西瓜书上的LVW（Las Vegas Wrapper）算法python实现

1. 理论部分

         参考12.3 包裹式选择（LVW）周志华《机器学习》

                                                        图1.LVW伪代码算法描述 

 2. 代码部分

        鉴于在网上找不到相关的算法，本人就自己写了一个。注意，本人的代码中，学习算法使用C4.5，使用MSE代替CrossValidation计算误差。

import random
import numpy as np

import matplotlib.pyplot as plt
from math import log
import operator
import pickle
from sklearn.model_selection import train_test_split

"""
函数说明:计算给定数据集的经验熵(香农熵)

Parameters:
	dataSet - 数据集
Returns:
	shannonEnt - 经验熵(香农熵)
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def calcShannonEnt(dataSet):
    numEntires = len(dataSet)  # 返回数据集的行数
    labelCounts = {}  # 保存每个标签(Label)出现次数的字典
    for featVec in dataSet:  # 对每组特征向量进行统计
        currentLabel = featVec[-1]  # 提取标签(Label)信息
        if currentLabel not in labelCounts.keys():  # 如果标签(Label)没有放入统计次数的字典,添加进去
            labelCounts[currentLabel] = 0
        labelCounts[currentLabel] += 1  # Label计数
    shannonEnt = 0.0  # 经验熵(香农熵)
    for key in labelCounts:  # 计算香农熵
        prob = float(labelCounts[key]) / numEntires  # 选择该标签(Label)的概率
        shannonEnt -= prob * log(prob, 2)  # 利用公式计算
    return shannonEnt  # 返回经验熵(香农熵)



"""
函数说明:按照给定特征划分数据集

Parameters:
	dataSet - 待划分的数据集
	axis - 划分数据集的特征
	value - 需要返回的特征的值
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def splitDataSet(dataSet, axis, value):
    retDataSet = []  # 创建返回的数据集列表
    for featVec in dataSet:  # 遍历数据集
        if featVec[axis] == value:
            reducedFeatVec = featVec[:axis]  # 去掉axis特征
            reducedFeatVec.extend(featVec[axis + 1:])  # 将符合条件的添加到返回的数据集
            retDataSet.append(reducedFeatVec)
    return retDataSet  # 返回划分后的数据集


"""
函数说明:选择最优特征

Parameters:
	dataSet - 数据集
Returns:
	bestFeature - 信息增益最大的(最优)特征的索引值
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-20
"""


def chooseBestFeatureToSplit(dataSet):
    numFeatures = len(dataSet[0]) - 1  # 特征数量
    baseEntropy = calcShannonEnt(dataSet)  # 计算数据集的香农熵
    bestInfoGain = 0.0  # 信息增益
    bestFeature = -1  # 最优特征的索引值
    for i in range(numFeatures):  # 遍历所有特征
        # 获取dataSet的第i个所有特征
        featList = [example[i] for example in dataSet]
        uniqueVals = set(featList)  # 创建set集合{},元素不可重复
        newEntropy = 0.0  # 经验条件熵
        for value in uniqueVals:  # 计算信息增益
            subDataSet = splitDataSet(dataSet, i, value)  # subDataSet划分后的子集
            prob = len(subDataSet) / float(len(dataSet))  # 计算子集的概率
            newEntropy += prob * calcShannonEnt(subDataSet)  # 根据公式计算经验条件熵
        infoGain = baseEntropy - newEntropy  # 信息增益
        # print("第%d个特征的增益为%.3f" % (i, infoGain))			#打印每个特征的信息增益
        if (infoGain > bestInfoGain):  # 计算信息增益
            bestInfoGain = infoGain  # 更新信息增益，找到最大的信息增益
            bestFeature = i  # 记录信息增益最大的特征的索引值
    return bestFeature  # 返回信息增益最大的特征的索引值


"""
函数说明:统计classList中出现此处最多的元素(类标签)

