Non-intrusive appliance load monitoring with bagging classifiers

Non-intrusive appliance load monitoring is an important problem class with interesting applications. Due to the difficulties that arise from the application of data mining techniques to real-world data sets, we close a gap in literature and focus on robust bagging classifiers. In this work, we answer the question, if the recognition rate of ensemble classifiers is significantly better than the recognition rate of the native classifiers. We analyse two types of bagging classifiers, i.e. (i) support vector machine and nearest neighbour ensembles and (ii) random forests. We compare their performance in terms of accuracy and robustness on a NIALM data set recorded in a field study. The experimental analysis concentrates on recognition rates w.r.t. various training set sizes, on the influence of neighbourhood sizes and the numbers of decision trees in random forest ensembles. It turns out that the decision tree ensembles belong to the best classifiers in the employed scenarios.