{"id":15124,"date":"2014-09-10T11:52:39","date_gmt":"2014-09-10T06:22:39","guid":{"rendered":"https:\/\/2thenew.today\/blog\/?p=15124"},"modified":"2016-12-19T14:54:44","modified_gmt":"2016-12-19T09:24:44","slug":"recommendation-engine-by-using-apache-mahout-cassandra","status":"publish","type":"post","link":"https:\/\/2thenew.today\/blog\/recommendation-engine-by-using-apache-mahout-cassandra\/","title":{"rendered":"Recommendation Engine by using Apache Mahout & Cassandra"},"content":{"rendered":"

As a part of R&D we are trying to build a recommendations for users by using Apache Mahout & Cassandra.There was a direct Cassandra\u00a0 DataModel<\/code>CassandraDataModel based on a Cassandra keyspace, but there was a lot of confusions which involves so many methods and functions,then we are tried to minimize it by using Cassandra Indexes\u00a0 and Mahout FileDataModel<\/p>\n

Here we are having 1 millions of Records in Cassandra Database in a testkeyspace with 2 tables and their schema are<\/p>\n

CREATE TABLE SampleUser (
\nuser_id bigint,
\nratings float,
\nitem_id bigint,
\nPRIMARY KEY (user_id, ratings, item_id)
\n)<\/p>\n

CREATE TABLE SampleItems (
\nitem_id bigint,
\nuser_id bigint,
\nratings float,
\nPRIMARY KEY (item_id, user_id)
\n)<\/p>\n

We want to provide User based Recommendations, Item based recommendation with help of similar users (For example, based on the rating giving by other users in a common locality or other similarity).<\/p>\n

They are<\/p>\n

1)Customers Who Bought This Item Also Bought\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 2)Customers Who Viewed This Item Also Viewed<\/p>\n

The java code for Mahout Recommendations in Cassandra Database<\/p>\n

[html]
\npublic class RecommendationsFromCassandra {
\nprivate Cluster cluster;
\nprivate Session session;<\/p>\n

\/*
\n* Connection to Cassandra Database
\n*\/<\/p>\n

public void connect(String node) {
\ncluster = Cluster.builder().addContactPoint(node).build();
\nMetadata metadata = cluster.getMetadata();
\nsession = cluster.connect();
\n}<\/p>\n

\/*
\n* Executing the query to get the user ids who are having particular item
\n*\/<\/p>\n

public void query1() throws IOException {
\nString userids = null;
\nResultSet results = session
\n.execute("SELECT user_id FROM testkeyspace.sampleitems WHERE item_id = 914");
\nfor (Row row : results)
\n {
\nif (userids == null)
\nuserids = Long.toString(row.getLong("user_id"));
\nelse
\nuserids = userids + "," + row.getLong("user_id");
\n}<\/p>\n

\/*
\n* Transfer the all user ids to the 2nd query to get the all users and
\n* their item id’s , Ratings
\n*\/<\/p>\n

String stmt = "SELECT user_id,item_id,ratings FROM testkeyspace.sampleuser WHERE user_id in ("
\n+ userids + ") ;";<\/p>\n

Statement s1 = new SimpleStatement(stmt);
\ns1.setFetchSize(Integer.MAX_VALUE);<\/p>\n

ResultSet results1 = session.execute(s1);
\nFile newTextFile = new File("${env:HOME}\/thetextfile.txt");
\nFileWriter fw = new FileWriter(newTextFile);<\/p>\n

for (Row row2 : results1) {
\nfw.write(row2.getLong("user_id") + "\\t" + row2.getLong("item_id")
\n+ "\\t" + row2.getFloat("ratings") + "\\n");
\n}
\nfw.close();<\/p>\n

\/*
\n* Recommendation Part
\n*\/<\/p>\n

UserSimilarity similarity;
\ntry {
\nDataModel datamodel = new FileDataModel(new File(
\n"${env:HOME}\/thetextfile.txt"));
\nsimilarity = new PearsonCorrelationSimilarity(datamodel);
\nUserNeighborhood neighbourhood = new NearestNUserNeighborhood(100,
\nsimilarity, datamodel);<\/p>\n

Recommender recommender = new GenericUserBasedRecommender(
\ndatamodel, neighbourhood, similarity);
\nlong start = System.currentTimeMillis();
\nList<RecommendedItem> recommendations = recommender.recommend(10,
\n10);
\nfor (RecommendedItem recommendation : recommendations) {
\nSystem.out.println(recommendation);<\/p>\n

}
\nlong stop = System.currentTimeMillis();
\nSystem.out.println("Took: " + (stop – start) + " millis");<\/p>\n

} catch (TasteException e) {
\ne.printStackTrace();
\n}<\/p>\n

}<\/p>\n

public void close() {
\ncluster.close();
\n}<\/p>\n

public static void main(String[] args) throws IOException {
\nRecommendationsFromCassandra client = new RecommendationsFromCassandra();
\nclient.connect("127.0.0.1");
\nclient.query1();
\n\/\/ CassandraRecommender r = new CassandraRecommender();
\nclient.close();
\nFile file = new File("${env:HOME}\/thetextfile.txt");
\nfile.delete();
\n}
\n}
\n[\/html]<\/p>\n


\nThe time consumming for Recommendations to particular user is <\/span><\/span><\/span>2.40<\/span><\/span><\/span> sec for 1m records<\/span><\/span><\/span><\/b><\/p>\n

\u00a0<\/span><\/span><\/span><\/strong><\/p>\n

\u00a0<\/span><\/span><\/span><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"

As a part of R&D we are trying to build a recommendations for users by using Apache Mahout & Cassandra.There was a direct Cassandra\u00a0 DataModelCassandraDataModel based on a Cassandra keyspace, but there was a lot of confusions which involves so many methods and functions,then we are tried to minimize it by using Cassandra Indexes\u00a0 and […]<\/p>\n","protected":false},"author":114,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"iawp_total_views":3,"footnotes":""},"categories":[1395],"tags":[],"class_list":["post-15124","post","type-post","status-publish","format-standard","hentry","category-big-data"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t