Nonlinear Method for Stock Market Trend Prediction Based on Deep Learning and ARIAM

Abstract

The stock market is often seen as a barometer of a country’s economic situation. The overall performance of the market can reflect investors’ confidence in the economic outlook. The volatility of stock returns has gradually become the most concerned issue for many institutional and retail investors. Based on past research, traditional models in econometrics are not capable of predicting stock prices over the long term. The ARIMA model cannot describe nonlinearity and cannot achieve satisfactory results. Recently with the development of artificial intelligence technology, the importance of deep learning in the field of computer science and other areas has become increasingly prominent. Considering the strong nonlinear generalization ability of neural networks, we innovatively proposes a hybrid model DL-ARIAM based on attention CNN-LSTM and XGBoost to predict stock market trend, which integrates time series models, convolutional neural networks with attention mechanisms, long short-term memory networks, and XGBoost regressors in a nonlinear relationship to improve prediction accuracy. The results indicate that the DL-ARIAM model owns good performance and high prediction accuracy, which can help investors or institutions make decisions.