Cluster-Based Smoothing and Linear-Function Fusion for Collaborative Filtering

Memory-based and model-based approaches are two approaches for collaborative filtering. In the past, the memory-based approaches have been shown to suffer from data sparsity and difficulty in scalability. Alternatively, the model-based approaches limit the range of users. Existing research on hybrid approaches try to solve these problems by combining advantages of these two methods. However, without taking full use of known information, they still have the problem of low prediction accuracy. This paper proposes a novel hybrid framework to solve these problems. The first step is to combine the memory-based and the model-based approaches to compute prediction ratings in three different aspects: predictions based on ratings of the same item by other users, predictions based on different item ratings made by the same user, predictions based on the data of different but similar users rating on different but similar items, and the second step is to employ a linear function to get the final rating by fusing the three prediction ratings. As a result, our approach conquers data sparsity, provides higher accuracy, and increases efficiency in recommendations.