Figure 8

Schematic illustration of the anomaly detection algorithm and determination of the detection score. This schematic illustration shows the training phase and anomaly detection algorithm of the deliberative control loop. The dataset is split up in a trainingset and a test set. The query execution times are labeled with behavior normal or as an anomaly. The trainingset is used to find the optimal detection score. Precision and recall are used to choose the score. The figure illustrates the detection step in the deliberative control loop.