Improving on recommend speed of recommender systems by using expert users
Wenqiang Xiao, Yao Shijun, Shanming Wu · 2016
Owing to the fact that many algorithms aimed at improving the accuracy of the recommendation have been continuously proposed, it becomes more and more difficult to continue to improve the accuracy of the recommendation results. However, in real online recommender systems, besides the accuracy, the speed of recommendation is also a major factor. We found that the dimension of item vector is enormous when we need to calculate the similarity between two items. In this paper, to solve the problem, we introduce four methods to select small parts of user data and test items-based collaborative filtering algorithm. It not only reduced the impact of noise user data on the results, but also increased the recommend speed. The most interesting finding is that the accuracy is very close to the original result, and the speed is much faster than the basic items-based collaborative filtering recommendation algorithm.