Electricity Retail Plan Recommendation Method Based on Multigranular Hesitant Fuzzy Sets and an Improved Non-Negative Latent Factor Model

Electricity retail companies can derive significant benefits from precise recommendations of electricity retail plans (ERPs). However, existing recommendation methods often assume that customers are proficient in evaluating all the attributes of ERPs, and overlook the fact that the accuracy of predicting missing information is closely tied to the objective function of customers’ satisfaction, which degrades the recommendation results significantly. In light of the challenge, an ERP recommendation method based on multigranular hesitant fuzzy sets (MHFSs) and an improved non-negative latent factor model (INLFM) is proposed. First, a quantitative model for customer satisfaction based on MHFSs is established, which provides a foundation for estimating target customers’ satisfaction. Secondly, an INLFM-based prediction model is developed to fill in the missing values of customers’ satisfaction. Additionally, an estimation model for target customer satisfaction based on a customer portrait label system and a dual-layer affinity propagation (DLAP) clustering algorithm is proposed, and a top-H ERPs recommendation method is developed, facilitating precise ERP recommendation tailored to the needs of electricity retail company. Finally, case studies on customers in a high-tech development zone in eastern China show that the proposed method can characterize customers’ satisfaction more accurately and equitably, meanwhile reduce the recommendation deviation effectively.