ID3算法实验

翻译 2016年05月30日 10:56:50

一、实验目的:

理解ID3算法的基本原理,并且编程实现。

二、实验要求:

使用C/C++/MATLAB实现ID3算法。

输入:

若干行,每行 5 个字符串,表示

Outlook    Temperature    Humidity   Wind    Play ball

如上表。

输出:

决策树。

 

实验结果如下:

输入:

 Sunny       Hot    High          Weak    No

 Sunny       Hot    High          Strong    No

 Overcast     Hot    High          Weak    Yes

 Rain        Mild    High          Weak    Yes

 Rain       Cool     Normal        Weak   Yes

 Rain       Cool     Normal        Strong    No

 Overcast    Cool     Normal        Strong    Yes

 Sunny       Mild    High          Weak    No

 Sunny       Cool    Normal        Weak   Yes

 Rain        Mild     Normal        Weak   Yes

 Sunny        Mild   Normal        Strong    Yes

 Overcast    Mild     High          Strong    Yes

 Overcast    Hot     Normal         Weak   Yes

 Rain       Mild     High          Strong    No

 

输出:

Outlook

                Rain            Wind

                                               Strong          No 

                                               Weak            Yes

                Overcast        Yes

                Sunny           Humidity

                                               Normal          Yes

                                               High            No 

三、具体实现:

实现算法如下:

#include <iostream>

#include <fstream>

#include <math.h>

#include <string>

using namespace std;

 

#define ROW 14

#define COL 5

#define log2 0.69314718055

 

typedef struct TNode

{

  char data[15];

  charweight[15];

  TNode *firstchild,*nextsibling;

}*tree;

 

typedef struct LNode

{

  char  OutLook[15];

  char  Temperature[15];

  char  Humidity[15];

  char  Wind[15];

  char  PlayTennis[5];

 

  LNode *next;

}*link;

typedef struct AttrNode

{

  char   attributes[15];//属性

  int        attr_Num;//属性的个数

 

  AttrNode*next;

}*Attributes;

 

char * Examples[ROW][COL] ={//"OverCast","Cool","High","Strong","No",

          //      "Rain","Hot","Normal","Strong","Yes",

           "Sunny","Hot","High","Weak","No",

             "Sunny","Hot","High","Strong","No",

             "OverCast","Hot","High","Weak","Yes",

             "Rain","Mild","High","Weak","Yes",

           "Rain","Cool","Normal","Weak","Yes",

           "Rain","Cool","Normal","Strong","No",

           "OverCast","Cool","Normal","Strong","Yes",

           "Sunny","Mild","High","Weak","No",

           "Sunny","Cool","Normal","Weak","Yes",

           "Rain","Mild","Normal","Weak","Yes",

           "Sunny","Mild","Normal","Strong","Yes",

           "OverCast","Mild","Normal","Strong","Yes",

           "OverCast","Hot","Normal","Weak","Yes",

           "Rain","Mild","High","Strong","No"

            };

char * Attributes_kind[4] ={"OutLook","Temperature","Humidity","Wind"};

int      Attr_kind[4] = {3,3,2,2};

 

char * OutLook_kind[3] ={"Sunny","OverCast","Rain"};

char * Temperature_kind[3] ={"Hot","Mild","Cool"};

char * Humidity_kind[2] ={"High","Normal"};

char * Wind_kind[2] ={"Weak","Strong"};

 

/*int i_Exampple[14][5] = {0,0,0,0,1,

           0,0,0,1,1,

            1,0,0,1,0,

           2,1,0,0,0,

           2,2,1,0,0,

           2,2,1,1,1,

           1,2,1,1,0,

           0,1,0,0,1,

           0,2,1,0,0,

           2,1,1,0,0,

           0,1,1,1,0,

           1,1,1,1,0,

           1,1,1,0,0,

           2,1,0,0,1

            };*/

void treelists(tree T);

void InitAttr(Attributes &attr_link,char *Attributes_kind[],int Attr_kind[]);

void InitLink(link &L,char * Examples[][COL]);

void ID3(tree &T,link L,linkTarget_Attr,Attributes attr);

void PN_Num(link L,int &positve,int&negative);

double Gain(int positive,int negative,char *atrribute,link L,Attributes attr_L);

 

void main()

{

  link LL,p;

  Attributesattr_L,q;

 

  tree T;

  T = new TNode;

 T->firstchild = T->nextsibling = NULL;

 strcpy(T->weight,"");

  strcpy(T->data,"");

