Research Article
Time-Based Distributed Collaborative Filtering Recommendation Algorithm
@INPROCEEDINGS{10.1007/978-3-030-94763-7_19, author={Qiao Li and Xiantong Hu and Linfei Zhou and Xiao Zheng and Wei Zhao and Yunquan Gao and Xuangou Wu}, title={Time-Based Distributed Collaborative Filtering Recommendation Algorithm}, proceedings={Mobile Networks and Management. 11th EAI International Conference, MONAMI 2021, Virtual Event, October 27-29, 2021, Proceedings}, proceedings_a={MONAMI}, year={2022}, month={1}, keywords={Collaborative filtering ALS Time factor RMSE}, doi={10.1007/978-3-030-94763-7_19} }- Qiao Li
Xiantong Hu
Linfei Zhou
Xiao Zheng
Wei Zhao
Yunquan Gao
Xuangou Wu
Year: 2022
Time-Based Distributed Collaborative Filtering Recommendation Algorithm
MONAMI
Springer
DOI: 10.1007/978-3-030-94763-7_19
Abstract
Recommendation systems based on collaborative filtering are widely used in many fields. Alternating Least Squares (ALS) in the Mlib Library is a distributed and parallel algorithm in Spark framework, which can solve the problems of scalability and speedup in a limited hardware resources of stand-alone systems. However, it does not consider the influence of the factor of time on the recommendation accuracy. Taking restaurant ratings as an example, this month ratings are more reliable than those from last year. Thus, the motivation in our proposal re-scores ratings with different time weights. We improve ALS in its process of data preparation according to requirements on the structure of data input. Consequently, our improvement does not need to modify the main body of ALS. Experimental results validate effectiveness that our proposal outperforms the original ALS in recommendation accuracy.
