# -*- coding: utf-8 -*-
"""
Created on Sun Aug 12 12:19:34 2018

@author: Luove
"""
from sklearn import svm
from sklearn import metrics
import pandas as pd 
import numpy as np
from numpy.random import shuffle
#from random import seed
#import pickle #保存模型和加載模型
import os


os.getcwd()
os.chdir('D:/Analyze/Python Matlab/Python/BookCodes/Python數(shù)據(jù)分析與挖掘?qū)崙?zhàn)/圖書配套數(shù)據(jù)、代碼/chapter9/demo/code')
inputfile = '../data/moment.csv'
data=pd.read_csv(inputfile)

data.head()
data=data.as_matrix()
#seed(10)
shuffle(data) #隨機重排，按列，同列重排，因是隨機的每次運算會導(dǎo)致結(jié)果有差異，可在之前設(shè)置seed
n=0.8
train=data[:int(n*len(data)),:]
test=data[int(n*len(data)):,:]

#建模數(shù)據(jù) 整理
#k=30 
m=100
record=pd.DataFrame(columns=['acurrary_train','acurrary_test']) 
for k in range(1,m+1):
  # k特征擴大倍數(shù)，特征值在0-1之間，彼此區(qū)分度太小，擴大以提高區(qū)分度和準確率
  x_train=train[:,2:]*k
  y_train=train[:,0].astype(int)
  x_test=test[:,2:]*k
  y_test=test[:,0].astype(int)
  
  model=svm.SVC()
  model.fit(x_train,y_train)
  #pickle.dump(model,open('../tmp/svm1.model','wb'))#保存模型
  #model=pickle.load(open('../tmp/svm1.model','rb'))#加載模型
  #模型評價 混淆矩陣
  cm_train=metrics.confusion_matrix(y_train,model.predict(x_train))
  cm_test=metrics.confusion_matrix(y_test,model.predict(x_test))
  
  pd.DataFrame(cm_train,index=range(1,6),columns=range(1,6))
  accurary_train=np.trace(cm_train)/cm_train.sum()   #準確率計算
#  accurary_train=model.score(x_train,y_train)             #使用model自帶的方法求準確率
  pd.DataFrame(cm_test,index=range(1,6),columns=range(1,6))
  accurary_test=np.trace(cm_test)/cm_test.sum()
  record=record.append(pd.DataFrame([accurary_train,accurary_test],index=['accurary_train','accurary_test']).T)

record.index=range(1,m+1)
find_k=record.sort_values(by=['accurary_train','accurary_test'],ascending=False) # 生成一個copy 不改變原變量
find_k[(find_k['accurary_train']>0.95) & (find_k['accurary_test']>0.95) & (find_k['accurary_test']>=find_k['accurary_train'])]
#len(find_k[(find_k['accurary_train']>0.95) & (find_k['accurary_test']>0.95)])
''' k=33
  accurary_train accurary_test
33    0.950617    0.95122
'''
''' 計算一下整體 
 accurary_data
 0.95073891625615758
'''
k=33
x_train=train[:,2:]*k
y_train=train[:,0].astype(int)
model=svm.SVC()
model.fit(x_train,y_train)
model.score(x_train,y_train)
model.score(datax_train,datay_train)
datax_train=data[:,2:]*k
datay_train=data[:,0].astype(int)
cm_data=metrics.confusion_matrix(datay_train,model.predict(datax_train))
pd.DataFrame(cm_data,index=range(1,6),columns=range(1,6))
accurary_data=np.trace(cm_data)/cm_data.sum()
accurary_data