Hey guys.
I am currently trying to improve the performance of an Apache Mahout
Recommender, that I am working on. Due to the fact that my users'
preferences are stored in a MYSQL database, I am using the following
code in order to produce recommendations:
MysqlDataSource dataSource = new MysqlDataSource();
dataSource.setServerName("localhost");
dataSource.setUser("user");
dataSource.setPassword("user_pass");
dataSource.setDatabaseName("user_history_DB");
dataSource.setCachePreparedStatements(true);
dataSource.setCachePrepStmts(true);
dataSource.setCacheResultSetMetadata(true);
dataSource.setAlwaysSendSetIsolation(false);
dataSource.setElideSetAutoCommits(true);
JDBCDataModel dataModel = new MySQLJDBCDataModel(dataSource,
"user_preferences", "user_id", "item_id", "preference",
"timestamp");
UserSimilarity userSimilarity =
new PearsonCorrelationSimilarity(dataModel);
UserNeighborhood neighborhood =
new NearestNUserNeighborhood(3, userSimilarity,
dataModel);
Recommender recommender =
new GenericUserBasedRecommender(dataModel,
neighborhood, userSimilarity);
Recommender cachingRecommender = new
CachingRecommender(recommender);
List<RecommendedItem> recommendations =
cachingRecommender.recommend(1234, 10);
Also, I have to inform you that I have used Indexes on the database
table, on both user_id and item_id to increase performance. Moreover, I
have applied the following configuration settings for the MySQL server
(they are mentioned in the MYSQLJDBCDataModel documentation):
* innodb_buffer_pool_size=64M
* myisam_sort_buffer_size=64M
* query_cache_limit=64M
* query_cache_min_res_unit=512K
* query_cache_type=1
* query_cache_size=64M
Finally, I am using Netbeans 7.1 and the project is a standard java
application.
Unfortunately, the performance is very bad. What can I do in order to
improve performance? Am I missing something here?
Thank you,
Nick Katsipoulakis.
