使用RFE（反向特征消除）进行数据降维

1 直接使用skl库即可

skl含有的rfe库可以直接以多个变量为目标进行数据降维（主要特征提取），该代码源于2020年华为杯数模比赛b题，项目代码以及数据见：https://github.com/xychen5/mathModel1/blob/master/README.md

# -*- utf:8 -*-
#!/usr/bin/bash
import pandas as pd
import numpy as np
import dataWashing
import matplotlib.pyplot as plt
# 导入随机森林基模型
from sklearn.ensemble import RandomForestRegressor
# 导入RFE方法和线性回归基模型
from sklearn.feature_selection import RFE
from sklearn.linear_model import LinearRegression,Ridge,Lasso# 加载数据
data = pd.read_csv('sampleData/sampleOldData_del3.csv')# 自变量特征
feature = data[["硫含量_原料_μg/g", "辛烷值RON_原料", "饱和烃_v%（烷烃+环烷烃）", "烯烃_v%", "芳烃_v%", "溴值_gBr/100g", "密度(20℃)_kg/m³", "辛烷值RON_产品", "焦炭_wt%", "S__wt%", "焦炭_wt%", "S__wt%", "S-ZORB.CAL_H2.PV", "S-ZORB.PDI_2102.PV", "S-ZORB.PT_2801.PV", "S-ZORB.FC_2801.PV", "S-ZORB.TE_2103.PV", "S-ZORB.TE_2005.PV", "S-ZORB.PT_2101.PV", "S-ZORB.PDT_2104.PV", "S-ZORB.SIS_PDT_2103B.PV", "S-ZORB.TC_2101.PV", "S-ZORB.TE_2301.PV", "S-ZORB.PT_2301.PV", "S-ZORB.FC_2301.PV", "S-ZORB.PC_2105.PV", "S-ZORB.PC_5101.PV", "S-ZORB.TC_5005.PV", "S-ZORB.LC_5001.PV", "S-ZORB.LC_5101.PV", "S-ZORB.TE_5102.PV", "S-ZORB.TE_5202.PV", "S-ZORB.FC_5202.PV", "S-ZORB.AT_5201.PV", "S-ZORB.PT_9301.PV", "S-ZORB.FT_9301.PV", "S-ZORB.FT_1501.PV", "S-ZORB.FT_5104.PV", "S-ZORB.FT_5101.PV", "S-ZORB.FT_9101.PV", "S-ZORB.TE_9001.PV", "S-ZORB.PT_9001.PV", "S-ZORB.FT_9001.PV", "S-ZORB.FT_9403.PV", "S-ZORB.PT_9403.PV", "S-ZORB.TE_9301.PV", "S-ZORB.FT_9201.PV", "S-ZORB.FT_9202.PV", "S-ZORB.FT_9302.PV", "S-ZORB.FT_3301.PV", "S-ZORB.FT_9402.PV", "S-ZORB.PT_9402.PV", "S-ZORB.FT_9401.PV", "S-ZORB.PT_9401.PV", "S-ZORB.PDC_2502.PV", "S-ZORB.FC_2501.PV", "S-ZORB.FT_1001.PV", "S-ZORB.FT_1002.PV", "S-ZORB.FT_1003.PV", "S-ZORB.FT_1004.PV", "S-ZORB.TE_1001.PV", "S-ZORB.FC_1005.PV", "S-ZORB.FC_1101.PV", "S-ZORB.FC_1102.PV", "S-ZORB.AT_1001.PV", "S-ZORB.TE_1105.PV", "S-ZORB.PDI_1102.PV", "S-ZORB.TE_1601.PV", "S-ZORB.SIS_TE_6010.PV", "S-ZORB.PC_6001.PV", "S-ZORB.AC_6001.PV", "S-ZORB.TE_1608.PV", "S-ZORB.TC_1606.PV", "S-ZORB.PT_6002.PV", "S-ZORB.PC_1603.PV", "S-ZORB.