DSLOB

DSLOB数据集是由南加州大学和J.P.Morgan AI Research合作创建的合成限价订单簿数据集，旨在为金融市场的预测算法提供一个基准测试平台。该数据集通过多代理市场模拟器ABIDES生成，包含365天的交易数据，涵盖了正常市场条件、小冲击和大冲击三种不同的市场状态。数据集的创建过程涉及精确控制市场冲击的参数，以模拟真实世界中的分布偏移。DSLOB数据集特别适用于测试和改进机器学习模型在面对高频时间序列数据中的分布偏移时的鲁棒性，尤其是在金融趋势预测等应用领域。The DSLOB Dataset is a synthetic limit order book dataset co-developed by the University of Southern California and J.P. Morgan AI Research, designed as a benchmark platform for predictive algorithms in financial markets. This dataset is generated via the multi-agent market simulator ABIDES, containing 365 days of trading data covering three distinct market states: normal market conditions, small market shocks, and large market shocks. The creation process of the DSLOB dataset involves precise control of market impact parameters to simulate distribution shifts in real-world scenarios. The DSLOB dataset is particularly suitable for testing and enhancing the robustness of machine learning models against distribution shifts in high-frequency time series data, especially in application domains such as financial trend forecasting.