Reuters & Bloomberg (R&B) dataset

该数据集包含了552,909篇新闻文章，总计覆盖了2,605天，其中有1,794个交易日，用于根据新闻文章预测股票价格变动。在创建样本的交易日集合时，该数据集剔除了模糊不清的样本，最终包含了32,204个清晰样本。数据集被随机划分为60%的训练集、20%的验证集以及20%的测试集。其规模涉及552,909篇文章和2,605天，任务是对股票价格变动进行分类。

This dataset includes 552,909 news articles spanning 2,605 days in total, among which 1,794 are trading days, and is utilized for stock price movement prediction based on news articles. When building the trading day sample set, ambiguous samples were removed, leaving 32,204 unambiguous valid samples in the final dataset. The dataset is randomly divided into three subsets: 60% for training, 20% for validation, and 20% for testing. With a scale involving 552,909 articles across 2,605 days, the core task of this dataset is to classify stock price movements.