Causal neural fuzzy inference modeling of missing data in implicit recommendation system

Highlights：

• The contributions of this paper are summarized as follows: 1. The impact factors of exposure probability are analyzed and processed from causal perspective using fuzzy set theory. 2. Causal neural fuzzy inference network is designed to predict the exposure probability in implicit recommendation. 3. A novel optimization method is proposed and used for the CNFI and MF mixed recommendation model; 4. Extensive experiments are performed on three reality datasets and validate the effectiveness of CNFI.

摘要

•In this paper, we propose a novel algorithm to model missing data for implicit recommendation, named Causal Neural Fuzzy Inference (CNFI). It designs a causal neural fuzzy inference network to predict the probability of users exposure to items (also named exposure probability for convenience). It considers the causal relationship in the recommendation system, and applies fuzzy theory and approach to model the missing data in implicit feedback.•The contributions of this paper are summarized as follows: 1. The impact factors of exposure probability are analyzed and processed from causal perspective using fuzzy set theory. 2. Causal neural fuzzy inference network is designed to predict the exposure probability in implicit recommendation. 3. A novel optimization method is proposed and used for the CNFI and MF mixed recommendation model; 4. Extensive experiments are performed on three reality datasets and validate the effectiveness of CNFI.