Spectral Clustering-based Matrix Completion Method for Top-n Recommendation
Qingmei Zhou, Xin Chen, Jiuya Zhang · 2019
In the era of big data on the Internet, top-N recommender systems constitute a mission-critical technology in e-commerce applications. However, sparse data is one of the key factors leading to poor performance of the recommendation algorithm. In this paper, we propose a low-rank matrix completion recommendation scheme based on spectral clustering. The underlying idea is that spectral clustering based on rating and item classification information is used to achieve the division of the similar user group. Furthermore, we use low-rank matrix completion to effectively predict the unrated items in the sub-matrix of the spectral clustering. The real dataset experimental results show that our proposed scheme often outperform other state-of-the-art algorithms in top-N recommendation.