回归模型中的误差度量方式

前三种最常用

1 MAE（Mean Absolute Error）平均绝对差值

$$MAE = \frac{1}{N}\sum_{i=0}^n|Y_i - Yhat_i|$$

2 MSE(Mean Square Error)均方误差

$$MSE = \frac{1}{N}\sum_{i=0}^n(Y^2 - {Yhat}^2)$$

3 RMSE(Root Mean Square error)均方根误差

$$RMSE = \sqrt{\frac{1}{N}\sum_{i=0}^n(Y^2-{Yhat}^2)}$$
$$RMSD = \sqrt{\frac{1}{N}\sum_{i=0}^n(x_{1,i}- x_{2,i})^2 }$$

4 Normalized root-mean-square deviation归一化均方差跟偏差

5 coefficient of variation of the RMSD

6 Pearson’s Correlation Coefficient(皮尔逊相关系数)

7 concordance correlation coefficient(一致性相关系数)

参考：
http://blog.csdn.net/computerme/article/details/38871467
http://blog.csdn.net/zryxh1/article/details/53161011
http://blog.csdn.net/reallocing1/article/details/56292877