Abstract
To facilitate effective and efficient recommendations of digital contents over a large customer base, a customerdriven recommender is developed utilizing the fast diffusion and information sharing capability of customer networks. A traditional centralized recommender system based on collaborative filtering collects for a host customer a product recommendation list from most similar customers (called neighbors) that have been determined globally by checking the whole customer base. Such a procedure is a very effective method, but its computational overhead can be prohibitive when the customer base is large. The proposed method, called the customer-driven recommender system, follows the collaborative filtering principle, but performs distributed and local search for neighbors from whom a recommendation list for a host customer is generated. In order to validate the effectiveness and performance of the proposed method, we build a customer network for digital contents recommendation using real data in a mobile commerce and compare it against the traditional collaborative filtering-based system. Furthermore, experiments are designed and performed to show why the local search mechanism of the customer-driven recommender system is as accurate as, but much more computationally efficient than, the global search mechanism of the collaborative filtering-based recommender system.
Original language | English |
---|---|
Pages | 158-164 |
Number of pages | 7 |
Publication status | Published - 2007 |
Event | 17th Workshop on Information Technologies and Systems, WITS 2007 - Montreal, QC, Canada Duration: 8 Dec 2007 → 9 Dec 2007 |
Conference
Conference | 17th Workshop on Information Technologies and Systems, WITS 2007 |
---|---|
Country/Territory | Canada |
City | Montreal, QC |
Period | 8/12/07 → 9/12/07 |
Keywords
- Customer Network
- Recommendation Accuracy and Performance
- Recommender System