Performance ​
How to assess this?
- Split data in train and test set
- Build model on train set and evaluate the performance on test set
Accuracy = Correctly Classified Examples/Total Examples
Precision = TP/TP+FP
Recall = TP/TP+FN
| -- | Cancer (Actual) | Not Cancer (Actual) |
|---|---|---|
| Cancer (Predicted ) | 50 | 50 |
| Not Cancer (Predicted) | 100 | 800 |
So in our example:
Accuracy = 50 + 800 / 50 + 50 + 100 + 800
Precision = 50/ 50 + 50
Recall = 50/ 50 + 100