Collaborative Filtering based Electricity Plan Recommender System in Python

The deregulation of the electricity market enables residential customers to select suitable electricity retailing plans. This paper proposes a Bayesian hybrid collaborative filtering-based electricity plan recommender system (BHCF-EPRS), which is constructed in a two-stage model integrated with model-based and memory-based collaborative filtering methods. Bayesian inference is developed for missing feature estimation and user classification. Free from the requirements on total electricity use data and historical plan transaction data, the BHCF-EPRS can recommend suitable retailers and plans based on some easily obtainable features quantifying home appliance usage patterns. The BHCF-EPRS is verified to be a reliable recommender system with low error in full-ranking recommendation and high precision in top-N recommendation, which can improve the competitive operation of the electricity market.