After training a classification or prediction model, the next phase is to test the model. This consists of two parts: first applying the trained model to a test set, and then comparing the predicted values with the actual values.
In this section, we introduce model evaluation techniques for a predictive model. The target variable of the model can be categorical or continuous. Possible model evaluation techniques include both a number of different accuracy statistics and graphs.
Evaluating Classification Model Performance
A classification model can be evaluated by different classification accuracy statistics:
- confusion matrix
- class prediction statistics
- overall accuracy statistics
In addition to these statistics, graphs such as the receiver operating characteristic (ROC) curve and lift chart provide additional information about the classification model performance.