Package quickml.supervised.crossValidation

Examples of quickml.supervised.crossValidation.StationaryCrossValidator

28
29
30
31
32
33
34
        return determineAttributeImportance(new TreeBuilder(), trainingData);
    }


    public TreeSet<AttributeScore> determineAttributeImportance(PredictiveModelBuilder predictiveModelBuilder, final Iterable<? extends Instance<AttributesMap>> trainingData) {
        return determineAttributeImportance(new StationaryCrossValidator(4, new ClassifierLogCVLossFunction()), predictiveModelBuilder, trainingData);
    }
View Full Code Here
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49

    }

    private static void testWithInstances(String dsName, final List<Instance<AttributesMap>> instances) {
        StationaryCrossValidator crossValidator = new StationaryCrossValidator(new ClassifierLogCVLossFunction());

        for (final Scorer scorer : Lists.newArrayList(new SplitDiffScorer(), new MSEScorer(MSEScorer.CrossValidationCorrection.FALSE), new MSEScorer(MSEScorer.CrossValidationCorrection.TRUE))) {
            final TreeBuilder singleTreeBuilder = new TreeBuilder(scorer).binaryClassification(true);
            System.out.println(dsName+", single-tree, "+scorer+", "+crossValidator.getCrossValidatedLoss(singleTreeBuilder, instances));

            TreeBuilder forestTreeBuilder = new TreeBuilder(scorer).ignoreAttributeAtNodeProbability(0.5).binaryClassification(true);
            RandomForestBuilder randomForestBuilder = new RandomForestBuilder(forestTreeBuilder).numTrees(100).executorThreadCount(8);
            System.out.println(dsName+", random-forest, "+scorer+", "+crossValidator.getCrossValidatedLoss(randomForestBuilder, instances));
        }
    }
View Full Code Here
20
21
22
23
24
25
26
27
28
29
30
31
32
public class PredictiveAccuracyTests {
    private static final  Logger logger =  LoggerFactory.getLogger(PredictiveAccuracyTests.class);

    @Test
    public void irisTest() throws Exception {
        StationaryCrossValidator stationaryCrossValidator = new StationaryCrossValidator(new ClassifierRMSECrossValLossFunction());
        final List<Instance<AttributesMap>> irisDataset = Benchmarks.loadIrisDataset();
        final double crossValidatedLoss = stationaryCrossValidator.getCrossValidatedLoss(new RandomForestBuilder(), irisDataset);
        double previousLoss = 0.673;
        logger.info("Cross Validated Lost: {}", crossValidatedLoss);
        Assert.assertTrue(crossValidatedLoss <= previousLoss, String.format("Current loss is %s, but previous loss was %s, this is a regression", crossValidatedLoss, previousLoss));
        Assert.assertTrue(crossValidatedLoss > previousLoss * 0.95, String.format("Current loss is %s, but previous loss was %s, this is a significant improvement, previousLoss should be updated", crossValidatedLoss, previousLoss));

View Full Code Here
TOP

Related Classes of quickml.supervised.crossValidation.StationaryCrossValidator

  • quickml.Benchmarks
  • quickml.PredictiveAccuracyTests
  • quickml.supervised.inspection.AttributeImportanceFinder
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.