Long-Tail Recommendation Framework Using Frequent Neighbors
Jing Qin · 2020
The neighborhood-based recommendation method is a basic collaborative-filtering method, and one of the methods still used in the industry. The main basis for recommendation is the user's scoring data in the neighborhood-based recommendation method. However, due to the sparsity of user scoring data, the number of recommendations for long-tail items in the neighborhood-based method is relatively small. Although deep-learning technology has been widely applied to the neighborhood-based method, user scoring data are still needed in the methods, so the number of long-tail items in recommendation results is still very small. In order to increase the number of long-tail items recommended in the neighborhood-based method, this paper proposes a long-tail recommendation framework using frequent neighbors, and applies the framework to two real-world datasets. Experiments showed that the recommendation framework proposed in this paper effectively increases the number of long-tail items recommended in the neighborhood-based method and significantly improves the coverage metric.