The Impact of Imputation Methods on the Classification of Household Devices from Electricity Usage Time Series
Roberto Santana · 2023
The classification of household devices from electricity usage profile is a relevant problem that has been addressed using different time series classification algorithms. While missing data is common in time series from the energy domain, it is an open question how data-imputation impacts the performance of classification algorithms. In this paper we empirically investigate the influence of imputation strategies in the performance of state-of-the-art time series classifiers for this problem. We show that for some datasets the combined application of simple imputation strategies and advanced classifiers can improve the best accuracy results.