kmeans聚类算法

给出训练集，将所给数据集进行聚类算法处理并利用opencv进行图形化界面显示。

#include <opencv2/core/core.hpp>  
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/opencv.hpp>
#include <iostream>
#include <fstream>
#include <iterator>
#include <tuple>
#include <time.h>
#include <math.h>
#include <stdlib.h>
#define k 2

using namespace cv;
using namespace std;

vector<Point2f> readFile(string filePath) {
	vector<Point2f> result;
	ifstream fileIn(filePath, ios_base::in);
	if (!fileIn.is_open()) {
		cout << "读取文件失败" << endl;
	}
	else {
		double** data = new double*[2000];
		for (int i = 0; i < 2000; i++) {
			data[i] = new double[2];
			for (int j = 0; j < 2; j++) {
				fileIn >> data[i][j];
			}
			// 将数据做放大处理，方便后续用图形化界面展示
			Point2f p((data[i][0] + 4) * 100, (data[i][1] + 4) * 100);
			//cout << p.x << "\t " << p.y << endl;
			result.push_back(p);
		}
		fileIn.close();
		return result;
	}
}

void frogy(vector<Point2f> data, Point2f* means) {
	// 随机数种子
	srand((unsigned)time(NULL));
	int random[k];
	bool jump = true;
	for (int i = 0; i < k; i++) {
		do {
			jump = false;
			random[i] = rand() % data.size();
			for (int j = 0; j < i; j++)
				// 避免随机选择到重复的均值点
				if (random[i] == random[j])
					jump = true;
		} while (jump);
		means[i] = data[random[i]];
		//cout << means[i].x << "\t" << means[i].y << endl;
	}
}

double WCSS(Point2f data1, Point2f data2) {
	double sum = 0;
	sum += (data1.x - data2.x) * (data1.x - data2.x);
	sum += (data1.y - data2.y) * (data1.y - data2.y);
	return sqrt(sum);
}

int classifyCluster(Point2f* means, Point2f data) {
	int flag = 0;
	double min = 99999;
	for (int i = 0; i < k; i++) {
		if (WCSS(means[i], data) < min) {
			flag = i;
			min = WCSS(means[i], data);
		}
	}
	return flag;
}

Point2f getNewMeans(vector<Point2f> cluster) {
	int num = cluster.size();
	double sum[2] = { 0 };
	for (int i = 0; i < num; i++) {
		sum[0] += cluster[i].x;
		sum[1] += cluster[i].y;
	}
	Point2f newMean(sum[0] / num, sum[1] / num);
	return newMean;
}

void print(vector<Point2f> *clusters, Point2f *means) {
	Scalar colorTab[] = {
		Scalar(0, 0, 255), //红
		Scalar(0, 255, 0), // 绿
		Scalar(255, 100, 100), // 蓝
		Scalar(0, 255, 255), // 黄
		Scalar(255, 0, 255), // 紫
		Scalar(255, 255, 0), // 浅蓝
		Scalar(255, 255, 255), // 白
	};
	Mat img(800, 800, CV_8UC3);
	img = Scalar::all(0);
	for (int j = 0; j < k; j++) {
		for (int i = 0; i < clusters[j].size(); i++) {
			circle(img, clusters[j][i], 2, colorTab[j + 1], CV_FILLED, CV_AA);
		}
		circle(img, means[j], 4, colorTab[0], CV_FILLED, CV_AA);
	}
	cvNamedWindow("result");
	imshow("result", img);
	waitKey(NULL);
}

int main() {
	vector<Point2f> clusters[k]; // k个聚类
	Point2f means[k]; // k个均值点
	string filePath = "C:\\Users\\Administrator\\Desktop\\kMeansData.txt";
	// 将2k个样本读入data中
	vector<Point2f> data = readFile(filePath);
	//cout << data.size() << endl;

	// 1.Frogy方法——随机选择K个样本作为初始均值点
	frogy(data, means);
	//cout << get<0>(means[0]) << "\t" << get<1>(means[0]) << "\t" << get<1>(means[0]) << endl;
	//cout << get<0>(means[1]) << "\t" << get<1>(means[1]) << "\t" << get<1>(means[1]) << endl;

	// 将剩余样本归类
	int flag;
	for (int i = 0; i < data.size(); i++) {
		flag = classifyCluster(means, data[i]);
		clusters[flag].push_back(data[i]);
	}
	print(clusters, means);
	Point2f oldMeans[k] = { (0, 0) };
	//int counter = 0;
	//cout << WCSS(means[0], oldMeans[0]) << endl << WCSS(means[1], oldMeans[1]) << endl;;
	bool jump = true;;
	while (jump) {
		jump = false;
		for (int i = 0; i < k; i++)
			if (WCSS(means[i], oldMeans[i]) > 1)
				jump = true;
		//cout << counter++ << endl;
		// 重新计算新的均值点
		for (int i = 0; i < k; i++) {
			oldMeans[i] = means[i];
			means[i] = getNewMeans(clusters[i]);
		}
		// 清空每个聚类
		for (int i = 0; i < k; i++) {
			clusters[i].clear();
		}
		// 根据新的均值点获得新的聚类
		for (int i = 0; i < data.size(); i++) {
			flag = classifyCluster(means, data[i]);
			clusters[flag].push_back(data[i]);
		}
		print(clusters, means);
	}
	// 用图形化界面展示
	print(clusters, means);
}