Research on the Application of Long and Short-term Memory Network Algorithm in Corporate Financial Forecasting Driven by Big Data

Abstract

In the current context of rapid development of digital economy, corporate financial risks occur frequently, and traditional prediction means are difficult to cope with the complex data structure. In order to improve the accuracy of corporate financial risk prediction, this paper proposes a combined model based on ARIMA-IPOALSTM. The study firstly constructs a dataset of A-share listed companies in Shanghai and Shenzhen from 2012 to 2024, and screens 18 financial indicators for modeling. In the methodology, ARIMA is combined to extract trend features, and Improved Pelican Optimization Algorithm (IPOA) is used for LSTM parameter optimization to achieve model accuracy. The experimental results show that the accuracy of the training samples reaches 94.02%, the accuracy of the test samples is 93.48%, and the RMSE is as low as 0.574 × 10−⁵. The financial data of 15 listed companies in 2024 are further selected for simulation, and the model warning accuracy reaches up to 100%. The conclusion shows that ARIMA-IPOA-LSTM has strong adaptability under dealing with unbalanced data, can effectively improve the recall rate and F1 score, and has high practical value.