基于自回归神经网络的多维时间序列分析

Multidimensional Time Series Analysis Based on Autoregressive Neural Network

针对多维时间序列分析传统方法多数需要依靠手动建立时间依赖关系探索历史数据中隐含规律的问题,提出一种自回归神经网络方法.首先,通过卷积神经网络(CNN)和双向长短期记忆网络(LSTM)构成神经网络分别捕获多维输入特征和时间序列中存在的复杂依赖关系,并结合传统的自回归方法对线性关系进行特征提取;其次,在不同领域的两个数据集上与多个经典模型进行对比实验,结果表明,该模型预测性能最优,并能成功捕获数据中存在的重复模式;最后,用消融实验验证了该模型框架的高效性和稳定性.

Aiming at the problem that most traditional methods for multidimensional time series analysis relied on manually establishing temporal dependencies to explore the implicit rules in historical data, we proposed an autoregressive neural network method. Firstly, the neural network composed of convolution neural network(CNN) and bidirectional long short-term memory(LSTM) was used to capture the complex dependencies existing in multidimensional input features and time series, and the linear relationship was extracted by combining the traditional autoregressive method. Secondly, compared with several classical models on two datasets in different domains, the experimental results showed that the model had the best prediction performance and could successfully capture the repeated patterns in the data. Finally, the ablation experiments verified the efficiency and stability of the model framework.