It is unquestionable that time series forecasting is of paramount importance in many fields. The most used machine learning models to address time series forecasting tasks are Recurrent Neural Networks (RNNs). Typically, those models are built using one of the three most popular cells, ELMAN, Long-Short Term Memory (LSTM), or Gated Recurrent Unit (GRU) cells, each cell has a different structure and implies a different computational cost. However, it is not clear why and when to use each RNN-cell structure. Actually, there is no comprehensive characterization of all the possible time series behaviors and no guidance on what RNN cell structure is the most suitable for each behavior. The objective of this study is two-fold: it presents a comprehensive taxonomy of all-time series behaviors (deterministic, random-walk, nonlinear, long-memory, and chaotic), and provides insights into the best RNN cell structure for each time series behavior. We conducted two experiments: (1) The first experiment evaluates and analyzes the role of each component in the LSTM-Vanilla cell by creating 11 variants based on one alteration in its basic architecture (removing, adding, or substituting one cell component). (2) The second experiment evaluates and analyzes the performance of 20 possible RNN-cell structures. Our results showed that the MGU-SLIM3 cell is the most recommended for deterministic and nonlinear behaviors, the MGU-SLIM2 cell is the most suitable for random-walk behavior, FB1 cell is advocated for long-memory behavior, and LSTM-SLIM1 for chaotic behavior.
翻译:不可质疑的是,时间序列的预测在许多领域具有至关重要的意义。 用于处理时间序列预测任务的最常用的机器学习模型是经常性神经网络。 通常,这些模型是使用三种最受欢迎的细胞之一,即ELMAN、长期短期内存(LSTM)或Gated经常单元(GRU)建立,每个单元格的结构不同,意味着不同的计算成本。 然而,不清楚为什么和何时使用每个 RNN- 细胞结构。 事实上,对于所有可能的时间序列行为没有全面的描述,也没有关于 RNNER 细胞结构最适合每种行为的指南。 本研究的目标是两重:它展示了所有时间序列行为的全面分类(定式、随机行、非线性、长模和混乱),并且提供了对每个时间序列行为的最佳 RNNNE 单元格结构的洞察。 我们进行了两次实验:(1) 第一次对LSTM- Van 单元格中每个组件的作用进行了全面的描述和分析, 第一次实验是SLMSLSL 长期的模型, 和MM 行为, 以创建最基本的变式的模型结构为基础, 显示其最变式的M 。