Time Series Analysis of Household Electric Consumption with XGBoost Model

Abstract

Due to the improvement of population quality of life over the world and the following increase of energy demand in particularly the electricity, it has become necessary to follow the evolution of its consumption. Electricity consumption forecasting is considered as key factor in a process of improving energy efficiency, controlling consumption and reducing costs. The main objective of this paper consist to propose a forecast model for household electricity consumption using XGBoost regressor applied on a dataset which contains data collected from a house situated in Sceaux (Paris, France) between December 2006 and November 2010. The experimental results show that the proposed model achieved a higher performance for forecasting periods, particularly, in hourly and daily granularities in terms of RMSE and MAEP.