Synthetic Rating Datasets

本研究提出了一种生成合成评分数据集的方法，旨在为推荐系统的离线评估提供数据支持。该数据集通过模仿现有数据集的特性，生成具有可配置用户数量的合成数据。数据集的创建过程包括用户聚类和分布学习，以及评分采样，确保生成的数据能够反映不同用户群体的行为。该数据集主要应用于推荐系统的评估，特别是在缺乏公开可用评分数据集的情况下，用于验证推荐算法的性能。

This study proposes a method for generating synthetic rating datasets to provide data support for the offline evaluation of recommendation systems. This dataset generates synthetic data with configurable user counts by replicating the characteristics of existing real-world datasets. The dataset creation process includes user clustering, distribution learning, and rating sampling, ensuring that the generated data can reflect the behaviors of diverse user groups. This dataset is primarily utilized for the evaluation of recommendation systems, particularly for validating the performance of recommendation algorithms when publicly available rating datasets are scarce.