Enabling to output the rating predictions

librec/src/main/java/librec/data/DataDAO.java

@@ -275,6 +275,10 @@ public void writeData(String toPath, String sep) throws Exception {
 
 		List<String> lines = new ArrayList<>(1500);
 		for (MatrixEntry me : rateMatrix) {
+
+			if (me.get() <= 0) // enable sparse output
+				continue;
+
 			String line = Strings.toString(
 					new Object[] { me.row() + 1, me.column() + 1,
 							(float) me.get() }, sep);
@@ -543,6 +547,13 @@ public SparseMatrix getRateMatrix() {
 		return rateMatrix;
 	}
 
+	/**
+	 * @param ratingMatrix
+	 */
+	public void setRateMatrix(SparseMatrix ratingMatrix) {
+		this.rateMatrix = ratingMatrix;
+	}
+
 	/**
 	 * @return whether "items" are users, useful for social reltions
 	 */

librec/src/main/java/librec/intf/Recommender.java

@@ -429,6 +429,12 @@ private Map<Measure, Double> evalRatings() {
 
 		double sum_maes = 0, sum_mses = 0, sum_asyms = 0;
 		int numCount = 0;
+
+		//cf.getString("printTestPredictions");
+
+		SparseMatrix predMatrix = testMatrix.clone();
+
+
 		for (MatrixEntry me : testMatrix) {
 			double rate = me.get();
 			if (rate <= 0)
@@ -443,9 +449,10 @@ private Map<Measure, Double> evalRatings() {
 			double pred = predict(u, j, true);
 			if (Double.isNaN(pred))
 				continue;
-
+
+			predMatrix.set(u, j, pred);
 			double err = rate - pred;
-
+			
 			sum_maes += Math.abs(err);
 			sum_mses += err * err;
 			sum_asyms += Measures.ASYMMLoss(rate, pred, minRate, maxRate);
@@ -463,6 +470,21 @@ private Map<Measure, Double> evalRatings() {
 		measures.put(Measure.RMSE, rmse);
 		measures.put(Measure.ASYMM, asymm);
 
+		if (cf.isOn("has.predictions.rating.output")) {
+			DataDAO testDao = new DataDAO(cf.getPath("dataset.testing"), rateDao.getUserIds(), rateDao.getItemIds());
+			testDao.setRateMatrix(predMatrix);
+			try {
+				String foldString = "";
+				if (this.fold>-1) {
+					foldString = "_"+Integer.toString(this.fold);
+				}
+				testDao.writeData(cf.getPath("predictions.rating.output") + foldString);
+			} catch (Exception e) {
+				// TODO Auto-generated catch block
+				e.printStackTrace();
+			}
+		}
+
 		return measures;
 	}
 

librec/src/main/resources/librec.conf

@@ -1,13 +1,13 @@
 ################################################### General Settings #############################################
 # dataset: item ratings and user social inforamtion
-dataset.training.wins=D:\\path\\to\\dataset\\ratings.txt
+dataset.training.wins=D:\\temp\\ratings\\filmtrust-ratings.txt
 dataset.training.lins=/home/path/to/datasets/ratings.txt
 
 dataset.social.wins=-1
 dataset.social.lins=-1
 
-dataset.testing.wins=-1
-dataset.testing.lins=-1
+dataset.testing.wins=D:\\temp\\ratings\\filmtrust-test.txt
+dataset.testing.lins=/home/path/to/datasets/ratings.txt
 
 # Item Recommendation: threshold to convert rating values to binary ones (during data loading); 
 # Note that not all item recommendation models require binary ratings; -1: disable conversion
@@ -27,7 +27,7 @@ mail.smtp.password=xxxx
 mail.to=xxx@email.address
 
 # validation priority: 1. cross-validation; 2. (training) ratio; 3. given n; 4: given ratio
-is.cross.validation=on
+is.cross.validation=off
 is.parallel.folds=on
 num.kfold=5
 val.ratio=0.2
@@ -43,10 +43,15 @@ is.diverse.used=off
 num.reclist.len=-1
 num.ignor.items=-1
 
+
+# prediction output
+has.predictions.rating.output = on
+predictions.rating.output.wins =D:\\temp\\ratings\\predictions.txt
+
 # baseline & exts: GlobalAvg, UserAvg, ItemAvg, Random, Constant, MostPop; NMF, SlopeOne, Hybrid, PD, AR, PRankD; 
 # ranking: CLiMF, WRMF, BPR, RankALS, RankSGD; 
 # rating:  UserKNN, ItemKNN, RegSVD, BiasedMF, PMF, BPMF, SVD++, SocialMF, RSTE, TrustMF, SoRec, SoReg; 
-recommender=SVD++
+recommender=UserKNN
 
 ################################################### Model-based Methods ##########################################
 # general parameters for matrix factorization