 

  attr_L = newAttrNode;

 attr_L->next = NULL;

 

  LL = newLNode;

  LL->next =NULL;

 

  //成功建立两个链表

 InitLink(LL,Examples);

 InitAttr(attr_L,Attributes_kind,Attr_kind);

 

 ID3(T,LL,NULL,attr_L);

 

 cout<<"决策树以广义表形式输出如下:"<<endl;

 treelists(T);//以广义表的形式输出树

//     cout<<Gain(9,5,"OutLook",LL,attr_L)<<endl;

 

 cout<<endl;

}

 

//以广义表的形式输出树

void treelists(tree T)

{

  tree p;

 

  if(!T)

    return;

 cout<<"{"<<T->weight<<"}";

 cout<<T->data;

 

  p =T->firstchild;

  if (p)

  {

   cout<<"(";

    while (p)

    {

     treelists(p);

      p =p->nextsibling;

      if(p)cout<<',';

    }

   cout<<")";

  }

}

void InitAttr(Attributes &attr_link,char *Attributes_kind[],int Attr_kind[])

{

  Attributes p;

 

  for (int i=0;i < 4;i++)

  {

    p = newAttrNode;

    p->next =NULL;

   

   strcpy(p->attributes,Attributes_kind[i]);

   p->attr_Num = Attr_kind[i];

 

    p->next = attr_link->next;

   attr_link->next = p;

  }

}

void InitLink(link &LL,char * Examples[][COL])

{

  link p;     

 

  for (int i =0;i < ROW;i++)

  {

    p = newLNode;

    p->next =NULL;

 

   strcpy(p->OutLook,Examples[i][0]);

   strcpy(p->Temperature,Examples[i][1]);

   strcpy(p->Humidity,Examples[i][2]);

   strcpy(p->Wind,Examples[i][3]);

   strcpy(p->PlayTennis,Examples[i][4]);

 

    p->next =LL->next;

    LL->next= p;   

  }

}

void PN_Num(link L,int &positve,int &negative)

{

  positve = 0;

  negative = 0;

  link p;

 

  p =L->next;

  while (p)

  {

    if(strcmp(p->PlayTennis,"No") == 0)

     negative++;

    elseif(strcmp(p->PlayTennis,"Yes") == 0)

      positve++;

 

    p =p->next;

  }       

}

 

//计算信息增益

//link L: 样本集合S

//attr_L:属性集合

double Gain(int positive,int negative,char *atrribute,link L,Attributes attr_L)

{

  intatrr_kinds;//每个属性中的值的个数

 

  Attributes p =attr_L->next;

  link q =L->next;

 

  int attr_th =0;//第几个属性

  while (p)

  {

    if(strcmp(p->attributes,atrribute) == 0)

    {

      atrr_kinds= p->attr_Num;       

      break;

    }

    p =p->next;

    attr_th++;

  }

 

  doubleentropy,gain=0;

 

  double p1 =1.0*positive/(positive + negative);

  double p2 =1.0*negative/(positive + negative);

 

  entropy =-p1*log(p1)/log2 - p2*log(p2)/log2;//集合熵

  gain =entropy;

 

  //获取每个属性值在训练样本中出现的个数

  //获取每个属性值所对应的正例和反例的个数

 

  //声明一个3*atrr_kinds的数组

  int ** kinds=new int * [3];

  for (int j=0;j < 3;j++)

  {

    kinds[j] =new int[atrr_kinds];//保存每个属性值在训练样本中出现的个数

  }

 

  //初始化

  for (int j =0;j< 3;j++)

  {

    for (int i=0;i < atrr_kinds;i++)

    {

     kinds[j][i] = 0;

    }

  }

  while (q)

  {

    if(strcmp("OutLook",atrribute) == 0)

    {

      for (int i= 0;i < atrr_kinds;i++)

      {

       if(strcmp(q->OutLook,OutLook_kind[i]) == 0)

        {

          kinds[0][i]++;

 

         if(strcmp(q->PlayTennis,"Yes") == 0)

           kinds[1][i]++;

          else

           kinds[2][i]++;

        }

      }

    }

    else if(strcmp("Temperature",atrribute) == 0)

    {

      for (int i= 0;i < atrr_kinds;i++)

      {

       if(strcmp(q->Temperature,Temperature_kind[i]) == 0)

        {

         kinds[0][i]++;

