An improved collaborative filtering recommendation algorithm not based on item rating

Abstract

As e-commerce grows fast nowadays, recommender systems have become an integral part of every electricity business. A number of the recommendation algorithms need score matrix (i.e., matrix that is used to record the data of the score that users value the item) as a mean of input. However, in many cases, the data only obtained the user's record matrix (i.e., matrix that contained only whether users have purchased or downloaded the item, without a score that is about a particular range), instead of the users' score matrix. Under this circumstance, the record matrix fails to reflect the preference of the user, the function of the recommendation algorithm declined. The feature of the improved algorithm the paper presents that, by recording a neighbor user (i.e., a similar user) data of purchase or download history, the current users' preference of the item can be predicted, and by record matrix authors can predict users' preferences of an item, thereby improve the effectiveness of recommendation algorithm which requires score matrix as an input.