/**
 * Copyright (C) 2013-2017 Vasilis Vryniotis 
 *
 * Licensed 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.datumbox.framework.core.machinelearning.regression;

import com.datumbox.framework.common.Configuration;
import com.datumbox.framework.core.common.dataobjects.Dataframe;
import com.datumbox.framework.common.storage.interfaces.StorageEngine;
import com.datumbox.framework.core.common.utilities.MapMethods;
import com.datumbox.framework.core.machinelearning.MLBuilder;
import com.datumbox.framework.core.machinelearning.common.abstracts.AbstractTrainer;
import com.datumbox.framework.core.machinelearning.common.abstracts.modelers.AbstractRegressor;
import com.datumbox.framework.core.machinelearning.common.dataobjects.TrainableBundle;
import com.datumbox.framework.core.machinelearning.common.interfaces.StepwiseCompatible;

import java.util.HashSet;
import java.util.Map;
import java.util.Set;


/**
 * Implementation of Stepwise Regression algorithm. The method takes as argument
 * an other regression model which uses to estimate the best model. This implementation
 * uses the backwards elimination algorithm.
 *
 * @author Vasilis Vryniotis 
 */
public class StepwiseRegression extends AbstractRegressor  {

    private static final String REG_KEY = "reg";

    private final TrainableBundle bundle;

    /** {@inheritDoc} */
    public static class ModelParameters extends AbstractRegressor.AbstractModelParameters {
        private static final long serialVersionUID = 1L;

        /** 
         * @param storageEngine
         * @see AbstractTrainer.AbstractModelParameters#AbstractModelParameters(StorageEngine)
         */
        protected ModelParameters(StorageEngine storageEngine) {
            super(storageEngine);
        }

    } 

    /** {@inheritDoc} */
    public static class TrainingParameters extends AbstractRegressor.AbstractTrainingParameters {
        private static final long serialVersionUID = 1L;
        
        //primitives/wrappers
        private int maxIterations = Integer.MAX_VALUE;
        
        private double aout = 0.05;

        //Parameter Objects
        private AbstractRegressor.AbstractTrainingParameters regressionTrainingParameters;

        //Field Getters/Setters

        /**
         * Getter for the maximum permitted iterations during training.
         * 
         * @return 
         */
        public int getMaxIterations() {
            return maxIterations;
        }

        /**
         * Setter for the maximum permitted iterations during training.
         * 
         * @param maxIterations 
         */
        public void setMaxIterations(int maxIterations) {
            this.maxIterations = maxIterations;
        }
        
        /**
         * Getter for the threshold of the maximum p-value; a variable must
         * have a p-value less or equal than the threshold to be retained in the 
         * model. Variables with higher p-value than this threshold are removed.
         * 
         * 
         * @return 
         */
        public double getAout() {
            return aout;
        }
        
        /**
         * Setter for the threshold of the maximum p-value; a variable must
         * have a p-value less or equal than the threshold to be retained in the 
         * model. Variables with higher p-value than this threshold are removed.
         * 
         * @param aout 
         */
        public void setAout(double aout) {
            this.aout = aout;
        }

        /**
         * Getter for the training parameters of the regression algorithm that is
         * used internally by the Stepwise Regression.
         * 
         * @return 
         */
        public AbstractRegressor.AbstractTrainingParameters getRegressionTrainingParameters() {
            return regressionTrainingParameters;
        }
        
        /**
         * Setter for the training parameters of the regression algorithm that is
         * used internally by the Stepwise Regression.
         * 
         * @param regressionTrainingParameters 
         */
        public void setRegressionTrainingParameters(AbstractRegressor.AbstractTrainingParameters regressionTrainingParameters) {
            this.regressionTrainingParameters = regressionTrainingParameters;
        }
        
    }

    /**
     * @param trainingParameters
     * @param configuration
     * @see AbstractTrainer#AbstractTrainer(AbstractTrainingParameters, Configuration)
     */
    protected StepwiseRegression(TrainingParameters trainingParameters, Configuration configuration) {
        super(trainingParameters, configuration);
        bundle  = new TrainableBundle(configuration.getStorageConfiguration().getStorageNameSeparator());
    }

