@Override
public double evaluate(
JavaSparkContext sparkContext, PMML model, Path modelParentPath, JavaRDD<String> testData) {
log.info("Evaluating model");
JavaRDD<Rating> testRatingData = parsedToRatingRDD(toParsedRDD(testData));
testRatingData = aggregateScores(testRatingData);
MatrixFactorizationModel mfModel = pmmlToMFModel(sparkContext, model, modelParentPath);
double eval;
if (implicit) {
double auc = AUC.areaUnderCurve(sparkContext, mfModel, testRatingData);
log.info("AUC: {}", auc);
eval = auc;
} else {
double rmse = RMSE.rmse(mfModel, testRatingData);
log.info("RMSE: {}", rmse);
eval = 1.0 / rmse;
}
return eval;
}
public void evaluate(List<Sample> sampleList, Algorithm algorithm) {
List<Score> scoreList = new ArrayList<Score>();
for (Sample sample : sampleList) {
scoreList.add(new Score(algorithm.predict(sample.getFeature()), sample.getLabel()));
}
Collections.sort(scoreList, new ComparatorScore());
double counter = 0;
double rankSum = 0;
double positiveSample = 0;
double negtiveSample = 0;
for (Score s : scoreList) {
if (s.label > 0) {
rankSum += scoreList.size() - counter;
positiveSample++;
} else {
negtiveSample++;
}
counter++;
}
// report.
System.out.println("AUC: " + AUC.measure(algorithm, sampleList));
System.out.println(
"AUC: "
+ (positiveSample > 0 && negtiveSample > 0
? 1.0d
* (rankSum - positiveSample * (positiveSample + 1) / 2)
/ (positiveSample * negtiveSample)
: 1));
System.out.println("Positive sample number: " + positiveSample);
System.out.println("Negative sample number: " + negtiveSample);
System.out.println("Max score: " + scoreList.get(0).score);
System.out.println("Middle score: " + scoreList.get(scoreList.size() / 2).score);
System.out.println("Min score: " + scoreList.get(scoreList.size() - 1).score);
}
