Abstract
Personalized recommendations enable firms to effectively target customers with products and services. Such systems are often based on association rules. While there has been considerable work done on mining rules more efficiently, there is very little prior research that examines how to use rules effectively when making recommendations. Traditional association rule-based recommendation systems have relied on identifying one rule from the several eligible ones to make the recommendation. This ignores information from other eligible rules that can potentially improve the recommendation. We propose a method to combine multiple rules when making recommendations. In doing so, we also present an approach to select the best combination of rules from the many that might be available.
Original language | English (US) |
---|---|
Pages | 13-18 |
Number of pages | 6 |
State | Published - 2009 |
Externally published | Yes |
Event | 19th Workshop on Information Technologies and Systems, WITS 2009 - Phoenix, AZ, United States Duration: Dec 14 2009 → Dec 15 2009 |
Conference
Conference | 19th Workshop on Information Technologies and Systems, WITS 2009 |
---|---|
Country/Territory | United States |
City | Phoenix, AZ |
Period | 12/14/09 → 12/15/09 |
Keywords
- Bayesian estimation
- Maximum likelihood
- Mutual information
- Personalization
ASJC Scopus subject areas
- Information Systems
- Control and Systems Engineering