yield a final prediction. It is this final prediction that you will receive after sending a query.</p><p>For example, you could choose to slightly modify the implementation to return class probabilities coming from a mixture of model estimates for class probabilities, or any other technique you could conceive for combining your results. The default engine setting has this set to yield the label from the model predicting with greater confidence.</p><h2 id='evaluation:-model-assessment-and-selection' class='header-anchors'>Evaluation: Model Assessment and Selection</h2><p> A predictive model needs to be evaluated to see how it will generalize to future observations. PredictionIO uses cross-validation to perform model performance metric estimates needed to assess your particular choice of model. The script <code>Evaluation.scala</code> available with the engine template exemplifies what a usual evaluator setup will look like. First, you must define an appropriate metric. In the engine template, since the topic is text classification, the default metric implemented is category accuracy.</p><p> Second you must define an evaluation object (i.e. extends the class <a href="https://predictionio.apache.org/api/current/#or g.apache.predictionio.controller.Evaluation">Evaluation</a>). Here, you must specify the actual engine and metric components that are to be used for the evaluation. In the engine template, the specified engine is the TextManipulationEngine object, and metric, Accuracy. Lastly, you must specify the parameter values that you want to test in the cross validation. You see in the following block of 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 @@ -542,12 +542,12 @@ <span class="o">)</span> </pre></td></tr></tbody></table> </div> <h2 id='engine-deployment' class='header-anchors'>Engine Deployment</h2><p>Once an engine is ready for deployment it can interact with your web application in real-time. This section will cover how to send and receive queries from your engine, gather more data, and re-training your model with the newly gathered data.</p><h3 id='sending-queries' class='header-anchors'>Sending Queries</h3><p>Recall that one of the greatest advantages of using the PredictionIO platform is that once your engine is deployed, you can respond to queries in real-time. Recall that our queries are of the form</p><p><code>{"text" : "..."}</code>.</p><p>To actually send a query you can use our REST API by typing in the following shell command:</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>curl -H <span class="s2">"Content- Type: application/json"</span> -d <span class="s1">'{ "text":"I like speed and fast motorcycles." }'</span> http://localhost:8000/queries.json </pre></td></tr></tbody></table> </div> <p>There are a number of <a href="https://github.com/PredictionIO";>SDK's</a> you can use to send your queries and obtain a response. Recall that our predicted response is of the form</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><span class="o">{</span><span class="s2">"category"</span> : <span class="s2">"class"</span>, <span class="s2">"confidence"</span> : 1.0<span class="o">}</span> -</pre></td></tr></tbody></table> </div> <p>which is what you should see upon inputting the latter command for querying.</p><h3 id='gathering-more-data-and-retraining-your-model' class='header-anchors'>Gathering More Data and Retraining Your Model</h3><p>The importing data section that is included in this tutorial uses a sample data set for illustration purposes, and uses the PredictionIO Python SDK to import the data. However, there are a variety of ways that you can <a href="//predictionio.incubator.apache.org/datacollection/eventapi/">import</a> your collected data (via REST or other SDKs).</p><p>As you continue to collect your data, it is quite easy to retrain your model once you actually import your data into the Event Server. You simply repeat the steps listed in the Quick Start guide. We re-list them here again:</p><p><strong>1.</strong> Build your engine.</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><span class="gp">$ </span>pio build +</pre></td></tr></tbody></table> </div> <p>which is what you should see upon inputting the latter command for querying.</p><h3 id='gathering-more-data-and-retraining-your-model' class='header-anchors'>Gathering More Data and Retraining Your Model</h3><p>The importing data section that is included in this tutorial uses a sample data set for illustration purposes, and uses the PredictionIO Python SDK to import the data. However, there are a variety of ways that you can <a href="//predictionio.apache.org/datacollection/eventapi/">import</a> your collected data (via REST or other SDKs).</p><p>As you continue to collect your data, it is quite easy to retrain your model once you actually import your data into the Event Server. You simply repeat the steps listed in the Quick Start guide. We re-list them here again:</p><p><strong>1.</strong> Build your engine.</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre cla ss="lineno">1</pre></td><td class="code"><pre><span class="gp">$ </span>pio build </pre></td></tr></tbody></table> </div> <p><strong>2.a.</strong> Evaluate your training model and tune parameters.</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><span class="gp">$ </span>pio <span class="nb">eval </span>org.template.textclassification.AccuracyEvaluation org.template.textclassification.EngineParamsList </pre></td></tr></tbody></table> </div> <p><strong>2.b.</strong> Train your model and deploy.