网站编辑,wordpress文章版权声明,做ppt会去什么网站找图,大连关键词快速排名2019独角兽企业重金招聘Python工程师标准 关于K-Means介绍很多#xff0c;还不清楚可以查一些相关资料。 个人对其实现步骤简单总结为4步: 1.选出k值,随机出k个起始质心点。 2.分别计算每个点和k个起始质点之间的距离,就近归类。 3.最终中心点集可以划分为… 2019独角兽企业重金招聘Python工程师标准 关于K-Means介绍很多还不清楚可以查一些相关资料。 个人对其实现步骤简单总结为4步: 1.选出k值,随机出k个起始质心点。 2.分别计算每个点和k个起始质点之间的距离,就近归类。 3.最终中心点集可以划分为k类,分别计算每类中新的中心点。 4.重复2,3步骤对所有点进行归类,如果当所有分类的质心点不再改变,则最终收敛。 下面贴代码。 1.入口类基本读取数据源进行训练然后输出。 数据源文件和源码后面会补上。 package com.hyr.kmeans;import au.com.bytecode.opencsv.CSVReader;import java.io.FileReader;
import java.io.FileWriter;
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;public class KmeansMain {public static void main(String[] args) throws IOException {// 读取数据源文件CSVReader reader new CSVReader(new FileReader(src/main/resources/data.csv)); // 数据源FileWriter writer new FileWriter(src/main/resources/out.csv);ListString[] myEntries reader.readAll(); // 6.8, 12.6// 转换数据点集ListPoint points new ArrayListPoint(); // 数据点集for (String[] entry : myEntries) {points.add(new Point(Float.parseFloat(entry[0]), Float.parseFloat(entry[1])));}int k 6; // K值int type 1;KmeansModel model Kmeans.run(points, k, type);writer.write( K is model.getK() , Object Funcion Value is model.getOfv() , calc_distance_type is model.getCalc_distance_type() \n);int i 0;for (Cluster cluster : model.getClusters()) {i;writer.write( classification i \n);for (Point point : cluster.getPoints()) {writer.write(point.toString() \n);}writer.write(\n);writer.write(centroid is cluster.getCentroid().toString());writer.write(\n\n);}writer.close();}} 2.最终生成的模型类,也就是最终训练好的结果。K值,计算的点距离类型以及object function value值。 package com.hyr.kmeans;import java.util.ArrayList;
import java.util.List;public class KmeansModel {private ListCluster clusters new ArrayListCluster();private Double ofv;private int k; // k值private int calc_distance_type;public KmeansModel(ListCluster clusters, Double ofv, int k, int calc_distance_type) {this.clusters clusters;this.ofv ofv;this.k k;this.calc_distance_type calc_distance_type;}public ListCluster getClusters() {return clusters;}public Double getOfv() {return ofv;}public int getK() {return k;}public int getCalc_distance_type() {return calc_distance_type;}
} 3.数据集点对象,包含点的维度,代码里只给出了x轴,y轴二维。以及点的距离计算。通过类型选择距离公式。给出了几种常用的距离公式。 package com.hyr.kmeans;public class Point {private Float x; // x 轴private Float y; // y 轴public Point(Float x, Float y) {this.x x;this.y y;}public Float getX() {return x;}public void setX(Float x) {this.x x;}public Float getY() {return y;}public void setY(Float y) {this.y y;}Overridepublic String toString() {return Point{ x x , y y };}/*** 计算距离** param centroid 质心点* param type* return*/public Double calculateDistance(Point centroid, int type) {// TODODouble result null;switch (type) {case 1:result calcL1Distance(centroid);break;case 2:result calcCanberraDistance(centroid);break;case 3:result calcEuclidianDistance(centroid);break;}return result;}/*计算距离公式*/private Double calcL1Distance(Point centroid) {double res 0;res Math.abs(getX() - centroid.getX()) Math.abs(getY() - centroid.getY());return res / (double) 2;}private double calcEuclidianDistance(Point centroid) {return Math.sqrt(Math.pow((centroid.getX() - getX()), 2) Math.pow((centroid.getY() - getY()), 2));}private double calcCanberraDistance(Point centroid) {double res 0;res Math.abs(getX() - centroid.getX()) / (Math.abs(getX()) Math.abs(centroid.getX())) Math.abs(getY() - centroid.getY()) / (Math.abs(getY()) Math.abs(centroid.getY()));return res / (double) 2;}Overridepublic boolean equals(Object obj) {Point other (Point) obj;if (getX().equals(other.getX()) getY().equals(other.getY())) {return true;}return false;}
} 4.训练后最终得到的分类。包含该分类的质点,属于该分类的点集合该分类是否收敛。 package com.hyr.kmeans;import java.util.ArrayList;
import java.util.List;public class Cluster {private ListPoint points new ArrayListPoint(); // 属于该分类的点集private Point centroid; // 该分类的中心质点private boolean isConvergence false;public Point getCentroid() {return centroid;}public void setCentroid(Point centroid) {this.centroid centroid;}Overridepublic String toString() {return centroid.toString();}public ListPoint getPoints() {return points;}public void setPoints(ListPoint points) {this.points points;}public void initPoint() {points.clear();}public boolean isConvergence() {return isConvergence;}public void setConvergence(boolean convergence) {isConvergence convergence;}