Parameters:
	classList - 类标签列表
Returns:
	sortedClassCount[0][0] - 出现此处最多的元素(类标签)
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def majorityCnt(classList):
    classCount = {}
    for vote in classList:  # 统计classList中每个元素出现的次数
        if vote not in classCount.keys(): classCount[vote] = 0
        classCount[vote] += 1
    sortedClassCount = sorted(classCount.items(), key=operator.itemgetter(1), reverse=True)  # 根据字典的值降序排序
    return sortedClassCount[0][0]  # 返回classList中出现次数最多的元素


"""
函数说明:创建决策树

Parameters:
	dataSet - 训练数据集
	labels - 分类属性标签
	featLabels - 存储选择的最优特征标签
Returns:
	myTree - 决策树
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-25
"""


def createTree(dataSet, labels, featLabels):
    classList = [example[-1] for example in dataSet]  # 取分类标签(是否放贷:yes or no)
    if classList.count(classList[0]) == len(classList):  # 如果类别完全相同则停止继续划分
        return classList[0]
    if len(dataSet[0]) == 1 or len(labels) == 0:  # 遍历完所有特征时返回出现次数最多的类标签
        return majorityCnt(classList)
    bestFeat = chooseBestFeatureToSplit(dataSet)  # 选择最优特征
    bestFeatLabel = labels[bestFeat]  # 最优特征的标签
    featLabels.append(bestFeatLabel)
    myTree = {bestFeatLabel: {}}  # 根据最优特征的标签生成树
    del (labels[bestFeat])  # 删除已经使用特征标签
    featValues = [example[bestFeat] for example in dataSet]  # 得到训练集中所有最优特征的属性值
    uniqueVals = set(featValues)  # 去掉重复的属性值
    for value in uniqueVals:  # 遍历特征，创建决策树。
        subLabels = labels[:]
        myTree[bestFeatLabel][value] = createTree(splitDataSet(dataSet, bestFeat, value), subLabels, featLabels)

    return myTree


"""
函数说明:获取决策树叶子结点的数目

Parameters:
	myTree - 决策树
Returns:
	numLeafs - 决策树的叶子结点的数目
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def getNumLeafs(myTree):
    numLeafs = 0  # 初始化叶子
    firstStr = next(iter(
        myTree))  # python3中myTree.keys()返回的是dict_keys,不在是list,所以不能使用myTree.keys()[0]的方法获取结点属性，可以使用list(myTree.keys())[0]
    secondDict = myTree[firstStr]  # 获取下一组字典
    for key in secondDict.keys():
        if type(secondDict[key]).__name__ == 'dict':  # 测试该结点是否为字典，如果不是字典，代表此结点为叶子结点
            numLeafs += getNumLeafs(secondDict[key])
        else:
            numLeafs += 1
    return numLeafs


"""
函数说明:获取决策树的层数

Parameters:
	myTree - 决策树
Returns:
	maxDepth - 决策树的层数
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def getTreeDepth(myTree):
    maxDepth = 0  # 初始化决策树深度
    firstStr = next(iter(
        myTree))  # python3中myTree.keys()返回的是dict_keys,不在是list,所以不能使用myTree.keys()[0]的方法获取结点属性，可以使用list(myTree.keys())[0]
    secondDict = myTree[firstStr]  # 获取下一个字典
    for key in secondDict.keys():
        if type(secondDict[key]).__name__ == 'dict':  # 测试该结点是否为字典，如果不是字典，代表此结点为叶子结点
            thisDepth = 1 + getTreeDepth(secondDict[key])
        else:
            thisDepth = 1
        if thisDepth > maxDepth: maxDepth = thisDepth  # 更新层数
    return maxDepth


