Start Date
12-18-2013
Description
Collaborative filtering recommender systems suffer from lack of diversity as well as the scalability and the sparsity problems. Introduced is a new recommendation idea called LCSC that generates recommendations from those who previously recommended the target customers successfully while CF generates recommendations from those who are most similar to the target customer. The LCSC method is characterized by (1) the establishment and use of recommender networks to maintain recommendation history that serves as the basis of recommendation decision and (2) a small scope of search for supportive neighbors from whom recommended items are collected. Experiments with real transactional data confirm that LCSC's recommendation accuracy is as good as CF's, while LCSC has significant advantages over CF in computational efficiency and recommendation diversity.
Recommended Citation
Ryu, Young; Kim, Jae Kyeong; and Kim, Hyea Kyeong, "A Local Scoring Model for Diversified Recommendation" (2013). ICIS 2013 Proceedings. 19.
https://aisel.aisnet.org/icis2013/proceedings/KnowledgeManagement/19
A Local Scoring Model for Diversified Recommendation
Collaborative filtering recommender systems suffer from lack of diversity as well as the scalability and the sparsity problems. Introduced is a new recommendation idea called LCSC that generates recommendations from those who previously recommended the target customers successfully while CF generates recommendations from those who are most similar to the target customer. The LCSC method is characterized by (1) the establishment and use of recommender networks to maintain recommendation history that serves as the basis of recommendation decision and (2) a small scope of search for supportive neighbors from whom recommended items are collected. Experiments with real transactional data confirm that LCSC's recommendation accuracy is as good as CF's, while LCSC has significant advantages over CF in computational efficiency and recommendation diversity.