AIML-TUDA/dlam-ts-project-data-2026

针对匿名操作单元的多变量小时级预测数据集，用于预测每个系列ID的未来小时操作负载指数，较高值表示该单元操作压力更大。数据集包含训练、验证输入和预测索引文件，覆盖96个系列，每个系列有4992个时间步，目标列为target，使用WAPE作为主要评估指标。

Multivariate hourly forecasting for anonymized operations units. Predict the future hourly operational load index for each series_id, with higher values indicating more operational pressure. The dataset includes training, validation input, and forecast index files, covering 96 series with 4992 timesteps per series, target column target, and uses WAPE as the primary metric.