In multi-criteria collaborative filtering, the data sparsity is a critical factor affecting the effectiveness of the algorithm. Existing solutions generally remove low-efficiency data through dimensionality reduction,...
详细信息
ISBN:
(纸本)9781538683408
In multi-criteria collaborative filtering, the data sparsity is a critical factor affecting the effectiveness of the algorithm. Existing solutions generally remove low-efficiency data through dimensionality reduction, decomposition, etc. While at lower data dimension, they also lose original data information, which impacts the accuracy of predicted ratings. This paper adopts a new processing strategy and proposes to use the matrix filling method to handle the missing data. Firstly, an improved similarity algorithm combined with the Jaccard Similarity Coefficient is used to calculate the predicted rating. Secondly, the Reliable Factor is introduced to prevent the error of filling data. Finally, multiple linear regression is employed for aggregating the ratings of the criteria to eventually yield preferred predicted ratings. A series of experiments show that the proposed method can precisely reflect the user's similarity and achieve better results in predicted ratings, in contrast to traditional methods.
暂无评论