本文代码来之《数据分析与挖掘实战》,在此基础上补充完善了一下~
代码是基于SVM的分类器Python实现,原文章节题目和code关系不大,或者说给出已处理好数据的方法缺失、源是图像数据更是不见踪影,一句话就是练习分类器(▼㉨▼メ)
源代码直接给好了K=30,就试了试怎么选的,挑选规则设定比较单一,有好主意请不吝赐教哟
1 # -*- coding: utf-8 -*- 2 """ 3 Created on Sun Aug 12 12:19:34 2018 4 5 @author: Luove 6 """ 7 from sklearn import svm 8 from sklearn import metrics 9 import pandas as pd 10 import numpy as np 11 from numpy.random import shuffle 12 #from random import seed 13 #import pickle #保存模型和加载模型 14 import os 15 16 17 os.getcwd() 18 os.chdir('D:/Analyze/Python Matlab/Python/BookCodes/Python数据分析与挖掘实战/图书配套数据、代码/chapter9/demo/code') 19 inputfile = '../data/moment.csv' 20 data=pd.read_csv(inputfile) 21 22 data.head() 23 data=data.as_matrix() 24 #seed(10) 25 shuffle(data) #随机重排,按列,同列重排,因是随机的每次运算会导致结果有差异,可在之前设置seed 26 n=0.8 27 train=data[:int(n*len(data)),:] 28 test=data[int(n*len(data)):,:] 29 30 #建模数据 整理 31 #k=30 32 m=100 33 record=pd.DataFrame(columns=['acurrary_train','acurrary_test']) 34 for k in range(1,m+1): 35 # k特征扩大倍数,特征值在0-1之间,彼此区分度太小,扩大以提高区分度和准确率 36 x_train=train[:,2:]*k 37 y_train=train[:,0].astype(int) 38 x_test=test[:,2:]*k 39 y_test=test[:,0].astype(int) 40 41 model=svm.SVC() 42 model.fit(x_train,y_train) 43 #pickle.dump(model,open('../tmp/svm1.model','wb'))#保存模型 44 #model=pickle.load(open('../tmp/svm1.model','rb'))#加载模型 45 #模型评价 混淆矩阵 46 cm_train=metrics.confusion_matrix(y_train,model.predict(x_train)) 47 cm_test=metrics.confusion_matrix(y_test,model.predict(x_test)) 48 49 pd.DataFrame(cm_train,index=range(1,6),columns=range(1,6)) 50 accurary_train=np.trace(cm_train)/cm_train.sum() #准确率计算 51 # accurary_train=model.score(x_train,y_train) #使用model自带的方法求准确率 52 pd.DataFrame(cm_test,index=range(1,6),columns=range(1,6)) 53 accurary_test=np.trace(cm_test)/cm_test.sum() 54 record=record.append(pd.DataFrame([accurary_train,accurary_test],index=['accurary_train','accurary_test']).T) 55 56 record.index=range(1,m+1) 57 find_k=record.sort_values(by=['accurary_train','accurary_test'],ascending=False) # 生成一个copy 不改变原变量 58 find_k[(find_k['accurary_train']>0.95) & (find_k['accurary_test']>0.95) & (find_k['accurary_test']>=find_k['accurary_train'])] 59 #len(find_k[(find_k['accurary_train']>0.95) & (find_k['accurary_test']>0.95)]) 60 ''' k=33 61 accurary_train accurary_test 62 33 0.950617 0.95122 63 ''' 64 ''' 计算一下整体 65 accurary_data 66 0.95073891625615758 67 ''' 68 k=33 69 x_train=train[:,2:]*k 70 y_train=train[:,0].astype(int) 71 model=svm.SVC() 72 model.fit(x_train,y_train) 73 model.score(x_train,y_train) 74 model.score(datax_train,datay_train) 75 datax_train=data[:,2:]*k 76 datay_train=data[:,0].astype(int) 77 cm_data=metrics.confusion_matrix(datay_train,model.predict(datax_train)) 78 pd.DataFrame(cm_data,index=range(1,6),columns=range(1,6)) 79 accurary_data=np.trace(cm_data)/cm_data.sum() 80 accurary_data
REF:
《数据分析与挖掘实战》
源代码及数据需要可自取:https://github.com/Luove/Data