Once the training is complete, the user can click on the See results button to see how the models behave. The metrics are computed on molecules belonging to a test set, never seen by the model during the optimization. We split the initial dataset randomly into train and test sets (detailed in the following page of the documentation).
The performance page contains a summary of performance metrics for the model trained for each individual target, associated plots, and the confusion matrix (see images below).
Explanations to interpret the performances of the models and select good models are given in the following article: QSAR model scores interpretation.