Adaptive recommendation: putting the best foot forward
John O’Donovan, John Dunnion · 2004
Information is becoming increasingly available in digital formats such as Web Pages, MP3 files and many others. This puts more emphasis on the need for reliable information filtering techniques. New recommendation algorithms are continuously being developed to deal with the problem of information overload. In this paper we present a new, regression-based approach to the application of recommendation algorithms. We classify five different datasets based on a range of metrics, including sparsity, user-item ratio and the distribution of user ratings. From performance analysis tests of four predictive algorithms over these sets, we develop a regression function to predict the suitability of a particular recommendation algorithm for a previously unseen dataset. Our results show that the best-performing algorithm on the new set is the one predicted by our regression analysis.