Design Matrix Factorization and Knowledge-Guided Clustering for Trust-Aware Cross-Domain Recommendation Systems

The core challenge in designing an algorithm for trust-aware cross-domain recommendation systems lies in effectively integrating ratings from all domains (target-domain and source-domains) and trust relationships to improve recommendation performance in the target-domain. These systems involve two types of information fusion: intra-domain information fusion, which integrates intra-domain trust relationships and ratings within each domain (target-domain or source-domain), and cross-domain information fusion, which fuses information from all domains through inter-domain trust relationships as a bridge. The existing algorithms are limited in that they only address either intra-domain information fusion or cross-domain information fusion, failing to comprehensively utilize both intra-domain and inter-domain trust relationships for integrating information from all domains. To address this limitation, we propose a novel approach that combines inter-domain information fusion and cross-domain information fusion. Based on this approach, we design a matrix factorization and knowledge-guided clustering based trust-aware cross-domain collaborative filtering algorithm (MFKC _ TCCFA). Experimental results on a dataset show that MFKC_TCCFA outperforms existing state-of-the-art algorithms.