Performance of the ensemble learner.
Top: the Receiver Operating Characteristic (ROC) curve for both current and forecast failure prediction. The ROC represents the range of possible trade-offs between the classifiers’ true positive (truly failed pump classified as a failure) and false positive (truly functional pump classified as a failure) rates when choosing a threshold to operationalize the classifier. Bottom: the true positive rate (TPR), true negative rate (TNR), positive predictive value (PPV), and negative predictive value (NPV, not to be confused with net present value) are plotted as a function of the learner’s probability threshold. This bottom panel illustrates the relationship between learner performance and the implementer-defined probability threshold to decide of a pump is broken (current) or will break (forecast) on any given day.