LAUSR

Item-based collaborative filtering (ItemCF) is a proven algorithm in recommendation systems that is based on a neighborhood algorithm, but it neglects the influence of sentiment in different aspects. However, customers always express their opinion in reviews, and these personalized data will influence the recommendation effect. This paper proposes an aspect sentiment collaborative filtering algorithm (ASCF), which combines sentiment analysis with a fuzzy Kano model. ASCF obtains the users’ different attitudes toward aspects of the product by fine-grained sentiment analysis from the user’s purchase records, and then analyzes the user’s degree of desire and importance for each feature based on the fuzzy Kano model, proposing a novel similarity measure method with user preferences for a collaborative filtering algorithm. Experiments with Amazon data sets show that ASCF effectively improves the precision of ItemCF and opinion-enhanced collaborative filtering; it provides higher recommendation precision and fewer product recommendations at the similarity precision. The experiments use the smartphone catalog as an example to analyze the aspect-characteristic words distribution matrix.