 

         if(strcmp(q->PlayTennis,"Yes") == 0)

           kinds[1][i]++;

          else

           kinds[2][i]++;

        }

      }

    }

    else if(strcmp("Humidity",atrribute) == 0)

    {

      for (int i= 0;i < atrr_kinds;i++)

      {

       if(strcmp(q->Humidity,Humidity_kind[i]) == 0)

        {

         kinds[0][i]++;

 

         if(strcmp(q->PlayTennis,"Yes") == 0)

           kinds[1][i]++;//

          else

           kinds[2][i]++;

        }

      }

    }

    else if(strcmp("Wind",atrribute) == 0)

    {

      for (int i= 0;i < atrr_kinds;i++)

      {

       if(strcmp(q->Wind,Wind_kind[i]) == 0)

        {

         kinds[0][i]++;

 

         if(strcmp(q->PlayTennis,"Yes") == 0)

           kinds[1][i]++;

          else

           kinds[2][i]++;

        }

      }

    }

    q =q->next;

  }

 

  //计算信息增益

  double *gain_kind = new double[atrr_kinds];

  intpositive_kind = 0,negative_kind = 0;

 

  for (int j =0;j < atrr_kinds;j++)

  {

    if(kinds[0][j] != 0 && kinds[1][j] != 0 && kinds[2][j] != 0)

    {

      p1 =1.0*kinds[1][j]/kinds[0][j];

      p2 = 1.0*kinds[2][j]/kinds[0][j];

     

     gain_kind[j] = -p1*log(p1)/log2-p2*log(p2)/log2;

     

      gain =gain - (1.0*kinds[0][j]/(positive + negative))*gain_kind[j];

    }

    else

     gain_kind[j] = 0;

  }

  return gain;

}

 

//在ID3算法中的训练样本子集合与属性子集合的链表需要进行清空

void FreeLink(link &Link)

{

  link p,q;

 

  p =Link->next;

 

  Link->next= NULL;

  while (p)

  {

    q = p;

    p =p->next;

    free(q);

  }

}

void ID3(tree &T,link L,linkTarget_Attr,Attributes attr)

{

  Attributesp,max,attr_child,p1;

  linkq,link_child,q1;

  tree r,tree_p;

 

  int positive=0,negative =0;

 

 PN_Num(L,positive,negative);

 

  //初始化两个子集合

  attr_child =new AttrNode;

 attr_child->next = NULL;

 

  link_child =new LNode;

 link_child->next = NULL;

 

  if (positive== 0)//全是反例

  {

   strcpy(T->data,"No");

    return;

  }

  else if(negative == 0)//全是正例

  {

   strcpy(T->data,"Yes");

    return;

  }

 

  p =attr->next; //属性链表

 

  double gain,g= 0;

 

 /************************************************************************/

  /* 建立属性子集合与训练样本子集合有两个方案:

    一:在原来链表的基础上进行删除;

    二:另外申请空间进行存储子集合;

    采用第二种方法虽然浪费了空间,但也省了很多事情,避免了变量之间的应用混乱

  */

 /************************************************************************/

 

  if(p)

  {

    while (p)

    {

      gain =Gain(positive,negative,p->attributes,L,attr);

     cout<<p->attributes<<"  "<<gain<<endl;

      if(gain> g)

      {

        g =gain;

        max =p;//寻找信息增益最大的属性

      }

      p =p->next;

    }

   strcpy(T->data,max->attributes);//增加决策树的节点

   

   cout<<"信息增益最大的属性:max->attributes ="<<max->attributes<<endl<<endl;

    //下面开始建立决策树

 

    //创建属性子集合

    p =attr->next;

    while (p)

    {

      if(strcmp(p->attributes,max->attributes) != 0)

      {

        p1 = newAttrNode;

       strcpy(p1->attributes,p->attributes);

       p1->attr_Num = p->attr_Num;

       p1->next = NULL;

 

       p1->next = attr_child->next;

       attr_child->next = p1;          

      }

      p =p->next;

    }

   

    //需要区分出是哪一种属性

    //建立每一层的第一个节点

    if(strcmp("OutLook",max->attributes) == 0)

    {

      r = newTNode;

     r->firstchild = r->nextsibling = NULL;

     strcpy(r->weight,OutLook_kind[0]);

     T->firstchild = r;

      //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

      q =L->next;

      while (q)

      {

        if(strcmp(q->OutLook,OutLook_kind[0]) == 0)

