锐客网

  1. 首页 > 睿知 > it技术 > >

python信息熵函数 python 熵值法( 五 )

IT技术Rou


1
,x
2
,x
3
,...x
n
为信息集合X的n个取值，则x i x_ix
i
的概率：
P ( X = i ) = p i , i = 1 , 2 , 3 , . . . , n P(X=i) = p_i, i=1,2,3,...,n
P(X=i)=p
i
,i=1,2,3,...,n
信息集合X的信息熵为：
H ( X ) = ? ∑ i = 1 n p i log ? p i H(X) =- \sum_{i=1}^{n}{p_i}\log{p_i}
H(X)=?
i=1
∑
n
p
i
logp
i
条件熵：指已知某个随机变量的情况下，信息集合的信息熵 。
设信息集合X中有y 1 , y 2 , y 3 , . . . y m {y_1, y_2, y_3, ...y_m}y
1
,y
2
,y
3
,...y
m
组成的随机变量集合Y，则随机变量（X，Y）的联合概率分布为
P ( x = i , y = j ) = p i j P(x=i,y=j) = p_{ij}
P(x=i,y=j)=p
ij
条件熵：
H ( X ∣ Y ) = ∑ j = 1 m p ( y j ) H ( X ∣ y j ) H(X|Y) = \sum_{j=1}^m{p(y_j)H(X|y_j)}
H(X∣Y)=
j=1
∑
m
p(y
j
)H(X∣y
j
)
由
H ( X ∣ y j ) = ? ∑ j = 1 m p ( y j ) ∑ i = 1 n p ( x i ∣ y j ) log ? p ( x i ∣ y j ) H(X|y_j) = - \sum_{j=1}^m{p(y_j)}\sum_{i=1}^n{p(x_i|y_j)}\log{p(x_i|y_j)}
H(X∣y
j
)=?
j=1
∑
m
p(y
j
)
i=1
∑
n
p(x
i
∣y
j
)logp(x
i
∣y
j
)
和贝叶斯公式：
p ( x i y j ) = p ( x i ∣ y j ) p ( y j ) p(x_iy_j) = p(x_i|y_j)p(y_j)
p(x
i
y
j
)=p(x
i
∣y
j
)p(y
j
)
可以化简条件熵的计算公式为:
H ( X ∣ Y ) = ∑ j = 1 m ∑ i = 1 n p ( x i , y j ) log ? p ( x i ) p ( x i , y j ) H(X|Y) = \sum_{j=1}^m \sum_{i=1}^n{p(x_i, y_j)\log\frac{p(x_i)}{p(x_i, y_j)}}
H(X∣Y)=
j=1
∑
m
i=1
∑
n
p(x
i
,y
j
)log
p(x
i
,y
j
)
p(x
i
)
信息增益：信息熵-条件熵 ， 用于衡量在知道已知随机变量后，信息不确定性减小越大 。
d ( X , Y ) = H ( X ) ? H ( X ∣ Y ) d(X,Y) = H(X) - H(X|Y)
d(X,Y)=H(X)?H(X∣Y)
python代码实现
import numpy as np
import math
def calShannonEnt(dataSet):
""" 计算信息熵 """
labelCountDict = {}
for d in dataSet:
label = d[-1]
if label not in labelCountDict.keys():
labelCountDict[label] = 1
else:
labelCountDict[label] += 1
entropy = 0.0
for l, c in labelCountDict.items():
p = 1.0 * c / len(dataSet)
entropy -= p * math.log(p, 2)
return entropy
def filterSubDataSet(dataSet, colIndex, value):
"""返回colIndex特征列label等于value，并且过滤掉改特征列的数据集"""
subDataSetList = []
for r in dataSet:
if r[colIndex] == value:
newR = r[:colIndex]
newR = np.append(newR, (r[colIndex + 1:]))
subDataSetList.append(newR)
return np.array(subDataSetList)
def chooseFeature(dataSet):
""" 通过计算信息增益选择最合适的特征"""
featureNum = dataSet.shape[1] - 1
entropy = calShannonEnt(dataSet)
bestInfoGain = 0.0
bestFeatureIndex = -1
for i in range(featureNum):
uniqueValues = np.unique(dataSet[:, i])
condition_entropy = 0.0
for v in uniqueValues:#计算条件熵
subDataSet = filterSubDataSet(dataSet, i, v)
p = 1.0 * len(subDataSet) / len(dataSet)
condition_entropy += p * calShannonEnt(subDataSet)
infoGain = entropy - condition_entropy#计算信息增益
if infoGain = bestInfoGain:#选择最大信息增益
bestInfoGain = infoGain
bestFeatureIndex = i
return bestFeatureIndex
def creatDecisionTree(dataSet, featNames):

推荐阅读


上一篇：thinkphp2.2.1的简单介绍

下一篇：电脑怎么看显示器多少寸的，电脑怎么看显示器多少寸的