PT_1602A.PV", "S-ZORB.PC_1301.PV", "S-ZORB.PT_1201.PV", "S-ZORB.LC_1201.PV", "S-ZORB.FC_1201.PV", "S-ZORB.TE_1201.PV", "S-ZORB.TE_1203.PV", "S-ZORB.LC_1202.PV", "S-ZORB.FC_1203.PV", "S-ZORB.FC_1202.PV", "S-ZORB.PC_1202.PV", "S-ZORB.TC_2801.PV", "S-ZORB.FC_3101.PV", "S-ZORB.FC_3103.PV", "S-ZORB.FC_2601.PV", "S-ZORB.PC_2601.PV", "S-ZORB.PDT_2604.PV", "S-ZORB.TE_2601.PV", "S-ZORB.TC_2607.PV", "S-ZORB.AI_2903.PV", "S-ZORB.PDI_2703A.PV", "S-ZORB.PDC_2607.PV", "S-ZORB.FT_9102.PV", "S-ZORB.PT_1501.PV", "S-ZORB.FT_1002.TOTAL", "S-ZORB.FT_1004.TOTAL", "S-ZORB.FT_9001.TOTAL", "S-ZORB.FT_5104.TOTAL", "S-ZORB.FT_5201.TOTAL", "S-ZORB.FT_5101.TOTAL", "S-ZORB.FT_9101.TOTAL", "S-ZORB.FT_1501.TOTAL", "S-ZORB.FT_1003.TOTAL", "S-ZORB.FT_3301.TOTAL", "S-ZORB.FT_9201.TOTAL", "S-ZORB.FT_9202.TOTAL", "S-ZORB.FT_9301.TOTAL", "S-ZORB.FT_9302.TOTAL", "S-ZORB.FT_9401.TOTAL", "S-ZORB.FT_9402.TOTAL", "S-ZORB.FT_9403.TOTAL", "S-ZORB.FT_1202.TOTAL", "S-ZORB.FT_5201.PV", "S-ZORB.FC_1101.TOTAL", "S-ZORB.FT_1204.PV", "S-ZORB.FT_5102.PV", "S-ZORB.FT_1204.TOTAL", "S-ZORB.FT_5102.TOTAL", "S-ZORB.FC_1202.TOTAL", "S-ZORB.FT_9102.TOTAL", "S-ZORB.FT_1001.TOTAL", "S-ZORB.TE_1101.DACA", "S-ZORB.PT_1102.DACA", "S-ZORB.PT_1103.DACA", "S-ZORB.TE_1104.DACA", "S-ZORB.TE_1107.DACA", "S-ZORB.TE_1103.DACA", "S-ZORB.TE_1106.DACA", "S-ZORB.LI_9102.DACA", "S-ZORB.TE_9003.DACA", "S-ZORB.TE_9002.DACA", "S-ZORB.FT_9002.DACA", "S-ZORB.PC_9002.DACA", "S-ZORB.LT_9001.DACA", "S-ZORB.LC_5002.DACA", "S-ZORB.LC_5102.DACA", "S-ZORB.LT_3801.DACA", "S-ZORB.LT_3101.DACA", "S-ZORB.PC_3101.DACA", "S-ZORB.TE_3101.DACA", "S-ZORB.FT_3303.DACA", "S-ZORB.LC_3301.DACA", "S-ZORB.PC_3301.DACA", "S-ZORB.FT_3304.DACA", "S-ZORB.LT_1501.DACA", "S-ZORB.TE_1501.DACA", "S-ZORB.TE_1502.DACA", "S-ZORB.LC_1203.DACA", "S-ZORB.LT_2101.DACA", "S-ZORB.FT_3001.DACA", "S-ZORB.FT_2701.DACA", "S-ZORB.SIS_PT_2703", "S-ZORB.FC_2702.DACA", "S-ZORB.TC_2702.DACA", "S-ZORB.PT_2905.DACA", "S-ZORB.LT_2901.DACA", "S-ZORB.FT_2901.DACA", "S-ZORB.TE_2901.DACA", "S-ZORB.TE_2902.DACA", "S-ZORB.FT_2502.DACA", "S-ZORB.TE_2