Stock Price Prediction Based on ARIMA-GM Hybrid and LSTM Model

The stock market has grown to be an essential component of China's economic market as the country's economy has developed. And stock prices are closely watched by investors and managers. This paper uses Python to crawl the monthly stock prices of Ping An of China from January 2016 to December 2024 and decides to estimate its closing price using the ARIMA, GM, and LSTM models. Furthermore, this paper mixes the ARIMA model with the GM model to build a hybrid model. The empirical results show that the ARIMA models mean square errors is 35.79, while the GM models is 125.63. The MSE loss function of the LSTM model is 14.186, and the MSE of the ARIMA-GM hybrid model is 7.575. The results of this paper show that the hybrid models prediction accuracy surpasses not only that of a single model but also that of the LSTM model in machine learning, providing a scientific and effective guide for investor and manager behavior.