A collaborative filtering recommendation algorithm based on biclustering

Collaborative filtering has been widely used in many fields such as movie recommendation and e-commerce. However, there are still some problems such as data sparsity which restrict its further development. To address the data sparsity problem we proposed a novel collaborative filtering recommendation algorithm based on biclustering. Firstly, we use biclustering algorithm simultaneous clustering of the rows and columns of the rating matrix to generate biclusters, then the missing data can be smoothed by using the information of the biclusters. Secondly, a weighted matrix is introduced to distinguish between the original data and the smoothing data. Lastly, the active user´s neighbors can be found based on the new similarity we proposed, and the recommendation to the active user is produced. The experiment results are offered, which show that the algorithm we presented can alleviate the data sparsity problem and improve the quality of the recommendation.