spark 机器学习包实现KMeans

KMeans将数据分为K类。第一次随机选择k个点做质心。求每个数据与每个质心（迪卡尔积）的欧式距离，并选择最短距离。如此将数据分为k类。每类数据取平均值将结果作为质心重新计算欧式距离。直到质心基本不变（误差平方和最小）为止。
public class KMean {
    public static void main(String[] args) {
        if(args.length<3){
            System.out.println("error ");
            System.exit(1);
        }
        String inputFile=args[0];
        int k=new Integer(args[1]);
        int iterations=new Integer(args[2]);
        int runs=1;
        if(args.length>4){
            runs=new Integer(args[3]);
        }
        SparkConf conf = new SparkConf();
        JavaSparkContext sc = new JavaSparkContext(conf);
        JavaRDD file = sc.textFile(inputFile);
        Pattern pattern=Pattern.compile(" ");//预编译 split（）每次都会编译，效率低
        JavaRDD map = file.map(x -> {
            String[] split = pattern.split(x);
            double[] point = new double[split.length];
            for (int i = 0; i < split.length; i++) {
                point[i] = new Double(split[i]);
            }
            return Vectors.dense(point);
        });
        KMeansModel train = KMeans.train(map.rdd(),k,iterations,runs,KMeans.K_MEANS_PARALLEL());
        for(Vector v:train.clusterCenters()){
            System.out.println(v);//质心
        }
        double cost = train.computeCost(map.rdd());
        System.out.println(cost);//代价 距离平方和
        sc.stop();