Contributions to Collaborative Filtering Research

In this report, I first introduced the three areas that of interest to collaborative filtering researchers are interested in and challenged by, namely: (a) how to solve the sparsity and scalability problems in recommendation systems; (b) how to synthesize variousrapidly develop and test collaborative filtering algorithms; and (c) how to apply collaborative filtering to implicit preference dataunbounded numeric preference data. In the next three chapters, I described my contributions to these three areas: (a) my work on the CoFE recommendation engine; (b) my work on the SVD based algorithm; and (c) my attempt to design a new algorithm for the iTunes play count data. A summary and prospects for future work are given in the final chapter. 3 Acknowledgements I would like to express my sincere gratitude to my advisor, Dr. Jon Herlocker, for his support during my two years study in computer science. It was he who gave me the opportunity to get to know the collaborative filtering and data mining. It was this opportunity that made me realize what I am interested in and would like to dedicate myself to. I am also grateful for his guidance during the writing phases of this report. I would also like to thank Dr. Francisco Martin for his support during my work on the iTunes recommendations project. His advice helped me not only to think deep into the algorithms, but also to grasp the big picture from the business perspectives. I would also like to thank Dr. Margaret Burnett for serving in my committee. I am also grateful for her support and encouragement. Thanks also go to my group members. I could not have had these project results without their teamwork. Finally, I would like to thank my husband for his understanding and support, but my appreciation would go beyond words.