Python分类算法交叉验证

我们使用Sklearn-train_test_split随机划分训练集和测试集

http://blog.csdn.net/cherdw/article/details/54881167

实验代码：

import gensimfrom sklearn.linear_model import LogisticRegressionimport pandas as pdfrom sklearn.model_selection import train_test_splitfile = open(u'/home/ubuntu/file/数据平衡无分类', encoding='utf-8')documents = gensim.models.doc2vec.TaggedLineDocument(file)fileresult=open(u'/home/ubuntu/file/调参结果','w')#怎么根据model计算新的向量#生成文本向量#使用逻辑回归进行预测def LR():    clf = LogisticRegression()    return clfdef getRecognitionRate(testPre, testClass):    testNum = len(testPre)    rightNum = 0    for i in range(0, testNum):        if testClass[i] == testPre[i]:            rightNum += 1    return float(rightNum) / float(testNum)def getData(model):    #生成pandas    tigs = []    data_dict = {}    # 生成pandas数据    fileclass = open(u'/home/ubuntu/file/数据平衡分类结果', encoding='utf-8')    for tig in fileclass:        tigs.append(tig.strip())    for i in range(len(model.docvecs)-1):        data_dict['p'+str(i)] = model.docvecs[i]    data = pd.DataFrame(data_dict)    data = data.T    # data['class0'] = tigs    X_train1, X_test1, y_train1, y_test1 = train_test_split(data, tigs, test_size=0.4, random_state=1)    fileclass.close()    return X_train1, y_train1, X_test1, y_test1#调参# for i in range(20,100):#   for j in range(10,100):# print('参数值：'+str(i)+":"+str(j))# fileresult.write('参数值：'+str(i)+":"+str(j)+'\n')model = gensim.models.Doc2Vec(documents, size=80, window=9, min_count=40, workers=8)T = getData(model)trainMatrix, trainClass, testMatrix, testClass = T[0], T[1], T[2], T[3]clf_LR=LR()clf_LR.fit(trainMatrix, trainClass)print('Logistic Regression recognition rate: '+str(getRecognitionRate(clf_LR.predict(testMatrix), testClass)))# fileresult.write('Logistic Regression recognition rate: '+str(getRecognitionRate(clf_LR.predict(testMatrix), testClass))+'\n')# fileresult.close()#怎么画出来ROC_curve# ROC_curve(clf_LR,testClass)

使用交叉验证方法计算平均正确率：

import gensimfrom sklearn.linear_model import LogisticRegressionimport pandas as pdfrom sklearn.model_selection import train_test_splitimport numpy as npfrom sklearn.model_selection import StratifiedKFoldfrom sklearn.metrics import roc_curve, aucfile = open(u'/home/ubuntu/file/数据平衡无分类', encoding='utf-8')fileclass=open(u'/home/ubuntu/file/数据平衡分类结果',encoding='utf-8')documents = gensim.models.doc2vec.TaggedLineDocument(file)model = gensim.models.Doc2Vec(documents, size=81, window=8, min_count=39, workers=8)#怎么根据model计算新的向量#生成文本向量#使用逻辑回归进行预测def LR():    clf = LogisticRegression()    return clfdef getRecognitionRate(testPre, testClass):    testNum = len(testPre)    rightNum = 0    for i in range(0, testNum):        if testClass[i] == testPre[i]:            rightNum += 1    return float(rightNum) / float(testNum)def getData():    #生成pandas    tigs = []    data_dict = {}    # 生成pandas数据    for tig in fileclass:        tigs.append(tig.strip())    for i in range(len(model.docvecs)):        data_dict['p' + str(i)] = model.docvecs[i]    data = pd.DataFrame(data_dict)    data = data.T    data['class0'] = tigs    X_train1, X_test1, y_train1, y_test1 = train_test_split(data, data['class0'], test_size=0.4, random_state=0)    return X_train1, y_train1, X_test1, y_test1# def ROC_curve(lr,y_test):#     pred_probas = lr.predict_proba(testMatrix)[:,1]#     fpr,tpr,_ = roc_curve(y_test, pred_probas)#     roc_auc = auc(fpr,tpr)#     plt.plot(fpr,tpr,label='area = %.2f' %roc_auc)#     plt.plot([0, 1], [0, 1], 'k--')#     plt.xlim([0.0, 1.0])#     plt.ylim([0.0, 1.05])#     plt.show()def getData_3():    tigs = []    data_dict = {}    # 生成pandas数据    for tig in fileclass:        tigs.append(tig.strip())    for i in range(len(model.docvecs)-1):        data_dict['p' + str(i)] = model.docvecs[i]    data = pd.DataFrame(data_dict)    dataMatrix = data.T    dataMatrix['class0'] = tigs    print(dataMatrix)    sampleData = []    sampleClass = []    for i in dataMatrix.index:        tempList = dataMatrix.loc[i].values        print(tempList)        sampleClass.append(tempList[-1])        sampleData.append(tempList[0:-1])    sampleM = np.array(sampleData)  # 二维矩阵，一行是一个样本，行数=样本总数，列数=样本特征数    classM = np.array(sampleClass)  # 一维列向量，每个元素对应每个样本所属类别    # 调用StratifiedKFold方法生成训练集和测试集    skf = StratifiedKFold(n_splits=2)    setDict = {}  # 创建字典，用于存储生成的训练集和测试集    count = 1    for trainI, testI in skf.split(sampleM, classM):        trainSTemp = []  # 用于存储当前循环抽取出的训练样本数据        trainCTemp = []  # 用于存储当前循环抽取出的训练样本类标        testSTemp = []  # 用于存储当前循环抽取出的测试样本数据        testCTemp = []  # 用于存储当前循环抽取出的测试样本类标        # 生成训练集        trainIndex = list(trainI)        for t1 in range(0, len(trainIndex)):            trainNum = trainIndex[t1]            trainSTemp.append(list(sampleM[trainNum, :]))            trainCTemp.append(list(classM)[trainNum])        setDict[str(count) + 'train'] = np.array(trainSTemp)        setDict[str(count) + 'trainclass'] = np.array(trainCTemp)        # 生成测试集        testIndex = list(testI)        for t2 in range(0, len(testIndex)):            testNum = testIndex[t2]            testSTemp.append(list(sampleM[testNum, :]))            testCTemp.append(list(classM)[testNum])        setDict[str(count) + 'test'] = np.array(testSTemp)        setDict[str(count) + 'testclass'] = np.array(testCTemp)        count += 1    return setDictclf_LR=LR()#怎么画出来ROC_curve# ROC_curve(clf_LR,testClass)setDict = getData_3()setNums = len(setDict.keys())print(setDict.keys())LR_rate = 0.0for i in range(1, 3):    print(i)    trainMatrix = setDict[str(i) + 'train']    trainClass = setDict[str(i) + 'trainclass']    print(len(trainClass))    testMatrix = setDict[str(i) + 'test']    testClass = setDict[str(i) + 'testclass']    print(len(testClass))    clf_LR.fit(trainMatrix, trainClass)    LR_rate += getRecognitionRate(clf_LR.predict(testMatrix), testClass)print('Logistic Regression mean recognition rate: ', LR_rate / 2)

getData_3方法把数据集进行分割，分别对于不同的分割进行计算正确率，最后计算平均正确率。