"""
函数说明:绘制结点

Parameters:
	nodeTxt - 结点名
	centerPt - 文本位置
	parentPt - 标注的箭头位置
	nodeType - 结点格式
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""

arrow_args = dict(arrowstyle="<-", color='g')


def plot_node(node_txt, center_point, parent_point, node_style):
    ''' 内部函数，外部不要调用
    绘制父子节点，节点间的箭头，并填充箭头中间上的文本
    :param node_txt:文本内容
    :param center_point:文本中心点
    :param parent_point:指向文本中心的点
    '''
    createPlot.ax1.annotate(node_txt,
                            xy=parent_point,
                            xycoords='axes fraction',
                            xytext=center_point,
                            textcoords='axes fraction',
                            va="center",
                            ha="center",
                            bbox=node_style,
                            arrowprops=arrow_args)


"""
函数说明:标注有向边属性值

Parameters:
	cntrPt、parentPt - 用于计算标注位置
	txtString - 标注的内容
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def plotMidText(cntrPt, parentPt, txtString):
    xMid = (parentPt[0] - cntrPt[0]) / 2.0 + cntrPt[0]  # 计算标注位置
    yMid = (parentPt[1] - cntrPt[1]) / 2.0 + cntrPt[1]
    createPlot.ax1.text(xMid, yMid, txtString, va="center", ha="center", rotation=30)


"""
函数说明:绘制决策树

Parameters:
	myTree - 决策树(字典)
	parentPt - 标注的内容
	nodeTxt - 结点名
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def plotTree(myTree, parentPt, nodeTxt):
    decisionNode = dict(boxstyle="sawtooth", fc="0.8")  # 设置结点格式
    leafNode = dict(boxstyle="round4", fc="0.8")  # 设置叶结点格式
    numLeafs = getNumLeafs(myTree)  # 获取决策树叶结点数目，决定了树的宽度
    depth = getTreeDepth(myTree)  # 获取决策树层数
    firstStr = next(iter(myTree))  # 下个字典
    cntrPt = (plotTree.xOff + (1.0 + float(numLeafs)) / 2.0 / plotTree.totalW, plotTree.yOff)  # 中心位置
    plotMidText(cntrPt, parentPt, nodeTxt)  # 标注有向边属性值
    plot_node(firstStr, cntrPt, parentPt, decisionNode)  # 绘制结点
    secondDict = myTree[firstStr]  # 下一个字典，也就是继续绘制子结点
    plotTree.yOff = plotTree.yOff - 1.0 / plotTree.totalD  # y偏移
    for key in secondDict.keys():
        if type(secondDict[key]).__name__ == 'dict':  # 测试该结点是否为字典，如果不是字典，代表此结点为叶子结点
            plotTree(secondDict[key], cntrPt, str(key))  # 不是叶结点，递归调用继续绘制
        else:  # 如果是叶结点，绘制叶结点，并标注有向边属性值
            plotTree.xOff = plotTree.xOff + 1.0 / plotTree.totalW
            plot_node(secondDict[key], (plotTree.xOff, plotTree.yOff), cntrPt, leafNode)
            plotMidText((plotTree.xOff, plotTree.yOff), cntrPt, str(key))
    plotTree.yOff = plotTree.yOff + 1.0 / plotTree.totalD


"""
函数说明:创建绘制面板

Parameters:
	inTree - 决策树(字典)
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-24
"""


def createPlot(inTree):
    fig = plt.figure(1, facecolor='white')  # 创建fig
    fig.clf()  # 清空fig
    axprops = dict(xticks=[], yticks=[])
    createPlot.ax1 = plt.subplot(111, frameon=False, **axprops)  # 去掉x、y轴
    plotTree.totalW = float(getNumLeafs(inTree))  # 获取决策树叶结点数目
    plotTree.totalD = float(getTreeDepth(inTree))  # 获取决策树层数
    plotTree.xOff = -0.5 / plotTree.totalW
    plotTree.yOff = 1.0  # x偏移
    plotTree(inTree, (0.5, 1.0), '')  # 绘制决策树
    plt.savefig('C4.5.jpg', dpi=500)
    plt.show()  # 显示绘制结果


