注意python版本，目前还在坑里,
10月29日，OK 搞定了将Spyder编译器改成3.2.3，Python版本是3.62，问题解决
# -*- coding: utf-8 -*-
"""
Spyder Editor
This is a temporary script file.
"""
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
dataset = pd.read_csv('50_Startups.csv')
X = dataset.iloc[:, :-1].values
Y = dataset.iloc[:,4].values
from sklearn.preprocessing import LabelEncoder, OneHotEncoder
labelencoder_x = LabelEncoder()
X[:, 3] = labelencoder_x.fit_transform(X[:,3])
onehotencoder = OneHotEncoder(categorical_features = [3])
X = onehotencoder.fit_transform(X).toarray()
X = X[:, 1:]
from sklearn.cross_validation import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size = 0.2, random_state = 0)
from sklearn.linear_model import LinearRegression
regressor = LinearRegression()
regressor.fit(X_train,y_train)
#查看结果
y_pre = regressor.predict(X_test)
#用后退梯度
import statsmodels.formula.api as sm
#axis = 1 最右边
X = np.append(arr = np.ones((50,1)).astype(int),values = X ,axis = 1)
X_opt = X[:,[0,1,2,3,4,5]]
#ALL in
regressor_OLS = sm.OLS(endog = Y, exog = X_opt).fit()
regressor_OLS.summary()
X_opt = X[:,[0,1,3,4,5]]
#梯度递减
regressor_OLS = sm.OLS(endog = Y, exog = X_opt).fit()
regressor_OLS.summary()
X_opt = X[:,[0,3,4,5]]
#梯度递减
regressor_OLS = sm.OLS(endog = Y, exog = X_opt).fit()
regressor_OLS.summary()
X_opt = X[:,[0,3,5]]
#梯度递减
regressor_OLS = sm.OLS(endog = Y, exog = X_opt).fit()
regressor_OLS.summary()
X_opt = X[:,[0,3]]
#梯度递减
regressor_OLS = sm.OLS(endog = Y, exog = X_opt).fit()
regressor_OLS.summary()
