基于用户显式与隐式反馈的购物推荐算法

A Recommendation Algorithm for Shopping Based on Explicit and Implicit Feedback of User

摘要: 商业网站通常采用推荐系统来处理大量信息。新用户行为数据的缺乏导致系统无法为其提供推荐，这一问题被称为冷启动问题。此外，这些算法严重依赖于稀疏的显式反馈数据，使得为用户提供精确推荐变得具有挑战性，从而引发了数据稀疏性问题。为了解决这些问题，本文提出了一种基于用户显式和隐式反馈的购物推荐算法。首先，该算法使用特殊过滤器提取特征并生成候选列表。其次，通过交互学习模型和多隐式反馈学习模型，从候选列表中每个项目的评分数据中提取隐式关系。多隐式反馈学习模型从三个角度分析隐式关系，并将这三种隐式反馈作为辅助数据来解决数据稀疏性问题。最后，使用Tafeng数据集和Book Crossing数据集验证了该算法在购物推荐中的有效性。

Abstract: Business websites typically employ recommendation systems to process large amounts of information. The lack of behavioral data for new users leads to the system’s inability to make recommendations for them, which is known as the cold-start problem. Additionally, these algorithms heavily rely on sparse explicit feedback data, making it challenging to provide users with precise recommendations, leading to the data sparsity problem. To address these issues, a recommendation algorithm for shopping based on explicit and implicit feedback from the user is proposed. Firstly, the algorithm uses special filters to extract features and generate a candidate list. Secondly, the interaction learning model and the multi-implicit feedback learning model are used to extract implicit relationships from the rating data for each item in the candidate list. The multi-implicit feedback learning model analyzes the implicit relationship from three perspectives. The three types of implicit feedback are used as auxiliary data to solve the data sparsity problem. The effectiveness of the algorithm in shopping recommendation is validated using the Tafeng dataset and the BookCrossing dataset.