        {

          q1 =new LNode;

         strcpy(q1->OutLook,q->OutLook);

         strcpy(q1->Humidity,q->Humidity);

         strcpy(q1->Temperature,q->Temperature);

          strcpy(q1->Wind,q->Wind);

          strcpy(q1->PlayTennis,q->PlayTennis);

         q1->next = NULL;

         

         q1->next = link_child->next;

         link_child->next = q1;

        }

        q =q->next;

      }

    }

    else if(strcmp("Temperature",max->attributes) == 0)

    {

      r = newTNode;

     r->firstchild = r->nextsibling = NULL;

     strcpy(r->weight,Temperature_kind[0]);

     T->firstchild = r;

      //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

      q =L->next;

      while (q)

      {

        if(strcmp(q->Temperature,Temperature_kind[0]) == 0)

        {

          q1 =new LNode;

         strcpy(q1->OutLook,q->OutLook);

         strcpy(q1->Humidity,q->Humidity);

         strcpy(q1->Temperature,q->Temperature);

          strcpy(q1->Wind,q->Wind);

          strcpy(q1->PlayTennis,q->PlayTennis);

         q1->next = NULL;

         

         q1->next = link_child->next;

         link_child->next = q1;

        }

        q =q->next;

      }

    }

    else if(strcmp("Humidity",max->attributes) == 0)

    {

      r = newTNode;

     r->firstchild = r->nextsibling = NULL;

     strcpy(r->weight,Humidity_kind[0]);

     T->firstchild = r;

      //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

      q =L->next;

      while (q)

      {

        if(strcmp(q->Humidity,Humidity_kind[0]) == 0)

        {

          q1 =new LNode;

         strcpy(q1->OutLook,q->OutLook);

         strcpy(q1->Humidity,q->Humidity);

         strcpy(q1->Temperature,q->Temperature);

          strcpy(q1->Wind,q->Wind);

          strcpy(q1->PlayTennis,q->PlayTennis);

         q1->next = NULL;

         

         q1->next = link_child->next;

         link_child->next = q1;

        }

        q =q->next;

      }

    }

    else if(strcmp("Wind",max->attributes) == 0)

    {

      r = newTNode;

     r->firstchild = r->nextsibling = NULL;

     strcpy(r->weight,Wind_kind[0]);

 

     T->firstchild = r;

      //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

      q =L->next;

      while (q)

      {

        if(strcmp(q->Wind,Wind_kind[0]) == 0)

        {

          q1 =new LNode;

         strcpy(q1->OutLook,q->OutLook);

         strcpy(q1->Humidity,q->Humidity);

         strcpy(q1->Temperature,q->Temperature);

          strcpy(q1->Wind,q->Wind);

          strcpy(q1->PlayTennis,q->PlayTennis);

          q1->next = NULL;

         

         q1->next = link_child->next;

         link_child->next = q1;

        }

        q =q->next;

      }

    }

 

    int p = 0,n= 0;

   PN_Num(link_child,p,n);

    if (p != 0&& n != 0)

    {

     ID3(T->firstchild,link_child,Target_Attr,attr_child);

     FreeLink(link_child);

    }

    else if(p ==0)

    {

     strcpy(T->firstchild->data,"No");

     FreeLink(link_child);

      //       strcpy(T->firstchild->data,q1->PlayTennis);//----此处应该需要修改----:)

    }

    else if(n ==0)

    {

     strcpy(T->firstchild->data,"Yes");

     FreeLink(link_child);

    }

 

    //建立每一层上的其他节点

    tree_p =T->firstchild;

    for (int i =1;i < max->attr_Num;i++)

    {

      //需要区分出是哪一种属性

      if(strcmp("OutLook",max->attributes) == 0)

      {

        r = newTNode;

       r->firstchild = r->nextsibling = NULL;

       strcpy(r->weight,OutLook_kind[i]);

       tree_p->nextsibling = r;

        //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

        q =L->next;

        while(q)

        {

          if(strcmp(q->OutLook,OutLook_kind[i]) == 0)

          {

            q1 =new LNode;

           strcpy(q1->OutLook,q->OutLook);

           strcpy(q1->Humidity,q->Humidity);

           strcpy(q1->Temperature,q->Temperature);

            strcpy(q1->Wind,q->Wind);

            strcpy(q1->PlayTennis,q->PlayTennis);

           q1->next = NULL;

 

           q1->next = link_child->next;

           link_child->next = q1;

          }

          q =q->next;

        }

      }

      else if(strcmp("Temperature",max->attributes) == 0)