    /**
     * @param storageName
     * @param configuration
     * @see AbstractTrainer#AbstractTrainer(String, Configuration)
     */
    protected StepwiseRegression(String storageName, Configuration configuration) {
        super(storageName, configuration);
        bundle  = new TrainableBundle(configuration.getStorageConfiguration().getStorageNameSeparator());
    }

    /** {@inheritDoc} */
    @Override
    protected void _predict(Dataframe newData) {
        //load all trainables on the bundles
        initBundle();

        //run the pipeline
        AbstractRegressor mlregressor = (AbstractRegressor) bundle.get(REG_KEY);
        mlregressor.predict(newData);
    }
    
    /** {@inheritDoc} */
    @Override
    protected void _fit(Dataframe trainingData) {
        TrainingParameters trainingParameters = knowledgeBase.getTrainingParameters();
        Configuration configuration = knowledgeBase.getConfiguration();

        //reset previous entries on the bundle
        resetBundle();

        //perform stepwise
        int maxIterations = trainingParameters.getMaxIterations();
        double aOut = trainingParameters.getAout();
        
        //copy data before starting
        Dataframe copiedTrainingData = trainingData.copy();
        
        //backword elimination algorithm
        for(int iteration = 0; iteration pvalues = runRegression(copiedTrainingData);
            
            if(pvalues.isEmpty()) {
                break; //no more features
            }
            
            //fetch the feature with highest pvalue, excluding constant
            pvalues.remove(Dataframe.COLUMN_NAME_CONSTANT);
            Map.Entry maxPvalueEntry = MapMethods.selectMaxKeyValue(pvalues);
            //pvalues=null;
            
            if(maxPvalueEntry.getValue()<=aOut) {
                break; //nothing to remove, the highest pvalue is less than the aOut
            }
            
            
            Set removedFeatures = new HashSet<>();
            removedFeatures.add(maxPvalueEntry.getKey());
            copiedTrainingData.dropXColumns(removedFeatures);
            //removedFeatures = null;
            
            if(copiedTrainingData.xColumnSize()==0) {
                break; //if no more features exit
            }
        }
        
        //once we have the dataset has been cleared from the unnecessary columns train the model once again
        AbstractRegressor mlregressor = MLBuilder.create(
                knowledgeBase.getTrainingParameters().getRegressionTrainingParameters(),
                configuration
        );
        mlregressor.fit(copiedTrainingData);
        bundle.put(REG_KEY, mlregressor);

        copiedTrainingData.close();
    }

    /** {@inheritDoc} */
    @Override
    public void save(String storageName) {
        initBundle();
        super.save(storageName);

        String knowledgeBaseName = createKnowledgeBaseName(storageName, knowledgeBase.getConfiguration().getStorageConfiguration().getStorageNameSeparator());
        bundle.save(knowledgeBaseName);
    }

    /** {@inheritDoc} */
    @Override
    public void delete() {
        initBundle();
        bundle.delete();
        super.delete();
    }

    /** {@inheritDoc} */
    @Override
    public void close() {
        initBundle();
        bundle.close();
        super.close();
    }

    private void resetBundle() {
        bundle.delete();
    }

    private void initBundle() {
        TrainingParameters trainingParameters = knowledgeBase.getTrainingParameters();
        Configuration configuration = knowledgeBase.getConfiguration();
        String storageName = knowledgeBase.getStorageEngine().getStorageName();
        String separator = configuration.getStorageConfiguration().getStorageNameSeparator();

        if(!bundle.containsKey(REG_KEY)) {
            AbstractTrainingParameters mlParams = trainingParameters.getRegressionTrainingParameters();

            bundle.put(REG_KEY, MLBuilder.load(mlParams.getTClass(), storageName + separator + REG_KEY, configuration));
        }
    }

    private Map runRegression(Dataframe trainingData) {
        AbstractRegressor mlregressor = MLBuilder.create(
                knowledgeBase.getTrainingParameters().getRegressionTrainingParameters(),
                knowledgeBase.getConfiguration()
        );
        mlregressor.fit(trainingData);

        Map pvalues = ((StepwiseCompatible)mlregressor).getFeaturePvalues();

        mlregressor.close();
        
        return pvalues;
    }
}

 
 


        
        

                    




        
      

      
Ads help maintain this website.