Crude Oil Benchmark (COB)
收藏资源简介:
Crude Oil Benchmark (COB) 数据集由加州大学伯克利分校电子工程与计算机科学系创建,包含30年的全球主要原油价格数据,特别是WTI、Brent和Dubai三种原油。数据集通过转换资产价格数据为波动性代理,使用期望最大化(EM)算法生成与实际事件对齐的上下文任务标签。该数据集旨在解决金融领域中时间序列数据分布变化的问题,特别是在处理全球重要资产的实际数据时。数据集的应用领域包括持续学习和分布外检测,旨在通过提供丰富的上下文标签来提高预测模型的性能。
The Crude Oil Benchmark (COB) dataset was created by the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley. It encompasses 30 years of global leading crude oil price data, specifically covering three major benchmarks: WTI, Brent, and Dubai. This dataset converts asset price data into volatility proxies, and leverages the Expectation-Maximization (EM) algorithm to generate contextual task labels aligned with real-world events. The dataset is designed to address the issue of distribution shift in time-series financial data, particularly when processing real-world data of globally critical assets. Its application domains include continual learning and out-of-distribution detection, with the goal of improving the performance of predictive models by providing rich contextual task labels.

- 1Real World Time Series Benchmark Datasets with Distribution Shifts: Global Crude Oil Price and Volatility加州大学伯克利分校电子工程与计算机科学系 · 2023年