</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><span class="gp">$ </span>pio train <span class="gp">$ </span>pio deploy -</pre></td></tr></tbody></table> </div> </div></div></div></div><footer><div class="container"><div class="seperator"></div><div class="row"><div class="col-md-6 footer-link-column"><div class="footer-link-column-row"><h4>Community</h4><ul><li><a href="//predictionio.incubator.apache.org/install/" target="blank">Download</a></li><li><a href="//predictionio.incubator.apache.org/" target="blank">Docs</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">GitHub</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to User Mailing List</a></li><li><a href="//stackoverflow.com/questions/tagged/predictionio" target="blank">Stackoverflow</a></li></ul></div></div><div class="col-md-6 footer-link-column"><div class="footer-link-column-row"><h4>Contribute</h4><ul><li><a href="//predictionio.incubator.apache.org/community/contribute-code/" target="blank">Contribute</a></li><li><a href="//github.com/apache/incubator-pred ictionio" target="blank">Source Code</a></li><li><a href="//issues.apache.org/jira/browse/PIO" target="blank">Bug Tracker</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to Development Mailing List</a></li></ul></div></div></div><div class="row"><div class="col-md-12 footer-link-column"><p>Apache PredictionIO, PredictionIO, Apache, the Apache feather logo, and the Apache PredictionIO project logo are either registered trademarks or trademarks of The Apache Software Foundation in the United States and other countries.</p><p>All other marks mentioned may be trademarks or registered trademarks of their respective owners.</p></div></div><div class="row"><div class="col-md-12 footer-link-column"><a class="pull-right" href="http://incubator.apache.org/projects/predictionio.html";><img alt="Apache Incubator" src="/images/logos/apache_incubator-6954bd16.png"/></a><span>Apache PredictionIO is an effort undergoing incubation at The Apache S oftware Foundation (ASF), sponsored by the Apache Incubator. Incubation is required of all newly accepted projects until a further review indicates that the infrastructure, communications, and decision making process have stabilized in a manner consistent with other successful ASF projects. While incubation status is not necessarily a reflection of the completeness or stability of the code, it does indicate that the project has yet to be fully endorsed by the ASF.</span></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"/><span>â¢</span></div><div id="social-icons-wrapper"><a class="github-button" href="https://github.com/apache/incubator-predictionio"; data-style="mega" data-count-href="/apache/incubator-predictionio/stargazers" data-count-api="/repos/apache/incubator-predictionio#stargazers_count" data-count-aria-label="# stargazers on GitHub" aria-label="Star apache/incubator-predictionio on GitHub">Star</a> <a class="github-button" href="https://github.com/apache/incubator-predictionio/fork"; data-icon="octicon-git-branch" data-style="mega" data-count-href="/apache/incubator-predictionio/network" data-count-api="/repos/apache/incubator-predictionio#forks_count" data-count-aria-label="# forks on GitHub" aria-label="Fork apache/incubator-predictionio on GitHub">Fork</a> <script id="github-bjs" async="" defer="" src="https://buttons.github.io/buttons.js";></script><a href="https://twitter.com/predictionio"; target="blank"><img alt="PredictionIO on Twitter" src="/images/icons/twitter-ea9dc152.png"/></a> <a href="https://www.facebook.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(){ +</pre></td></tr></tbody></table> </div> </div></div></div></div><footer><div class="container"><div class="seperator"></div><div class="row"><div class="col-md-6 footer-link-column"><div class="footer-link-column-row"><h4>Community</h4><ul><li><a href="//predictionio.apache.org/install/" target="blank">Download</a></li><li><a href="//predictionio.apache.org/" target="blank">Docs</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">GitHub</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to User Mailing List</a></li><li><a href="//stackoverflow.com/questions/tagged/predictionio" target="blank">Stackoverflow</a></li></ul></div></div><div class="col-md-6 footer-link-column"><div class="footer-link-column-row"><h4>Contribute</h4><ul><li><a href="//predictionio.apache.org/community/contribute-code/" target="blank">Contribute</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">Source Code</a> </li><li><a href="//issues.apache.org/jira/browse/PIO" target="blank">Bug Tracker</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to Development Mailing List</a></li></ul></div></div></div><div class="row"><div class="col-md-12 footer-link-column"><p>Apache PredictionIO, PredictionIO, Apache, the Apache feather logo, and the Apache PredictionIO project logo are either registered trademarks or trademarks of The Apache Software Foundation in the United States and other countries.</p><p>All other marks mentioned may be trademarks or registered trademarks of their respective owners.</p></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"/><span>â¢</span></div><div id="social-icons-wrapper"><a class="github-button" href="https://github.com/apache/incubator-predictionio"; data-style="mega" data-cou nt-href="/apache/incubator-predictionio/stargazers" data-count-api="/repos/apache/incubator-predictionio#stargazers_count" data-count-aria-label="# stargazers on GitHub" aria-label="Star apache/incubator-predictionio on GitHub">Star</a> <a class="github-button" href="https://github.com/apache/incubator-predictionio/fork"; data-icon="octicon-git-branch" data-style="mega" data-count-href="/apache/incubator-predictionio/network" data-count-api="/repos/apache/incubator-predictionio#forks_count" data-count-aria-label="# forks on GitHub" aria-label="Fork apache/incubator-predictionio on GitHub">Fork</a> <script id="github-bjs" async="" defer="" src="https://buttons.github.io/buttons.js";></script><a href="https://twitter.com/predictionio"; target="blank"><img alt="PredictionIO on Twitter" src="/images/icons/twitter-ea9dc152.png"/></a> <a href="https://www.facebook.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');
http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/452034c1/demo/textclassification/index.html.gz ---------------------------------------------------------------------- diff --git a/demo/textclassification/index.html.gz b/demo/textclassification/index.html.gz index 846214d..60635c4 100644 Binary files a/demo/textclassification/index.html.gz and b/demo/textclassification/index.html.gz differ
