Cross-Domain Recommendation: Challenges, Progress, and Prospects

Feng Zhu,Yan Wang,Chaochao Chen,Jun Zhou,Longfei Li,Guanfeng Liu

DOI: https://doi.org/10.48550/arXiv.2103.01696

2021-03-02

Abstract:To address the long-standing data sparsity problem in recommender systems (RSs), cross-domain recommendation (CDR) has been proposed to leverage the relatively richer information from a richer domain to improve the recommendation performance in a sparser domain. Although CDR has been extensively studied in recent years, there is a lack of a systematic review of the existing CDR approaches. To fill this gap, in this paper, we provide a comprehensive review of existing CDR approaches, including challenges, research progress, and future directions. Specifically, we first summarize existing CDR approaches into four types, including single-target CDR, multi-domain recommendation, dual-target CDR, and multi-target CDR. We then present the definitions and challenges of these CDR approaches. Next, we propose a full-view categorization and new taxonomies on these approaches and report their research progress in detail. In the end, we share several promising research directions in CDR.

Information Retrieval,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the data sparsity problem in recommendation systems. Specifically, Cross - Domain Recommendation (CDR) has been proposed to utilize the relatively rich information in a more information - rich domain (the source domain) to improve the recommendation performance in another domain with relatively scarce information (the target domain). Although CDR has been widely studied in recent years, there is a lack of a systematic review of existing CDR methods. To fill this gap, this paper provides a comprehensive review of existing CDR methods, including the challenges faced, research progress, and future prospects.

The paper points out that traditional Recommender Systems (RSs) have difficulties in dealing with the data sparsity problem, especially in the cold - start problem of new users or new items. These problems may lead to over - fitting when training models based on Collaborative Filtering (CF), thereby significantly reducing the recommendation accuracy. Therefore, CDR has become an effective means to solve these challenges by using the rich information in the source domain to improve the recommendation accuracy in the target domain.

The article further defines different types of CDR methods, proposes the definitions, challenges, classifications, and new classification systems of these methods, and reports in detail the research progress of each method. Finally, the article explores some promising development directions in the CDR field, providing guidance for future scientific research work.