"""
函数说明:使用决策树分类

Parameters:
	inputTree - 已经生成的决策树
	featLabels - 存储选择的最优特征标签
	testVec - 测试数据列表，顺序对应最优特征标签
Returns:
	classLabel - 分类结果
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-25
"""


def classify(inputTree, featLabels, testVec):
    classLabel = []
    firstStr = next(iter(inputTree))  # 获取决策树结点
    secondDict = inputTree[firstStr]  # 下一个字典
    featIndex = featLabels.index(firstStr)

    for key in secondDict.keys():
        if testVec[featIndex] == key:
            if type(secondDict[key]).__name__ == 'dict':
                classLabel.append(classify(secondDict[key], featLabels, testVec))
            else:
                classLabel.append(secondDict[key])
    return classLabel


"""
函数说明:存储决策树

Parameters:
	inputTree - 已经生成的决策树
	filename - 决策树的存储文件名
Returns:
	无
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-25
"""


def storeTree(inputTree, filename):
    with open(filename, 'wb') as fw:
        pickle.dump(inputTree, fw)


"""
函数说明:读取决策树

Parameters:
	filename - 决策树的存储文件名
Returns:
	pickle.load(fr) - 决策树字典
Author:
	Jack Cui
Blog:
	http://blog.csdn.net/c406495762
Modify:
	2017-07-25
"""


def grabTree(filename):
    fr = open(filename, 'rb')
    return pickle.load(fr)


def flatten_list(nested_list):
    # 创建一个新的列表，用于存放所有的值
    flattened_list = []
    none = [[[]], [[[]]], [[[[]]]]]
    # 遍历当前层级的列表
    for item in nested_list:
        if item in none:
            # 当元素为 None 时，跳过此元素
            flattened_list.append(2)
        elif type(item) == list:
            # 当遍历到列表类型时，递归调用该函数将其展开
            flattened_list.extend(flatten_list(item))
        else:
            # 当元素不为列表类型时，将其添加到新的列表中
            flattened_list.append(item)
    return flattened_list


def las_vegas_wrapper(data, features, T):
    Error = np.inf
    x_tolist = features[:-1]
    y = data.values[:, -1]
    d = len(features)
    A1 = features
    t = 0
    while t < T:
        # random.seed(10)
        subset_size = random.randint(2, 9)

        d1 = subset_size
        random_features_subset = random.sample(x_tolist, subset_size)
        labels = random_features_subset.copy()
        subset = []
        for i in range(subset_size):
            if random_features_subset[i] in features:
                a = list(data.columns).index(random_features_subset[i])
                subset.append(data.values[:, a].reshape(-1, 1))
        subset.append(y.reshape(-1, 1))
        subset = np.column_stack((subset[i] for i in range(subset_size + 1))).tolist()
        x_train, x_test = train_test_split(subset, test_size=0.2, random_state=42)
        featLabels = []
        myTree = createTree(x_train, random_features_subset, featLabels)

        for i in range(subset_size):
            if labels[i] not in random_features_subset:
                # print(labels[i])
                random_features_subset.insert(0, labels[i])
        testVec = x_test
        pred_label = []# 测试数据
        for i in range(len(testVec)):
            result = classify(myTree, random_features_subset, testVec[i])
            pred_label.append(result)
        # print(pred_label)
        pred_label = flatten_list(pred_label)
        ture_label = np.array(testVec)[:, -1]
        error = np.mean((ture_label - pred_label)**2)
        if (error < Error) & ((error == Error) | (d1 < d)):
            t = 0
            Error = error
            d = d1
            A1 = random_features_subset
        else:
            t = t + 1
    return sorted(A1)


if __name__ == '__main__':
    import pandas as pd

    data = pd.read_excel("breast_cancer_wisconsin.xlsx")
    features = list(data.columns)

    # 应用函数
    selected_features = las_vegas_wrapper(data, features, T=100)
    print("选中特征的索引:\n", selected_features)