501.DACA", "S-ZORB.PT_2501.DACA", "S-ZORB.PT_2502.DACA", "S-ZORB.PDT_2503.DACA", "S-ZORB.ZT_2533.DACA", "S-ZORB.FT_2433.DACA", "S-ZORB.TE_2401.DACA", "S-ZORB.FC_2432.DACA", "S-ZORB.FT_2303.DACA", "S-ZORB.FT_2302.DACA", "S-ZORB.LT_1301.DACA", "S-ZORB.SIS_TE_2802", "S-ZORB.LT_1002.DACA", "S-ZORB.TE_5002.DACA", "S-ZORB.TE_5004.DACA", "S-ZORB.FC_5203.DACA", "S-ZORB.TE_5006.DACA", "S-ZORB.TE_5003.DACA", "S-ZORB.TE_5201.DACA", "S-ZORB.TE_5101.DACA", "S-ZORB.FT_2431.DACA", "S-ZORB.TC_2201.PV", "S-ZORB.TC_2201.OP", "S-ZORB.FT_3201.DACA", "S-ZORB.SIS_PT_2602.PV", "S-ZORB.SIS_TE_2606.PV", "S-ZORB.SIS_TE_2605.PV", "S-ZORB.PDT_2704.DACA", "S-ZORB.PDT_2703B.DACA", "S-ZORB.PDC_2702.DACA", "S-ZORB.PDI_2501.DACA", "S-ZORB.AT_1001.DACA", "S-ZORB.PT_6009.DACA", "S-ZORB.LI_2107.DACA", "S-ZORB.LI_2104.DACA", "S-ZORB.TE_6002.DACA", "S-ZORB.TE_6001.DACA", "S-ZORB.PT_1101.DACA", "S-ZORB.FT_3501.DACA", "S-ZORB.PC_3001.DACA", "S-ZORB.FC_5103.DACA", "S-ZORB.TE_5001.DACA", "S-ZORB.FT_2002.DACA", "S-ZORB.PDT_3601.DACA", "S-ZORB.PDT_3602.DACA", "S-ZORB.PT_6006.DACA", "S-ZORB.SIS_TE_6009.PV", "S-ZORB.SIS_PT_6007.PV", "S-ZORB.TE_6008.DACA", "S-ZORB.PT_5201.DACA", "S-ZORB.FC_1104.DACA", "S-ZORB.PC_3501.DACA", "S-ZORB.FT_2001.DACA", "S-ZORB.FT_2803.DACA", "S-ZORB.LT_9101.DACA", "S-ZORB.PDI_2801.DACA", "S-ZORB.PT_6003.DACA", "S-ZORB.AT_6201.DACA", "S-ZORB.PDI_2301.DACA", "S-ZORB.PDI_2105.DACA", "S-ZORB.PC_2902.DACA", "S-ZORB.FT_1502.DACA", "S-ZORB.BS_LT_2401.PV", "S-ZORB.BS_AT_2402.PV", "S-ZORB.PC_2401.DACA", "S-ZORB.BS_AT_2401.PV", "S-ZORB.PC_2401B.DACA", "S-ZORB.FT_3701.DACA", "S-ZORB.FT_3702.DACA", "S-ZORB.PT_2603.DACA", "S-ZORB.LC_2601.DACA", "S-ZORB.PDT_2605.DACA", "S-ZORB.PT_2607.DACA", "S-ZORB.PDT_2606.DACA", "S-ZORB.ZT_2634.DACA", "S-ZORB.TE_2608.DACA", "S-ZORB.TE_2603.DACA", "S-ZORB.TE_2604.DACA", "S-ZORB.DT_2001.DACA", "S-ZORB.DT_2107.DACA", "S-ZORB.TE_2104.DACA", "S-ZORB.PDT_2001.DACA", "S-ZORB.TE_2002.DACA", "S-ZORB.TE_2001.DACA", "S-ZORB.TE_2004.DACA", "S-ZORB.TE_2003.DACA", "S-ZORB.PC_2401B.PIDA.SP", "S-ZORB.PC_2401B.PIDA.OP", "S-ZORB.PC_2401.PIDA.OP", "S-ZORB.PC_2401.PIDA.SP", "S-ZORB.FT_3302.DACA", "S-ZORB.PDT_1003.DACA", "S-ZORB.PDT_1002.DACA", "S-ZORB.PDT_2409.DACA", "S-ZORB.PDT_3503.DACA", "S-ZORB.PDT_3502.DACA", "S-ZORB.PDT_2906.DACA", "S-ZORB.PDT_3002.DACA", "S-ZORB.PDT_1004.DACA", "S-ZORB.PDI_2903.DACA", "S-ZORB.PT_2901.DACA", "S-ZORB.PT_2106.DACA", "S-ZORB.FT_1301.DACA", "S-ZORB.PT_7510B.DACA", "S-ZORB.TE_7508B.DACA", "S-ZORB.PT_7508B.DACA", "S-ZORB.TE_7506B.DACA", "S-ZORB.PT_7510.DACA", "S-ZORB.TE_7508.DACA", "S-ZORB.PT_7508.DACA", "S-ZORB.TE_7506.DACA", "S-ZORB.PT_7505B.DACA", "S-ZORB.TE_7504B.DACA", "S-ZORB.PT_7503B.DACA", "S-ZORB.TE_7502B.DACA", "S-ZORB.PT_7505.DACA", "S-ZORB.TE_7504.DACA", "S-ZORB.PT_7503.DACA", "S-ZORB.PT_7502.DACA", "S-ZORB.TE_7106B.DACA", "S-ZORB.TE_7108B.DACA", "S-ZORB.PT_7107B.DACA", "S-ZORB.PT_7103B.DACA", "S-ZORB.TE_7102B.DACA", "S-ZORB.TE_7106.DACA", "S-ZORB.PT_7107.DACA", "S-ZORB.PT_7103.DACA", "S-ZORB.TE_7102.DACA", "S-ZORB.HIC_2533.AUTOMANA.OP", "S-ZORB.FC_2432.PIDA.SP", "S-ZORB.PT_1604.DACA", "S-ZORB.TC_1607.DACA", "S-ZORB.PT_6005.DACA", "S-ZORB.PT_6008.DACA", "S-ZORB.PT_1601.DACA", "S-ZORB.TE_1605.DACA", "S-ZORB.TE_1604.DACA", "S-ZORB.TE_1603.DACA", "S-ZORB.TE_1602.DACA", "S-ZORB.SIS_FT_3202.PV", "S-ZORB.TXE_3202A.DACA", "S-ZORB.TXE_3201A.DACA", "S-ZORB.TC_3203.DACA", "S-ZORB.SIS_TEX_3103B.PV", "S-ZORB.TEX_3103A.DACA", "S-ZORB.TE_3111.DACA", "S-ZORB.TE_3112.DACA", "S-ZORB.TXE_2203A.DACA", "S-ZORB.TXE_2202A.DACA", "S-ZORB.TE_5008.DACA", "S-ZORB.TE_5009.DACA", "S-ZORB.FC_5001.DACA", "S-ZORB.TE_5007.DACA", "S-ZORB.TE_1504.DACA", "S-ZORB.TE_1503.DACA", "S-ZORB.TC_3102.DACA", "S-ZORB.TE_1102.DACA", "S-ZORB.AT-0001.DACA.PV", "S-ZORB.AT-0002.DACA.PV", "S-ZORB.AT-0003.DACA.PV", "S-ZORB.AT-0004.DACA.PV", "S-ZORB.AT-0005.DACA.PV", "S-ZORB.AT-0006.DACA.PV", "S-ZORB.AT-0007.DACA.PV", "S-ZORB.AT-0008.DACA.PV", "S-ZORB.AT-0009.DACA.PV", "S-ZORB.AT-0010.DACA.PV", "S-ZORB.AT-0011.DACA.PV", "S-ZORB.AT-0012.DACA.PV", "S-ZORB.AT-0013.DACA.PV", "S-ZORB.TE_2104.DACA.PV", "S-ZORB.SIS_PDT_2103A.PV", "S-ZORB.PT_2106.DACA