ClassifyRepositoryFilter.java

package com.matrobot.gha.insights.filter;

import java.io.IOException;

import com.matrobot.gha.archive.repo.RepositoryRecord;
import com.matrobot.gha.insights.ml.Dataset;
import com.matrobot.gha.insights.ml.Sample;

public class ClassifyRepositoryFilter {

	private static final int MIN_ACTIVITY = 5;
	private RepositoryArchiveList datasets = new RepositoryArchiveList();
	
	
	public ClassifyRepositoryFilter(String firstPath, String secondPath, String thirdPath) throws IOException{
		
		datasets.addFromFile(firstPath);
		datasets.addFromFile(secondPath);
		datasets.addFromFile(thirdPath);
	}
	
	/**
	 * Training data for single feature vector
	 * @return
	 */
	public Dataset getDataset() {
	
		Dataset dataset = new Dataset(2);
		
		for(RepositoryRecord record : datasets.getDataset(1).values()){
			RepositoryRecord nextRecord = datasets.findRepository(2, record.repoName); 
			RepositoryRecord prevRecord = datasets.findRepository(0, record.repoName); 
			if(record.pushEventCount > MIN_ACTIVITY && prevRecord.pushEventCount > MIN_ACTIVITY){
				Sample sample = createSample(record, nextRecord, prevRecord);
				dataset.addSample(sample);
			}
		}
		
		return dataset;
	}

	private Sample createSample(RepositoryRecord record, RepositoryRecord nextRecord, RepositoryRecord prevRecord) {
		
		Sample sample = new Sample(2);
		
		double prevActivity = prevRecord.pushEventCount;
		if(prevActivity > 0){
			sample.features[0] = record.pushEventCount/prevActivity-1;
		}
		else{
			sample.features[0] = 0;
		}

		sample.features[1] =  record.committers.size()-prevRecord.committers.size();
		sample.output = (nextRecord.pushEventCount > record.pushEventCount)? 1 : 0;
		return sample;
	}
	
}