Interest community-based recommendation via cognitive similarity and adaptive evolutionary clustering

The interest community based collaborative filtering recommendation model aims to make recommendations in user interest communities, thereby reducing time complexity while fully leveraging user preferences. Nevertheless, previous research has not delved into the discovery of interest communities in heterogeneous information networks (HINs) encompassing multiple decision influences. To address this issue, we present an interest community-based recommendation approach based on cognitive similarity and adaptive evolutionary clustering, denoted as IC-AEC. To mitigate data sparsity, we construct a HIN containing multiple decision-influencing factors, i.e., cognitive similarity, item genres, and user preferences, thus enriching the network information. Furthermore, we design a novel adaptive evolutionary clustering method to detect interest communities in this HIN. Our adaptive evolutionary clustering evolves user interest states based on the complex relations of HIN, and relies on stable state values to partition user communities. Eventually, we propose a similarity measurement method that combines user preferences and item influence to calculate the similarity between users within the community for rating prediction. Theoretical analysis and experimental results on six real datasets demonstrate that IC-AEC outperforms superior approaches in prediction ratings and recommendation performance.