The Effectiveness of Collaborative Filtering-Based Recommendation Systems across Different Domains and Search Modes

Abstract

Collaborative filtering (CF) is a personalization technology used by numerous e-commerce

websites to generate recommendations for users based on others evaluations. Although many

studies have considered ways to refine CF algorithms, little is known about the effects of user

and domain characteristics on the accuracy of CF systems. This study investigates the effects of

two factors, domain and user search mode, on the accuracy of collaborative-filtering systems,

using data collected from two different experiments ― one conducted in a consumer-product

domain and one in a knowledge domain. The results show that the search mode employed by

users strongly influences the accuracy of recommendations. CF works better when users are

looking for specific information than when they are browsing out of general interest. Accuracy

drops significantly when data from different search modes are mixed. The results also show that

CF is more accurate in knowledge domains than in consumer-product domains. The study implies

that CF systems in either domain will provide more accurate recommendations if they identify

and accommodate users search modes.