.PV", "S-ZORB.TE_6008.DACA.PV", "S-ZORB.TE_6001.DACA.PV", "S-ZORB.FT_1204.DACA.PV", "S-ZORB.LC_1203.PIDA.PV", "S-ZORB.FT_5102.DACA.PV", "S-ZORB.LC_5102.PIDA.PV", "S-ZORB.TE_1103.DACA.PV", "S-ZORB.TE_1104.DACA.PV", "S-ZORB.TE_1102.DACA.PV", "S-ZORB.TE_1106.DACA.PV", "S-ZORB.TE_1107.DACA.PV", "S-ZORB.TE_1101.DACA.PV", "S-ZORB.CAL.LINE.PV", "S-ZORB.CAL.CANGLIANG.PV", "S-ZORB.CAL.SPEED.PV", "S-ZORB.CAL.LEVEL.PV", "S-ZORB.RXL_0001.AUXCALCA.PV", "S-ZORB.CAL_1.CANGLIANG.PV", "S-ZORB.FT_1006.DACA.PV", "S-ZORB.FT_5204.DACA.PV", "S-ZORB.FT_1006.TOTALIZERA.PV", "S-ZORB.FT_5204.TOTALIZERA.PV", "S-ZORB.FT_1503.DACA.PV", "S-ZORB.FT_1503.TOTALIZERA.PV", "S-ZORB.FT_1504.DACA.PV", "S-ZORB.FT_1504.TOTALIZERA.PV", "S-ZORB.PC_1001A.PV"]]lr = Ridge(alpha=100000, fit_intercept=True, normalize=True,copy_X=True, max_iter=1500, tol=1e-4, solver='auto')# estimator=LinearRegression(),  # 选择lin线性回归为基模型
rfe = RFE(#estimator=LinearRegression(normalize=True),#estimator=Lasso(normalize=True),estimator=lr,#n_features_to_select=28  # 选区特征数n_features_to_select = 27  # 选区特征数
)# 由于选择了一个目标，导致geature中实际上只有 367 = 366 - 1 列
# fit 方法训练选择特征属性
#sFeature = rfe.fit_transform(feature, data["RON损失（不是变量）"])
sFeature = rfe.fit_transform(feature, data[["RON损失（不是变量）", "硫含量_产品_μg/g"]])# 2d matrix: clonum: opt Value, row:
FRMatrix = sFeature.tolist()
#print(rfe.get_support(), "type is: ", type(rfe.get_support()))
chosenFlags = rfe.get_support().tolist()
print(sFeature.tolist(), "\n\n1->", chosenFlags, "\n\n2", )
print(feature.columns[rfe.get_support()]) #查看满足条件的属性# we delelte column 0, 1, 64, (1, 2, 65)so the real index should be:
## RON损失（不是变量）'s original index is 11(12)，
#chosenFlags = [False, False] + chosenFlags[:11] + [False] +  chosenFlags[11:62] + [False] + chosenFlags[62:]# RON损失（不是变量）'s original index is 11(12)， 硫含量_产品_μg/g‘s original index is 9(10)
chosenFlags = [False, False] + chosenFlags[:9] + [False] + chosenFlags[9:11] + [False] +  chosenFlags[11:62] + [False] + chosenFlags[62:]chosenIndex = []
chosenEnName = []
chosenCnName = []
