#encoding:utf-8
"""
 Author:  njulpy
 Version:  1.0
 Data:  2018/04/10
 Project： Using Python to Implement LogisticRegression Algorithm
"""
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
#建立sigmoid函數
def sigmoid(x):
 x = x.astype(float)
 return 1./(1+np.exp(-x))
#訓練模型，采用梯度下降算法
def train(x_train,y_train,num,alpha,m,n):
 beta = np.ones(n)
 for i in range(num):
  h=sigmoid(np.dot(x_train,beta)) #計算預測值
  error = h-y_train.T    #計算預測值與訓練集的差值
  delt=alpha*(np.dot(error,x_train))/m #計算參數的梯度變化值
  beta = beta - delt
  #print('error',error)
 return beta
def predict(x_test,beta):
 y_predict=np.zeros(len(y_test))+0.5
 s=sigmoid(np.dot(beta,x_test.T))
 y_predict[s < 0.34] = 0
 y_predict[s > 0.67] = 1
 return y_predict
def accurancy(y_predict,y_test):
 acc=1-np.sum(np.absolute(y_predict-y_test))/len(y_test)
 return acc
if __name__ == "__main__":
 data = pd.read_csv('iris.csv')
 x = data.iloc[:,1:5]
 y = data.iloc[:,5].copy()
 y.loc[y== 'setosa'] = 0
 y.loc[y== 'versicolor'] = 0.5
 y.loc[y== 'virginica'] = 1
 x_train,x_test,y_train,y_test = train_test_split(x,y,test_size=0.3,random_state=15)
 m,n=np.shape(x_train)
 alpha = 0.01
 beta=train(x_train,y_train,1000,alpha,m,n)
 pre=predict(x_test,beta)
 t = np.arange(len(x_test))
 plt.figure()
 p1 = plt.plot(t,pre)
 p2 = plt.plot(t,y_test,label='test')
 label = ['prediction', 'true']
 plt.legend(label, loc=1)
 plt.show()
 acc=accurancy(pre,y_test)
 print('The predicted value is ',pre)
 print('The true value is ',np.array(y_test))
 print('The accuracy rate is ',acc)