A good model is a model that performs well on the Testing Dataset, i.e. generalizes well.
#Regression
#Classification
- Precision
- Recall
- Accuracy, and cost-sensitive and per-class accuracy.
- AUC-ROC
- Cohen’s Kappa Statistic
#Ranking
- Precision and Recall
- Cumulative Gain (CG)
- Discounted Cumulative Gain (DCG)
- Normalized discounted cumulative gain
- Ideal discounted cumulative gain
Model Calibration