#rawColumnsInfo = dataWashing.loadColumnsInfo("rawData/columnsInfo.csv")
rawColumnsInfo = dataWashing.loadColumnsInfo("sampleData/allCloumnInfo_en_cn.csv")for i in range(len(rawColumnsInfo)):if chosenFlags[i] == True:# 16 nonOptvalues exists, the index starts from 1chosenIndex.append(i+1)chosenEnName.append(rawColumnsInfo[i]["enName"])chosenCnName.append(rawColumnsInfo[i]["cnName"])# write the chosen flags into file
with open("demensionDecrease/chosenOptValues_rfeRidge2.csv", "w") as f:for i in chosenIndex:f.write(str(i) + " ")f.write("\n")for i in chosenEnName:f.write(str(i) + " ")f.write("\n")for i in chosenCnName:f.write(str(i) + " ")for i in range(len(chosenCnName)):print(chosenEnName[i], " ", chosenCnName[i])

使用RFE（反向特征消除）进行数据降维相关推荐

机器学习-3.数据特征预处理与数据降维
特征预处理定义:通过特定的统计方法(数学方法)将数据转换成算法要求的数据. 处理方法数值型数据:标准缩放(1.归一化,2.标准化):缺失值. 类别型数据:one-hot编码. 时间类型:时间的切分. ...
python降维方法_机器学习数据降维方法总结（附python代码）
介绍在机器学习实战时,如果面对一个数据集具有上千个特征,那么对于模型训练将是一个巨大的挑战.面对如此多的数据变量,如果我们认真的去分析每一个变量将耗费我们几周甚至几个月的时间,那么你估计也要被开除了 ...
数据降维的几种方法（转）
转载地址:http://blog.csdn.net/yimingsilence/article/details/53007828 一个简单的降维例子 public final static int c ...
python降维之时间类型数据的处理_【Python算法】数据降维概述
数据降维概述 1.数据降维概述所谓的数据降维就是指采用某种映射方法,将原高维空间中的数据点映射到低维度的空间中,可以解决大规模特征下的数据显示问题,使得数据集更易使用,降低后续算法的计算,消除噪声影 ...
机器学习基础-数据降维
http://blog.itpub.net/29829936/viewspace-2643140/ 2019-05-01 10:06:00 首先需要了解下:什么叫做数据降维,为什么要对数据进行降维,如 ...
机器学习——数据降维和相关性分析
特征工程中数据降维方法思想有两种: 1. 破坏数据原有的结构从而提取数据的主要特征,例如主成分分析(PCA): 2. 对数据进行相关性分析,按照一定的法则来对数据的属性进行取舍达到降维的目的. 相关性 ...
python卡方检验筛选特征原理_基于Python的遥感特征筛选—递归特征消除(RFE)与极限树(Extra-Trees)...
引言基于前几篇文章关于筛选方法的介绍,本篇同样给大家介绍两种python封装的经典特征降维方法,递归特征消除(RFE)与极限树(Extra-Trees, ET).其中,RFE整合了两种不同的超参数, ...
使用特征_R语言-使用caret包实现特征选择：递归特征消除(RFE)算法
在caret(short for classification and regression training)包中有多个函数可以实现特征选择,总的分为封装法和过滤法.封装法,将特征选择过程与训练过程 ...
R语言基于递归特征消除RFE（Recursive Feature Elimination）进行特征筛选（feature selection）
R语言基于递归特征消除RFE(Recursive Feature Elimination)进行特征筛选(feature selection) 对一个学习任务来说,给定属性集,有些属性很有用,另一些则可 ...

使用RFE（反向特征消除）进行数据降维

1 直接使用skl库即可

使用RFE（反向特征消除）进行数据降维相关推荐

最新文章

热门文章