kmeans.java

Kmeans for data mining, clustering

/*
 *Kmeans
 *Chalres
 *
 *
 *
 **/

import java.io.BufferedReader;
import java.io.FileReader;
import java.io.IOException;
import java.util.Random;


public class kmeans {

     private static int M=150;     //No. of instance        
     private static int k=3;      //k cluster        
     private static int N=4;     //No. of dimension           
     private static String str="iris_data.txt";
     int err = 0; 
     
     private static int Z=0;   //No. of loop
     private static float[][] arr;  //data array
     public String[] target_class; 
     float cent[][];    //centroid
     float ocent[][];   //old centroid
     boolean r=true; //set default r to stop the loop first
     double P1,P2=0;
     float class_data[];
     
     
     void Normalize(){
           float max[]=new float[N];
           float min[]=new float[N];
           int i,j;
           
           for(i=0;i<N;i++){
                 max[i]=min[i]=arr[0][i];
           }
           
           for(i=0;i<M;i++)
                 for(j=0;j<N;j++)
                 	{
                 if(max[j]<arr[i][j]) max[j]=arr[i][j];//find the max value
                 if(min[j]>arr[i][j]) min[j]=arr[i][j];//find the min value
                    }
           
           for(i=0;i<M;i++)
                 for(j=0;j<N;j++)
                 	{
                       if(max[j]==min[j]); //do nothing
                       else arr[i][j]=(arr[i][j]-min[j])/(max[j]-min[j]);
                    }//normailze the dataset

            }
     
     public void init(){
           
           arr=new float[M][N+1];
           cent= new float[k][N];
           ocent=new float[k][N];
           target_class = new String[M];
           class_data = new float[M];
           
           //store the original classification
           for (int i = 0; i < 49; i++)
           {
           	class_data[i] = 0;
           }
             for (int i = 49; i < 99; i++)
           {
           	class_data[i] = 1;
           }
           
             for (int i = 99; i < M; i++)
           {
           	class_data[i] = 2;
           }
           
           readFile(str);
           int i,j;
           //print raw dataset
           for(i=0;i<M;i++){
                 
                 System.out.print("No. "+i+"\t");
                 for(j=0;j<N;j++)
                 	{
                       System.out.print(arr[i][j]);
                       
                       System.out.print("\t");
                     //  if(j==3)
                             }
                      
            System.out.print(class_data[i]);
            System.out.print("\n");
                           }
           Normalize();
           calculate();
           System.out.print("Cluster:"+"\n");
           for(i=0;i<M;i++)
           	{
          if (arr[i][N] == 2.0) {target_class[i] = "+";}
           else if (arr[i][N] == 1.0) {target_class[i] = "-";}
           else if (arr[i][N] == 0.0) {target_class[i] = "*";}
           else {target_class [i] = "?";}
           			
           			
          // System.out.print((arr[i][N])+"\n");
           System.out.print((target_class[i]));
          if (arr[i][N] - class_data[i] !=0) err++;
           }
          //System.out.print(M);    
           	//System.out.print(N);  
           System.out.println();
           
           System.out.print("Centroid: "+"\n");
           for(i=0;i<cent.length;i++)//print centroid
                 for(j=0;j<cent[i].length;j++){
                 	   
                       System.out.print(cent[i][j]+"\t");
                       if(j==cent[i].length-1) 
                             System.out.print("\n");
                 }
          // System.out.print("Error of Cluster: "+err);
           System.out.print("No. of Loop:"+Z);
           //System.out.print("\n"+P1+"\n"+P2);
     }
     



           public void readFile(String fileName){        //read from file
             int count = 0,i,j;
             String str[] = new String[M];
             String str1 = null;
             String str2[] = new String[4];
             try{
                 FileReader fr = new FileReader(fileName);
                 BufferedReader br = new BufferedReader(fr);
                 while((str1=br.readLine())!=null){
                     str[count] = str1;
                     count++;
                 }
                 fr.close();
                 br.close();
             }
             catch(IOException e){
                 System.out.println("IOException");
             }
             for(i=0;i<M;i++){
                 str2=str[i].split(",");
                 for(j=0;j<N;j++){
                       Float ff = new Float(str2[j]);//convert string to float
                     arr[i][j]=ff.floatValue();//store into the array
                 }
             }
             //print out array, for testing
             /*for(i=0;i<M;i++){
                 for(j=0;j<N;j++){
                     System.out.print(arr[i][j]+",");
                 }
                 System.out.println();
             }*/
         }

     
     public void calculate(){
           int j;
           Random R= new Random();
           for(int i=0;i<M;i++){
           	//randomly centroid choose
                 j=Math.abs(R.nextInt())%k;
                 arr[i][N]=j;
           }
           
           //work out centroid
           calulatecent();
           while(r)
           	{
           Z++;
           //No. of loop +1
           distance();
           calulatecent();
           if(Math.abs(P1-P2)<0.0000001) break;//act as if the distance is not changed
           for(int a=0;a<k;a++)
                 for(int i=0;i<N;i++)
                 	{
                             if(cent[a][i]!=ocent[a][i])
                             	{ 
                                   r=true;
                                   break;
                                   //the centroid is not stable
                                 }
                             else 
                                   r=false;
                                   //try to break the loop
                    }
           }
     }

     
     public void calulatecent(){//calculate centroid
           float s[]=new float[N];//total sum for the dimension
           for(int i=0;i<k;i++)//save old centroid
                 for(int j=0;j<N;j++){
                       ocent[i][j]=cent[i][j];
                 }
           
           for(int i=0;i<k;i++){

                 for(int l=0;l<N;l++) s[l]=0;
                 int l=0;
                 
                 for(int j=0;j<M;j++)
                       if(arr[j][N]==i){//centroid for every dimension
                             l++;
                             for(int z=0;z<N;z++)
                                   s[z]+=arr[j][z];
                             }
                 for(int x=0;x<N;x++)
                       cent[i][x]=s[x]/l;//mean
                 }
     }
     
     public void distance(){
     	//distance
           int i,j,w;
           double dis,z;
           P1=P2;P2=0;
           for(i=0;i<M;i++)
           	{
                 dis=Double.MAX_VALUE;
                 //apporach to infinite value
                 for(j=0;j<k;j++)
                 	{
                       z=0;
                       for(w=0;w<N;w++){
                       	//Euclidean distance
                            z=z+(cent[j][w]-arr[i][w])*(cent[j][w]-arr[i][w]);
                       }
                       z=Math.sqrt(z);
                       if(dis>z) 
                       	{
                       	//distance is measureble
                             dis=z;
                             arr[i][N]=j;//the cluster is selected
                             P2+=z;
                       }
                 }
           }
     
     }
     
     public static void main(String []args){
           kmeans k = new kmeans();
           k.init();     
           
     }

}