源起:
1.我要做交叉验证,需要每个训练集和测试集都保持相同的样本分布比例,直接用sklearn提供的KFold并不能满足这个需求。
2.将生成的交叉验证数据集保存成CSV文件,而不是直接用sklearn训练分类模型。
3.在编码过程中有一的误区需要注意:
这个sklearn官方给出的文档
>>> import numpy as np
>>> from sklearn.model_selection import KFold
>>> X = ["a", "b", "c", "d"]
>>> kf = KFold(n_splits=2)
>>> for train, test in kf.split(X):
... print("%s %s" % (train, test))
[2 3] [0 1]
[0 1] [2 3]
我之前犯的一个错误是将train,test理解成原数据集分割成子数据集之后的子数据集索引。而实际上,它就是原始数据集本身的样本索引。
源码:
# -*- coding:utf-8 -*-
# 得到交叉验证数据集,保存成CSV文件
# 输入是一个包含正常恶意标签的完整数据集,在读数据的时候分开保存到datasetBenign,datasetMalicious
# 分别对两个数据集进行KFold,最后合并保存
from sklearn.model_selection import KFold
import csv
def writeInFile(benignKFTrain, benignKFTest, maliciousKFTrain, maliciousKFTest, i, datasetBenign, datasetMalicious):
newTrainFilePath = "E:\\hadoopExperimentResult\\5KFold\\AllDataSetIIR10\\dataset\\ImbalancedAllTraffic-train-%s.csv" % i
newTestFilePath = "E:\\hadoopExperimentResult\\5KFold\\AllDataSetIIR10\\dataset\\IImbalancedAllTraffic-test-%s.csv" % i
newTrainFile = open(newTrainFilePath, "wb")# wb 为防止空行
newTestFile = open(newTestFilePath, "wb")
writerTrain = csv.writer(newTrainFile)
writerTest = csv.writer(newTestFile)
for index in benignKFTrain:
writerTrain.writerow(datasetBenign[index])
for index in benignKFTest:
writerTest.writerow(datasetBenign[index])
for index in maliciousKFTrain:
writerTrain.writerow(datasetMalicious[index])
for index in maliciousKFTest:
writerTest.writerow(datasetMalicious[index])
newTrainFile.close()
newTestFile.close()
def getKFoldDataSet(datasetPath):
# CSV读取文件
# 开始从文件中读取全部的数据集
datasetFile = file(datasetPath, 'rb')
datasetBenign = []
datasetMalicious = []
readerDataset = csv.reader(datasetFile)
for line in readerDataset:
if len(line) > 1:
curLine = []
curLine.append(float(line[0]))
curLine.append(float(line[1]))
curLine.append(float(line[2]))
curLine.append(float(line[3]))
curLine.append(float(line[4]))
curLine.append(float(line[5]))
curLine.append(float(line[6]))
curLine.append(line[7])
if line[7] == "benign":
datasetBenign.append(curLine)
else:
datasetMalicious.append(curLine)
# 交叉验证分割数据集
K = 5
kf = KFold(n_splits=K)
benignKFTrain = []; benignKFTest = []
for train,test in kf.split(datasetBenign):
benignKFTrain.append(train)
benignKFTest.append(test)
maliciousKFTrain=[]; maliciousKFTest=[]
for train,test in kf.split(datasetMalicious):
maliciousKFTrain.append(train)
maliciousKFTest.append(test)
for i in range(K):
print "======================== "+ str(i)+ " ========================"
print benignKFTrain[i], benignKFTest[i]
print maliciousKFTrain[i],maliciousKFTest[i]
writeInFile(benignKFTrain[i], benignKFTest[i], maliciousKFTrain[i], maliciousKFTest[i], i, datasetBenign,
datasetMalicious)
datasetFile.close()
if __name__ == "__main__":
getKFoldDataSet(r"E:\hadoopExperimentResult\5KFold\AllDataSetIIR10\dataset\ImbalancedAllTraffic-10.csv")
以上这篇Python sklearn KFold 生成交叉验证数据集的方法就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持软件开发网。
您可能感兴趣的文章:对python sklearn one-hot编码详解Python基于sklearn库的分类算法简单应用示例python利用sklearn包编写决策树源代码Python时间序列处理之ARIMA模型的使用讲解深入浅析Python 中的sklearn模型选择