com.davidbracewell.ml.indexing.transform.TfIdfTransform.java Source code

Introduction

Here is the source code for com.davidbracewell.ml.indexing.transform.TfIdfTransform.java

Source

/*
 * (c) 2005 David B. Bracewell
 *
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF 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
 *
 * Unless required by applicable law or agreed to in writing,
 * software distributed under the License 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.davidbracewell.ml.indexing.transform;

import com.davidbracewell.collection.CollectionUtils;
import com.davidbracewell.collection.Counter;
import com.davidbracewell.math.CompactCounter;
import com.davidbracewell.math.DoubleEntry;
import com.davidbracewell.ml.Feature;
import com.davidbracewell.ml.FeatureSet;
import com.davidbracewell.ml.Instance;
import com.google.common.base.Function;
import org.apache.commons.math3.util.FastMath;

import javax.annotation.Nullable;
import java.util.ArrayList;
import java.util.List;

/**
 * <p>A <code>Transform</code> that converts the values of a given set of features to TFIDF using the formula:
 * <pre>(0.5 + (0.5*TF)/maxTF) * log((1.0+N)/DF)</pre> where TF is the frequency of a given term, maxTF is the maximum
 * term frequency in the document, N is the total documents in the corpus, and DF is the number of documents the term
 * occurs in. </p>
 *
 * @author David B. Bracewell
 */
public class TfIdfTransform extends RestrictedFeatureTransform {

    private static final long serialVersionUID = 1L;
    Counter<Feature> docFrequency = new CompactCounter<>();
    double maxTF = 0;

    public TfIdfTransform(String featurePrefix) {
        super(featurePrefix);
    }

    @Override
    public void collectImpl(Instance instance) {
        FeatureSet features = instance.getFeatures();
        for (DoubleEntry entry : CollectionUtils.asIterable(instance.nonZeroIterator())) {
            if (shouldTransformFeature(features.get(entry.index).getName())
                    && features.get(entry.index).getType().isReal()) {
                docFrequency.increment(features.get(entry.index));
            }
        }
    }

    @Override
    protected void transformImpl(Instance input) {
        FeatureSet features = input.getFeatures();
        maxTF = Double.NEGATIVE_INFINITY;
        List<DoubleEntry> validEntries = new ArrayList<>();
        for (DoubleEntry entry : CollectionUtils.asIterable(input.nonZeroIterator())) {
            if (shouldTransformFeature(features.get(entry.index).getName())) {
                maxTF = Math.max(maxTF, entry.value);
                validEntries.add(entry);
            }
        }
        for (DoubleEntry entry : validEntries) {
            Feature feature = features.get(entry.index);
            input.set(entry.index, (0.5 + (0.5 * entry.value) / maxTF) * docFrequency.get(feature));
        }
    }

    @Override
    protected void finishImpl() {
        docFrequency = docFrequency.adjustValues(new Function<Double, Double>() {
            @Nullable
            @Override
            public Double apply(@Nullable Double input) {
                return FastMath.log((getCollectionSize() + 1.0) / (input + 1.0));
            }
        });
    }

}//END OF TfIdfTransform