Recommendation as Classification: Using Social and Content-Based Information in Recommendation

Recommendationsystems makesuggestions about artifacts to a user. For instance, they maypredict whether a user wouldbe interested in seeing a particular movie. Social recomendationmethodscollect ratings of artifacts from manyindividuals and use nearest-neighbor techniques to makerecommendationsto a user concerning newartifacts. However,these methodsdo not use the significant amountof other information that is often available about the nature of each artifact -- such as cast lists or moviereviews, for example.This paper presents an inductive learning approach to recommendationthat is able to use both ratings informationand other forms of information about each artifact in predicting user preferences. Weshow that our method outperformsan existing social-filtering methodin the domainof movierecommendationson a dataset of more than 45,000 movieratings collected from a community of over 250users.

• Type: Conference
• Year: 1998
• Where: AAAI
• Authors: Chumki Basu, Haym Hirsh, William W. Cohen
• Keywords: AAAI 1998, Inductive Learning Approach, Intelligent Agents, Social Recomendationmethodscollect Ratings, User Wouldbe