} 5.K-Meams训练类。按照上面所说四个步骤不断进行训练。 package com.hyr.kmeans;import java.util.ArrayList;
import java.util.List;
import java.util.Random;public class Kmeans {/*** kmeans** param points 数据集* param k K值* param k 计算距离方式*/public static KmeansModel run(ListPoint points, int k, int type) {// 初始化质心点ListCluster clusters initCentroides(points, k);while (!checkConvergence(clusters)) { // 所有分类是否全部收敛// 1.计算距离对每个点进行分类// 2.判断质心点是否改变,未改变则该分类已经收敛// 3.重新生成质心点initClusters(clusters); // 重置分类中的点classifyPoint(points, clusters, type);// 计算距离进行分类recalcularCentroides(clusters); // 重新计算质心点}// 计算目标函数值Double ofv calcularObjetiFuncionValue(clusters);KmeansModel kmeansModel new KmeansModel(clusters, ofv, k, type);return kmeansModel;}/*** 初始化k个质心点** param points 点集* param k K值* return 分类集合对象*/private static ListCluster initCentroides(ListPoint points, Integer k) {ListCluster centroides new ArrayListCluster();// 求出数据集的范围(找出所有点的x最小、最大和y最小、最大坐标。)Float max_X Float.NEGATIVE_INFINITY;Float max_Y Float.NEGATIVE_INFINITY;Float min_X Float.POSITIVE_INFINITY;Float min_Y Float.POSITIVE_INFINITY;for (Point point : points) {max_X max_X point.getX() ? point.getX() : max_X;max_Y max_Y point.getY() ? point.getY() : max_Y;min_X min_X point.getX() ? point.getX() : min_X;min_Y min_Y point.getY() ? point.getY() : min_Y;}System.out.println(min_X min_X ,max_X: max_X ,min_Y min_Y ,max_Y max_Y);// 在范围内随机初始化k个质心点Random random new Random();// 随机初始化k个中心点for (int i 0; i k; i) {float x random.nextFloat() * (max_X - min_X) min_X;float y random.nextFloat() * (max_Y - min_Y) min_X;Cluster c new Cluster();Point centroide new Point(x, y); // 初始化的随机中心点c.setCentroid(centroide);centroides.add(c);}return centroides;}/*** 重新计算质心点** param clusters*/private static void recalcularCentroides(ListCluster clusters) {for (Cluster c : clusters) {if (c.getPoints().isEmpty()) {c.setConvergence(true);continue;}// 求均值,作为新的质心点Float x;Float y;Float sum_x 0f;Float sum_y 0f;for (Point point : c.getPoints()) {sum_x point.getX();sum_y point.getY();}x sum_x / c.getPoints().size();y sum_y / c.getPoints().size();Point nuevoCentroide new Point(x, y); // 新的质心点if (nuevoCentroide.equals(c.getCentroid())) { // 如果质心点不再改变 则该分类已经收敛c.setConvergence(true);} else {c.setCentroid(nuevoCentroide);}}}/*** 计算距离,对点集进行分类** param points 点集* param clusters 分类* param type 计算距离方式*/private static void classifyPoint(ListPoint points, ListCluster clusters, int type) {for (Point point : points) {Cluster masCercano clusters.get(0); // 该点计算距离后所属的分类Double minDistancia Double.MAX_VALUE; // 最小距离for (Cluster cluster : clusters) {Double distancia point.calculateDistance(cluster.getCentroid(), type); // 点和每个分类质心点的距离if (minDistancia distancia) { // 得到该点和k个质心点最小的距离minDistancia distancia;masCercano cluster; // 得到该点的分类}}masCercano.getPoints().add(point); // 将该点添加到距离最近的分类中}}private static void initClusters(ListCluster clusters) {for (Cluster cluster : clusters) {cluster.initPoint();}}/*** 检查收敛** param clusters* return*/private static boolean checkConvergence(ListCluster clusters) {for (Cluster cluster : clusters) {if (!cluster.isConvergence()) {return false;}}return true;}/*** 计算目标函数值** param clusters* return*/private static Double calcularObjetiFuncionValue(ListCluster clusters) {Double ofv 0d;for (Cluster cluster : clusters) {for (Point point : cluster.getPoints()) {int type 1;ofv point.calculateDistance(cluster.getCentroid(), type);}}return ofv;}
} 最终训练结果: K is 6 , Object Funcion Value is 21.82857036590576 , calc_distance_type is 3 classification 1
Point{x3.5, y12.5}centroid is Point{x3.5, y12.5} classification 2
Point{x6.8, y12.6}
Point{x7.8, y12.2}
Point{x8.2, y11.1}
Point{x9.6, y11.1}centroid is Point{x8.1, y11.75} classification 3
Point{x4.4, y6.5}
Point{x4.8, y1.1}
Point{x5.3, y6.4}
Point{x6.6, y7.7}
Point{x8.2, y4.5}
Point{x8.4, y6.9}
Point{x9.0, y3.4}centroid is Point{x6.671428, y5.2142863} classification 4
Point{x6.0, y19.9}
Point{x6.2, y18.5}
Point{x5.3, y19.4}
Point{x7.6, y17.4}centroid is Point{x6.275, y18.800001} classification 5
Point{x0.8, y9.8}
Point{x1.2, y11.6}
Point{x2.8, y9.6}
Point{x3.8, y9.9}centroid is Point{x2.15, y10.225} classification 6
Point{x6.1, y14.3}centroid is Point{x6.1, y14.3} 代码下载地址: http://download.csdn.net/download/huangyueranbbc/10267041 github: https://github.com/huangyueranbbc/KmeansDemo 转载于:https://my.oschina.net/u/4074730/blog/3007470
