Figure 5.

Sequential backward selection (SBS). To improve computational efficiency and reduce generalization error, the sequential backward selection algorithm aims to reduce the dimensionality of the initial feature subspace from N to K features with minimal model performance loss. To obtain the list of K features, sequentially remove features from a given features list of N features. By including seven characteristics of the dataset, excluding cadence, we maximized the performance of the algorithm, which, in our case, turned out to be KNN.