Abstract

Applicability of neural nets in time series forecasting has been considered and researched. For this, training of fully connected and recurrent neural networks on various time series with preliminary selection of optimal hyperparameters (optimization algorithm, amount of neurons on hidden layers, amount of epochs during training) has been performed. Comparative analysis of received neural networking forecasting models with each other and regression models has been performed. Conditions, affecting on accuracy and stability of results of the neural networks, have been revealed.

中文翻译：

---

时间序列预测的神经网络和回归模型的比较分析

摘要

已经考虑并研究了神经网络在时间序列预测中的适用性。为此，已经进行了在各个时间序列上的全连接和循环神经网络的训练，并初步选择了最佳超参数（优化算法，隐藏层上的神经元数量，训练期间的时期）。已经对接收到的神经网络预测模型与回归模型进行了比较分析。已经揭示了影响神经网络结果的准确性和稳定性的条件。