基于深度网络CNN-LSTM模型的中国消费者信心指数预测

China’s Consumer Confidence Index Forecast Based on Deep Network CNN-LSTM Model

消费者信心指数是衡量区域社会经济发展的一个先行指标,对其进行预测研究有助于提前把握区域发展的未来经济走势和消费趋向。对此,文章尝试利用卷积神经网络(CNN)和长短期记忆网络(LSTM)在特征提取和时序非线性方面的独特优势,对直接融合CNN和LSTM的串联CNN-LSTM模型进行了构建,并通过该模型对中国消费者信心指数进行了预测分析。研究表明,CNN-LSTM串联模型相较于随机森林、支持向量机和BP神经网络这些传统机器学习模型,以及单一LSTM、CNN、Encoder-Decoder-LSTM、deep AR和深度状态空间(DSS)等深度学习模型都具有更高的预测精度,且在月度和季度这两种不同的数据样本期中均表现出优异的预测效果。

Consumer confidence index is a leading indicator for measuring regional socio-economic development, and is helpful to predict the future economic trend and consumption trend of regional development in advance. In view of this, the paper attempts to use the unique advantages of Convolutional Neural Network(CNN) and Long and Short Term Memory Network(LSTM)in feature extraction and time sequence nonlinearity to construct a tandem CNN-LSTM model integrating CNN and LSTM directly,and also makes predictive analysis on China’s Consumer Confidence Index based on this model. The research shows that compared with traditional machine learning models such as random forest, support vector machine and BP neural network, as well as deep learning models such as single LSTM, CNN, Encoder-Decoder-LSTM, deepAR and deep state space(DSS), CNN-LSTM tandem model has higher prediction accuracy, and shows excellent forecasting effect in both monthly and quarterly data samples.