com.cloudera.oryx.rdf.common.rule.NumericDecision.java Source code

Introduction

Here is the source code for com.cloudera.oryx.rdf.common.rule.NumericDecision.java

Source

/*
 * Copyright (c) 2013, Cloudera, Inc. All Rights Reserved.
 *
 * Cloudera, Inc. licenses this file to you under the Apache License,
 * Version 2.0 (the "License"). You may not use this file except in
 * compliance with the License. You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * This software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
 * CONDITIONS OF ANY KIND, either express or implied. See the License for
 * the specific language governing permissions and limitations under the
 * License.
 */

package com.cloudera.oryx.rdf.common.rule;

import com.google.common.collect.Lists;
import com.google.common.collect.Multiset;
import com.google.common.collect.TreeMultiset;
import org.apache.commons.math3.stat.descriptive.StorelessUnivariateStatistic;
import org.apache.commons.math3.stat.descriptive.moment.Mean;
import org.apache.commons.math3.util.FastMath;

import java.util.Collections;
import java.util.List;

import com.cloudera.oryx.rdf.common.example.Example;
import com.cloudera.oryx.rdf.common.example.FeatureType;
import com.cloudera.oryx.rdf.common.example.NumericFeature;

/**
 * Represents a decision over a numeric feature. Decisions are defined simply by a threshold value; the
 * decision is positive if the feature's value is greater than or equal to this decision's threshold value.
 *
 * @author Sean Owen
 * @see CategoricalDecision
 */
public final class NumericDecision extends Decision {

    private final float threshold;
    private final boolean defaultDecision;

    public NumericDecision(int featureNumber, float threshold, boolean defaultDecision) {
        super(featureNumber);
        this.threshold = threshold;
        this.defaultDecision = defaultDecision;
    }

    private NumericDecision(int featureNumber, float threshold, float mean) {
        super(featureNumber);
        this.threshold = threshold;
        this.defaultDecision = mean >= threshold;
    }

    /**
     * @return decision threshold; feature values greater than or equal are considered positive
     */
    public float getThreshold() {
        return threshold;
    }

    @Override
    public boolean getDefaultDecision() {
        return defaultDecision;
    }

    @Override
    public boolean isPositive(Example example) {
        NumericFeature feature = (NumericFeature) example.getFeature(getFeatureNumber());
        return feature == null ? defaultDecision : feature.getValue() >= threshold;
    }

    @Override
    public FeatureType getType() {
        return FeatureType.NUMERIC;
    }

    @Override
    public boolean equals(Object o) {
        if (!(o instanceof NumericDecision)) {
            return false;
        }
        NumericDecision other = (NumericDecision) o;
        return getFeatureNumber() == other.getFeatureNumber() && threshold == other.threshold;
    }

    @Override
    public int hashCode() {
        return getFeatureNumber() ^ Float.floatToIntBits(threshold);
    }

    @Override
    public String toString() {
        return "(#" + getFeatureNumber() + " >= " + threshold + ')';
    }

    static List<Decision> numericDecisionsFromExamples(int featureNumber, Iterable<Example> examples,
            int suggestedMaxSplitCandidates) {
        Multiset<Float> sortedFeatureValueCounts = TreeMultiset.create();
        StorelessUnivariateStatistic mean = new Mean();
        int numExamples = 0;
        for (Example example : examples) {
            NumericFeature feature = (NumericFeature) example.getFeature(featureNumber);
            if (feature == null) {
                continue;
            }
            numExamples++;
            float value = feature.getValue();
            sortedFeatureValueCounts.add(value, 1);
            mean.increment(value);
        }

        // Make decisions from split points that divide up input into roughly equal amounts of examples
        List<Decision> decisions = Lists.newArrayListWithExpectedSize(suggestedMaxSplitCandidates);
        int approxExamplesPerSplit = FastMath.max(1, numExamples / suggestedMaxSplitCandidates);
        int examplesInSplit = 0;
        float lastValue = Float.NaN;
        // This will iterate in order of value by nature of TreeMap
        for (Multiset.Entry<Float> entry : sortedFeatureValueCounts.entrySet()) {
            float value = entry.getElement();
            if (examplesInSplit >= approxExamplesPerSplit) {
                decisions.add(
                        new NumericDecision(featureNumber, (value + lastValue) / 2.0f, (float) mean.getResult()));
                examplesInSplit = 0;
            }
            examplesInSplit += entry.getCount();
            lastValue = value;
        }

        // The vital condition here is that if decision n decides an example is positive, then all subsequent
        // decisions in the list will also find it positive. So we need to order from highest threshold to lowest
        Collections.reverse(decisions);
        return decisions;
    }

}