A Hybrid of Local Optimization Selection for Recommendation Algorithm
Huiting Liu · Journal of Information and Computational Science · 2014
To overcome the impact on data sparsity of traditional collaborative filtering, we present a hybrid of local optimization selection for recommendation algorithm, named CCFHU. We firstly combine the similarity of feature attributes with the rating of items, to improve the computation result of the similarity of items. Then, we use local optimization options to select neighbors as a reference of the target group. Finally, we propose the content-based hybrid recommendation algorithm to amend abnormal predicted results in the collaborative filtering method. Our experimental results show that the CCFHU algorithm can reduce the negative impact of data sparsity on recommendations, improve the calculation accuracy of similarity between items, and effectively lower the mean absolute error of prediction consequences. Our experiments further contrast CCFHU with the content-boosted collaborative filtering and linear weighted method of collaborative filtering based on content prediction. It turns out that in terms of prediction accuracy, the CCFHU algorithm can achieve better results.