Application of wavelet-based ensemble tree classifier for non-intrusive load monitoring

This paper presents an application of discrete wavelet and ensemble decision tree classifier to the non-intrusive load monitoring (NILM). The effect of different order of Daubechies wavelet filter on the classification accuracy is investigated. Also the paper studies the effect of increasing the number of decision trees contained in the ensemble on the performance of the classifier by measuring the training and testing classification accuracies. The results have shown that the use of third order Daubechies wavelet filter can lead to highest classification accuracy compared other order of Daubechies filters. The results also have shown that when increasing the number of decision trees in the ensemble classifier can have significant effect on improving the classification accuracy in NILM.