Financial Profitability Evaluation and Forecasting Using the Deep Learning: A Case of the Chinese Petroleum Industry

This paper proposes to construct the GPCA-GLSTM hybrid model using deep learning to evaluate and predict the profitability of Chinese petroleum industry listed companies. Using the time-series global principal component analysis (GPCA) to assess the overall profitability of the petroleum industry com-prehensively, the grey relation analysis (GRA) is applied to determine the financial key influencing factors while the macro key influencing factors are also added as the feature variables of the long and short-term memory neural network (LSTM) model. The results show that the profitability of Chinese petroleum industry listed companies is affected by the stock market with the EPS being the most important influencing factor in profitability; the overall profitability of the petroleum industry is meager and seriously polarized. Sol-vency has the most significant impact on profitability, and the debt asset ratio has the highest correlation with the profitability composite index; the operating index and cash flow adequacy ratio have positive dynamic effects on the profitability composite index. The GPCA-GLSTM model can accurately capture dynamic dependencies in series and has a higher forecasting accuracy with RMSE=0,3978 and MAPE=1,1110 compared to univariate LSTM model, SARIMA model and support vector machine regression (SVR) model. There-fore, the proposed model has significant practical value and is an applicable method for assessing and forecasting the profitability of companies. © 2023, Economic Laboratory for Transition Research. All rights reserved.