Research Article
A Recommendation Algorithm for Improved Residual Networks Based on Matrix Factorization
@INPROCEEDINGS{10.1007/978-3-031-80713-8_2, author={Chengzhi Mao and Zhifeng Wu}, title={A Recommendation Algorithm for Improved Residual Networks Based on Matrix Factorization}, proceedings={Data Information in Online Environments. 4th EAI International Conference, DIONE 2023, Nanchang, China, November 25--27, 2023, Proceedings}, proceedings_a={DIONE}, year={2025}, month={2}, keywords={Recommendation algorithm Matrix factorization residual network}, doi={10.1007/978-3-031-80713-8_2} }- Chengzhi Mao
Zhifeng Wu
Year: 2025
A Recommendation Algorithm for Improved Residual Networks Based on Matrix Factorization
DIONE
Springer
DOI: 10.1007/978-3-031-80713-8_2
Abstract
Recommender system is widely used in e-commerce, news consulting, social networking, tourism, film, music and other fields because it can effectively deal with the problem of information overload. According to the practical problems in the recommendation system, various recommendation algorithms are produced. Traditional recommendation algorithms are divided into content-based recommendation, collaborative filtering based recommendation and hybrid recommendation. Collaborative filtering algorithms are favored because they can extract and process relevant information features to accurately predict user preferences. However, collaborative filtering algorithms generally have problems such as sparse data, cold start and data scalability. This paper presents the Associated Residual Matrix Factorization (ARMF) model, which can solve the common cold start problem and improve the recommended performance. The model uses associative residual neural networks (ARNs) to model higher-order interactions between the user and the underlying features of the item and to avoid network layer gradients exploding or disappearing and overfitting problems. In order to verify the validity and correctness of the model, this paper conducts tests on Movielens100k and Movielens10M datasets, and the experimental results show that the prediction results are 17%~ 23% better than other comparison algorithms.
