Collaborative filtering method based on bi-clustering filling

Abstract

The invention relates to a collaborative filtering method based on bi-clustering filling. The method comprises the steps of missing item filling based on bi-clustering: for each missing item, finding out all submatrixes including the missing item in a whole original scoring matrix, calculating a mean square residual Hmin (m, n) of all submatrixes, and estimating the value of the missing item according to a submatrix with the smallest mean square residual; user-based collaborative filtering recommendation: respectively giving different reliability weights to original data and filling data, calculating the similarity between a target user and other users according to the reliability weights, taking a set of a plurality of users with the highest similarity to the target user as a nearest neighbor set of the target user, and predicting the score to goods by the target user according to scoring information of the nearest neighbor set of the target user; and for each user, recommending a plurality of goods with the highest score. The collaborative filtering method based on the bi-clustering filling provided by the invention introduces a reliability matrix to distinguish a real scoring item and a filling item to improve a similarity function and a prediction scoring function, thereby improving the influence of data sparsity of the scoring matrix.

Description

technical field [0001] The invention belongs to the field of collaborative filtering recommendation, in particular to a collaborative filtering method based on bi-clustering filling. Background technique [0002] With the rapid development of the Internet and e-commerce, the information on the website has increased dramatically, and it has become increasingly difficult for people to quickly locate the information they need from the massive data. People are still suffering from a lack of information in the ocean of data, and the phenomenon of information overload is becoming more and more serious. In this context, personalized recommendation systems are increasingly valued. [0003] Data sparsity is the biggest challenge facing collaborative filtering. In actual commercial recommendation systems, the number of users and items is very large, and users often only have rating records on a few items, resulting in the actual rating matrix being very large. Sparse, usually the ev...

Images