决策树

下面给出决策树的一些知识

#include <stdio.h>
#include <stdio.h>
#include <math.h>
#include <string.h>
#include <vector>
#include <math.h>
#include <conio.h>
#define maxn 15
#define EPS 0.3
struct st
{
    int flag[5];

}stu[15],next1[15],next2[15];
int si[5];
struct Tree
{
	Tree *t[maxn];
	std::vector<int> ve1,ve2;
	int flag;
	int c;
}tree;
double cal_h(st s[],int len,int t)
{
	
	int vis[10]={0};
	std::vector <int> ve;
	for(int i = 0;i<len;i++)
	{
	
		vis[s[i].flag[t]]++;
		if(vis[s[i].flag[t]]>1) continue;
		ve.push_back(s[i].flag[t]);
	}
	double sum = 0;
	for(int i = 0;i<ve.size();i++)
	{
		if(vis[ve[i]]==0)
		{
			continue;
		}
		double temp  = 1.0*vis[ve[i]]/len;
		sum+=temp*(log(temp)/log(2));
	}
	return -sum;
}

void build_tree(Tree *root,int f_flag)
{
	
     for(int i = 0;i<root->ve1.size();i++)
     {
     	printf("%d ",root->ve1[i]);
     }
     printf("\n");
     for(int i = 0;i<root->ve2.size();i++)
     {
     	printf("%d ",root->ve2[i]);
     }
     printf("\n");
	 for(int i = 0;i<maxn;i++)
	 {
 		root->t[i] = NULL;
 	 }
	 int vis[2]={0};
	 if(root->ve1.size()==0)
	 {
 		 root->flag = f_flag;
 		 return;
 	 }
	 for(int i = 0;i<root->ve1.size();i++)
	 {
	 	  int temp = root->ve1[i];
          vis[stu[temp].flag[4]]++;     
 	 }
 	 if(vis[0]>=vis[1])
 	 {
 	 	root->flag = 0;
 	 }
 	 else
 	 {
 	 	root->flag = 1;
 	 }
     if(root->ve2.size()==0)
     {
     	return;
     }
     double h;
     for(int i= 0;i<root->ve1.size();i++)
     {
     	next1[i] = stu[root->ve1[i]];
     }
     h = cal_h(next1,root->ve1.size(),4);
     double ha[maxn];
     for(int i = 0;i<root->ve2.size();i++)
     {
        
     	int temp = root->ve2[i];
     	ha[temp] = 0;
     	if(temp==4) continue;
     	for(int j = 0;j<si[temp];j++)
     	{
     		int sum = 0;
	     	for(int k = 0;k<root->ve1.size();k++)
	     	{
	     		int ss = root->ve1[k];
	     		if(stu[ss].flag[temp]==j)
	     	    {
    	     		next1[sum] = stu[ss];
    	     		sum++;
    	     	}
	     	}
    	    ha[temp] += 1.0*sum/root->ve1.size()*cal_h(next1,sum,4);
	     	
        }
        ha[temp] = h-ha[temp];
        
     }
     int index;
     double maxnum = -111111111;
     for(int i = 0;i<root->ve2.size();i++)
     {
     	int temp = root->ve2[i];
     	if(ha[temp]>maxnum) maxnum = ha[temp],index = temp;
     }
     if(ha[index]<EPS) return;
     
     root->c = index;
    
     for(int i = 0;i<si[index];i++)
     {
        root->t[i] = new Tree;
     	for(int j = 0;j<root->ve1.size();j++)
        {
        	int a = root->ve1[j];
        	if(stu[a].flag[index]==i)
        	{
	        	root->t[i]->ve1.push_back(a);
	        }
        }
       for(int j = 0;j<root->ve2.size();j++)
        {
        	if(root->ve2[j]!=index)
        	{
	        	root->t[i]->ve2.push_back(root->ve2[j]);
	        }
        }
        
     }
    
     for(int i =0;i<si[index];i++)
     {
     	 build_tree(root->t[i],root->flag);
     }
    
}
void init()
{
    for(int i = 0;i<maxn;i++)
	{
	    for(int j = 0;j<=4;j++)
	    {
   		     scanf("%d",&stu[i].flag[j]);
    	}
	}	
	for(int i = 0;i<=4;i++)
	{
		scanf("%d",&si[i]);
	}
}
int main()
{
	Tree *root = (Tree*)malloc(sizeof(Tree));
	for(int i=0;i<15;i++)
	{
		root->ve1.push_back(i);
	}
	for(int i = 0;i<=4;i++)
	{
		root->ve2.push_back(i);
	}
	init();
	build_tree(root,1);
	return 0;
}

下面给出基于增益的代码，数据为李航的统计学习方法的数据。