      {

        r = newTNode;

       r->firstchild = r->nextsibling = NULL;

       strcpy(r->weight,Temperature_kind[i]);

       tree_p->nextsibling = r;

        //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

        q = L->next;

        while(q)

        {

          if(strcmp(q->Temperature,Temperature_kind[i]) == 0)

          {

            q1 =new LNode;

           strcpy(q1->OutLook,q->OutLook);

           strcpy(q1->Humidity,q->Humidity);

           strcpy(q1->Temperature,q->Temperature);

            strcpy(q1->Wind,q->Wind);

            strcpy(q1->PlayTennis,q->PlayTennis);

           q1->next = NULL;

           

           q1->next = link_child->next;

           link_child->next = q1;

          }

          q = q->next;

        }

       

      }

      else if(strcmp("Humidity",max->attributes) == 0)

      {

        r = newTNode;

       r->firstchild = r->nextsibling = NULL;

       strcpy(r->weight,Humidity_kind[i]);

       tree_p->nextsibling = r;

        //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

        q =L->next;

        while(q)

        {

          if(strcmp(q->Humidity,Humidity_kind[i]) == 0)

          {

            q1 =new LNode;

           strcpy(q1->OutLook,q->OutLook);

            strcpy(q1->Humidity,q->Humidity);

           strcpy(q1->Temperature,q->Temperature);

            strcpy(q1->Wind,q->Wind);

            strcpy(q1->PlayTennis,q->PlayTennis);

           q1->next = NULL;

           

           q1->next = link_child->next;

           link_child->next = q1;

          }

          q =q->next;

        }

      }

      else if(strcmp("Wind",max->attributes) == 0)

      {

        r = newTNode;

       r->firstchild = r->nextsibling = NULL;

       strcpy(r->weight,Wind_kind[i]);

       tree_p->nextsibling = r;

        //获取与属性值相关的训练样例Example(vi),建立一个新的训练样本链表link_child

        q =L->next;

        while(q)

        {

          if(strcmp(q->Wind,Wind_kind[i]) == 0)

          {

            q1 =new LNode;

           strcpy(q1->OutLook,q->OutLook);

           strcpy(q1->Humidity,q->Humidity);

           strcpy(q1->Temperature,q->Temperature);

            strcpy(q1->Wind,q->Wind);

            strcpy(q1->PlayTennis,q->PlayTennis);

           q1->next = NULL;

           

            q1->next = link_child->next;

           link_child->next = q1;

          }

          q =q->next;

        }

      }

      int p =0,n = 0;

     PN_Num(link_child,p,n);

      if (p != 0&& n != 0)

      {

       ID3(tree_p->nextsibling,link_child,Target_Attr,attr_child);

       FreeLink(link_child);

      }

      else if(p== 0)

      {

       strcpy(tree_p->nextsibling->data,"No");

       FreeLink(link_child);

      }

      else if(n== 0)

      {

       strcpy(tree_p->nextsibling->data,"Yes");

       FreeLink(link_child);

      }

 

      tree_p =tree_p->nextsibling;//建立所有的孩子结点

    }//建立决策树结束

  }

  else

  {

    q =L->next;

   strcpy(T->data,q->PlayTennis);

    return;//这个地方要赋以训练样本Example中最普遍的Target_attributes的值

  }

}

运行结果如下:


相关文章推荐

java实现ID3算法

ID3算法 机器学习

决策树,ID3算法

构建决策树的ID3算法 点击打开链接

id3经典算法

  • 2012-09-16 21:40
  • 670KB
  • 下载

决策树(decisions tree)和ID3算法

决策树和决策规则是解决实际应用中分类问题的强大的数据挖掘方法。 首先,介绍一下分类的基本原理: 分类是一个学习函数的过程,该函数把数据项映射到其中一个预定义的类中。若一个样本包含属性值向量...

ID3算法数据挖掘

  • 2012-10-29 15:39
  • 11KB
  • 下载

Python实现决策树ID3算法

机器学习 - 决策树ID3算法

最近研究树模型,从最简单的决策树开始研究,扼要的说下最简单的ID3算法。 对于机器学习的数据: train_x, train_y, test_x, test_y。当然test_y通常是被预测的值。对于...
内容举报
返回顶部
收藏助手
不良信息举报
您举报文章:深度学习:神经网络中的前向传播和反向传播算法推导
举报原因:
原因补充:

(最多只允许输入30个字)