JIITA, vol.7 no.1, p.654-665, 2023, DOI: 10.22664/ISITA.2021.7.1.654
Sang Suh*, Monika Singh, Chirag Dave
Department of Computer Science. Texas A&M Universit
Abstract: Recommender systems are widely used in online e-commerce applications to improve user engagement and increase revenue. Many recommendation systems employ collaborative filtering technology, which has been proven to be one of the most successful techniques in recommender systems in recent years. With the gradual increase of customers and products in electronic commerce systems, a key challenge for recommender systems is providing high-quality recommendations to users in “coldstart” situations. We consider three types of cold-start problems: 1) recommendation on existing items for new users; 2) recommendation on new items for existing users; 3) recommendation on new items for new users. To solve the problems of scalability and sparsity in collaborative filtering, this paper proposed a solution to the cold start problem by combining the association rules with the clustering.
First, the items are clustered based on user ratings and price. Then, the user profile is enriched with the association rules on clustered data. The proposed approach utilizes item clustering collaborative filtering to produce the recommendations. The recommendation, combining “association rules, user clustering, and item clustering collaborative filtering” is more scalable and accurate than the traditional system. For association rules, we use FP-Growth algorithm instead of Apriori to mine frequent items because FP-Growth algorithm only needs two scans compared with Aprior’s multiple scans, which is more efficient.
Keywords: recommender systems, collaborative filtering, cold start, association rule, user clustering, item clustering, scalability, sparsity, accuracy, mean absolute error
Fullpaper: