一樣是從從線性迴歸跟類別型兩大類來看Metrics optimization/ 評估指標最佳化

Regression

RMSE, MSE, R-squared

L2
.Tree-based
XGBoost, LightGBM sklearn.RandomForestRegressor

.Linear models
sklearn.<>Regression sklearn.SGDRegressor Vowpal Wabbit (quantile loss)

.Neural nets
Pytorch, Keras, TF, etc

MAE

L1
.Tree-based
LightGBM sklearn.RandomForestRegressor

.Linear models
Vowpal Wabbit (quantile loss)

.Neural nets
Pytorch, Keras, TF, etc

MSPE as weighted MSE; MAPE as weighted MAE

.使用 sample_weights, XGBoost, LightGBM 等 library 有, 但像 Neral nets 就不 support.
. resample the train set, 通常 resample 多次
df.sample(weights=sample_weights)