Prediction for User-Item using PMFRecommender or NMFRecommender

[srinivasaraov]

I am using the following code to build a model using PMFRecommender. If I want to get all recommendations for a particular user, instead of calling getRecommendedList() and filtering for the user, is there a method, say predict(user) that returns items with score for a particular user?

Also,

• RecommenderSimilarity is not required to be added to the RecommenderContext if we are using Matrix Factorization algorithms, right?

• What configuration parameters are available and possible values for PMFRecommender and NMFRecommender ?

• I get a precision and recall values as following for the following code and config. Is the code correct or is there anything I am missing? Precision: NaN Recall: 0.0

• Also, the recommended list keeps changing for every run? Is there a property I need to set to keep the recommendations constant?

• How can I specify ARFF input instead of text and get the predictions ?

         ```
  Configuration conf = new Configuration();
  Resource resource = new Resource("pmf-test.properties");
  conf.addResource(resource);
  
  // build data model
  DataModel dataModel = new TextDataModel(conf);
  dataModel.buildDataModel();
  
  // set recommendation context
  RecommenderContext context = new RecommenderContext(conf, dataModel);
  // RecommenderSimilarity similarity = new PCCSimilarity();
  // similarity.buildSimilarityMatrix(dataModel);
  // context.setSimilarity(similarity);
  
  // training
  Recommender recommender = new PMFRecommender();
  recommender.recommend(context);
  
  // evaluation
  RecommenderEvaluator maeEvaluator = new MAEEvaluator();
  double mae = recommender.evaluate(maeEvaluator);
  
  RecommenderEvaluator rmseEvaluator = new RMSEEvaluator();
  double rmse = recommender.evaluate(rmseEvaluator);
  
  RecommenderEvaluator precisionEvaluator = new PrecisionEvaluator();
  double precision = recommender.evaluate(precisionEvaluator);
  
  RecommenderEvaluator recallEvaluator = new RecallEvaluator();
  double recall = recommender.evaluate(recallEvaluator);
  
  System.out.println("MAE:" + mae);
  System.out.println("RMSE:" + rmse);
  System.out.println("Precision: " + precision);
  System.out.println("Recall: " + recall);
  
  // recommendation results
  List<RecommendedItem> recommendedItemList = recommender.getRecommendedList();
  List<String> userIds = new ArrayList<String>();
  userIds.add("836");
  RecommendedFilter filter = new GenericRecommendedFilter();
  ((GenericRecommendedFilter) filter).setUserIdList(userIds);
  recommendedItemList = filter.filter(recommendedItemList);
  
  Map<String, Double> recommendedItems = new HashMap<String, Double>();
  System.out.println("Recommendations for User 836");
  for(RecommendedItem item : recommendedItemList) {
      if(item.getUserId().equals("836")) {
          System.out.println(item.getUserId() + " : " + item.getItemId() + " : " + item.getValue());
      }
  }

//Config:

dfs.data.dir=/Users/sv/Datasets/
dfs.result.dir=/Users/sv/Datasets/

data.input.path=ml-100k/ratings.txt
data.column.format=UIR
data.model.splitter=ratio
data.splitter.ratio=rating
data.model.format=text
data.splitter.trainset.ratio=0.8

rec.random.seed=1
data.convert.binarize.threshold=-1.0

rec.eval.enable=true
rec.recommender.isranking=false

rec.recommender.class=pmf
rec.iterator.learnrate=0.01
rec.iterator.learnrate.maximum=0.01
rec.iterator.maximum=50
rec.user.regularization=0.08
rec.item.regularization=0.08
rec.factor.number=6
rec.learnrate.bolddriver=false
rec.learnrate.decay=1.0