@Test
public void test_perfectPrediction_NMAEequals0() {
// Phase 1: Preparation
RatingsDataset<? extends Rating> testDataset = new RatingsDatasetMock();
RelevanceCriteria relevanceCriteria = new RelevanceCriteria(4);
RecommendationResults recommendationResults = new RecommendationResults();
for (int idUser : testDataset.allUsers()) {
try {
ArrayList<Recommendation> recommendations = new ArrayList<>();
for (Map.Entry<Integer, ? extends Rating> entry :
testDataset.getUserRatingsRated(idUser).entrySet()) {
final int idItem = entry.getKey();
final Rating rating = entry.getValue();
final double prediction = rating.getRatingValue().doubleValue();
recommendations.add(new Recommendation(idItem, prediction));
}
Collections.sort(recommendations);
recommendationResults.add(idUser, recommendations);
} catch (UserNotFound ex) {
// Never happens if the dataset is properly implemented
}
}
NRMSE instance = new NRMSE();
// Phase 2: Execution
MeasureResult result =
instance.getMeasureResult(recommendationResults, testDataset, relevanceCriteria);
// Phase 3: Result checking
double expResult = 0;
double delta = 0.001f;
assertEquals(expResult, result.getValue(), delta);
}
@Test
public void test_perfectPredictionButOneMovedBy3_NMAEequals0dot433() {
// Phase 1: Preparation
RatingsDataset<? extends Rating> testDataset = new RatingsDatasetMock();
RelevanceCriteria relevanceCriteria = new RelevanceCriteria(4);
RecommendationResults recommendationResults = new RecommendationResults();
int idUser = 1;
ArrayList<Recommendation> recommendations = new ArrayList<>();
recommendations.add(new Recommendation(10, 4));
recommendations.add(new Recommendation(12, 5));
recommendations.add(new Recommendation(13, 2));
Collections.sort(recommendations);
recommendationResults.add(idUser, recommendations);
NRMSE instance = new NRMSE();
// Phase 2: Execution
MeasureResult nRMSEResult =
instance.getMeasureResult(recommendationResults, testDataset, relevanceCriteria);
// Phase 3: Result checking
double expResult = 0.4330127;
double result = nRMSEResult.getValue();
double delta = 0.001f;
assertEquals(expResult, result, delta);
}
@Test
public void test_allMovedTwo_NRMSEequals0point5() {
// Phase 1: Preparation
RatingsDataset<? extends Rating> testDataset = new RatingsDatasetMock();
RelevanceCriteria relevanceCriteria = new RelevanceCriteria(4);
RecommendationResults recommendationResults = new RecommendationResults();
int i = 0;
for (int idUser : testDataset.allUsers()) {
try {
ArrayList<Recommendation> recommendations = new ArrayList<>();
for (Map.Entry<Integer, ? extends Rating> entry :
testDataset.getUserRatingsRated(idUser).entrySet()) {
final int idItem = entry.getKey();
final double ratingValue = entry.getValue().getRatingValue().doubleValue();
double prediction;
if (i % 2 == 0) {
prediction = ratingValue + 2;
if (prediction > testDataset.getRatingsDomain().max().doubleValue()) {
prediction = ratingValue - 2;
}
} else {
prediction = ratingValue - 2;
if (prediction < testDataset.getRatingsDomain().min().doubleValue()) {
prediction = ratingValue + 2;
}
}
recommendations.add(new Recommendation(idItem, prediction));
i++;
}
Collections.sort(recommendations);
recommendationResults.add(idUser, recommendations);
} catch (UserNotFound ex) {
// Never happens if the dataset is properly implemented
ERROR_CODES.UNDEFINED_ERROR.exit(ex);
}
}
NRMSE instance = new NRMSE();
// Phase 2: Execution
MeasureResult result =
instance.getMeasureResult(recommendationResults, testDataset, relevanceCriteria);
// Phase 3: Result checking
double expResult = 0.5f;
double delta = 0.001f;
assertEquals(expResult, result.getValue(), delta);
}
/** Test of getMeasureResult method, of class NRMSE. */
@Test
public void test_noRecommendations_NRMSEequalsNaN() {
// Phase 1: Preparation
int idUser = 1;
RatingsDataset<? extends Rating> testDataset = new RatingsDatasetMock();
RelevanceCriteria relevanceCriteria = new RelevanceCriteria(4);
RecommendationResults recommendationResults = new RecommendationResults();
recommendationResults.add(idUser, new ArrayList<>());
NRMSE instance = new NRMSE();
// Phase 2: Execution
MeasureResult result =
instance.getMeasureResult(recommendationResults, testDataset, relevanceCriteria);
// Phase 3: Result checking
double expResult = Double.NaN;
double delta = 0.001f;
assertEquals(expResult, result.getValue(), delta);
}