Evaluation criteria for parameter estimation

1. Unbiased

For samples X1, X2...Xn to predict y, the prediction result y1 obtained by each different sample may be different, so we can find the expectation E(y1) of y1.
If E(y1) = the data y to be predicted, then it is an unbiased estimate.

2. Mean square error criterion

The smaller the mes, the better