Activities
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/alebuenoaz/lstm-and-gru-time-series-forecasting
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了10个合成时间序列，每个序列都有五天的高活动量和两天的低活动量，这样的设计模拟了每周的电话通话量。此外，在对这个数据集进行最佳结果分析时，发现门控循环单元（GRU）在一步预测均方根误差（RMSE）方面显著优于长短期记忆网络（LSTM）和基准模型。该数据集的规模为10个时间序列，任务是对时间序列进行预测。
This dataset contains 10 synthetic time series, each featuring five days of high activity paired with two days of low activity. This pattern is designed to simulate weekly telephone call volumes. Furthermore, in the analysis of optimal results derived from this dataset, the Gated Recurrent Unit (GRU) was found to significantly outperform both the Long Short-Term Memory network (LSTM) and baseline models in terms of one-step prediction Root Mean Squared Error (RMSE). The dataset comprises 10 time series, with the task being time series forecasting.
