“…In this context, the correct disaggregated load đ for the aggregate load đ is known in advance. Various supervised learning approaches, such as linear regression [212], genetic algorithms [153], [200], [249], SVM [122], [188], [189], [250], random decision forest [122], [183], [212], [251], hidden Markov models (HMM) [215], [252], decision trees [3], [212], [253], naive Bayes classifier [254], kNN [3], [171], [188], and deep learning neural networks [109], [110], [112], [116], [197], [240], [255], have been applied in the NILM literature. As outlined in Table IV, supervised NILM is mainly used for regression and classification tasks.…”