This paper presents methods for prediction of energy consumption of different appliances in homes. The aim is to predict the next day electricity consumption for some services in homes. Historical data for a set of homes in France was used. Two basic predictors are tested and a stochastic based predictor is proposed. The performance of the predictors is studied and it shows that the proposed predictor gives better results than other approaches. Two processings are proposed to improve the performance of the predictor, segmentation and aggregation of data. Application results are provided. (C) 2012 Elsevier Ltd. All rights reserved.