线性回归算法-3.线性回归算法的衡量标准

线性回归算法的衡量标准

  • 均方误差(Mean Squared Error)

\[\frac{1}{m}\sum_{i=1}^{m}(y_{test}^{(i)}- \hat y{_{test}^{(i)}})^2 \]

  • 均方根误差(Root Mean Squared Error)

\[\sqrt{\frac{1}{m}\sum_{i=1}^{m}(y_{test}^{(i)}- \hat y{_{test}^{(i)}})^2} \]

  • 平均绝对误差(Mean Absolute Error)

\[\frac{1}{m}\sum_{i=1}^{m} \left | (y_{test}^{(i)}- \hat y{_{test}^{(i)}}) \right| \]

加载波士顿房产数据

import numpy
import matplotlib.pyplot as plt
from sklearn import datasets

boston = datasets.load_boston()
# 打印此数据集的描述
print(boston.DESCR)

x = boston.data[:,5] #只取房间数量这个特征 

plt.scatter(x,y)
plt.show()

存在边缘极值点,用numpy中的"fancy index"去除数据中的上限值

x = x[y<50]
y = y[y<50]
plt.scatter(x,y)
plt.show()

from mylib.model_selection import train_test_split
from mylib.SimpleLineRegression import SimpleLineRegression

x_train,x_test,y_train,y_test = train_test_split(x,y,seed =666)
reg = SimpleLineRegression()
reg.fit(x_train,y_train)
y_predict = reg.predict(x_test)

通过训练,得到了a,b两个参数,从而确定了线性方程

plt.scatter(x,y)
plt.plot(x,reg.predict(x))
plt.show()

MSE 均方误差

mse_test = numpy.sum((y_predict-y_test)**2) / len(x_test)

RMSE 均方根误差

import math
mse_test = numpy.sum((y_predict-y_test)**2) / len(x_test)
rmse_test = math.sqrt(mse_test)

MAE 平均绝对误差

mae_test = numpy.sum(numpy.absolute(y_predict-y_test)) / len(x_test) #absolute 求绝对值

上述算法指标封装为库:

metrics文件:

调用封装的算法度量库:

from mylib.metrics import mean_squared_error,root_mean_squared_error,mean_absolute_error

mean_squared_error(y_predict,y_test)
root_mean_squared_error(y_predict,y_test)
mean_absolute_error(y_predict,y_test)

sk-learn 中的MSE和MAE调用:###

from sklearn.metrics import mean_squared_error,mean_absolute_error

mean_squared_error(y_predict,y_test)
mean_absolute_error(y_predict,y_test)

R Squared

**表达式为: $ R^2 = 1 - \frac{\Sigma(\hat{y}^{(i)} - y{(i)})2}{\Sigma(\overline{y}^{(i)} - y{(i)})2 } = 1 - \frac{MSE(\hat{y},y)}{Var(y)} $ **

  • $ \Sigma(\hat{y}^{(i)} - y{(i)})2 = SS_{residual} $ 使用得到的模型预测产生的错误
  • $ \Sigma(\overline{y}^{(i)} - y{(i)})2 = SS_{total} $ 使用\(y=\hat{y}\)预测产生的错误(baseline Model)
# var求方差
R2 = 1 - mean_squared_error(y_test,y_predict)/numpy.var(y_test)

封装到mylib中的metrics库中

from mylib.metrics import r2_score
r2_score(y_test,y_predict)

# 在线性回归类中封装score方法
reg.score(x_test,y_test)
posted @ 2019-07-14 18:02  凌晨四点的洛杉矶  阅读(229)  评论(0编辑  收藏  举报