Research on Enterprise Total Assets Prediction Model Based on Deep Learning: A Comparative Study of Ensemble Machine Learning Models and LSTM

In the digital era, enterprise digital transformation is an indispensable booster for the high-quality development of enterprise. It is critical to make accurate prediction of total assets as they help in optimizing resource allocation and operational efficiency. We use the balance sheets of China's A-share listed companies from 2001 to 2022 in this study. The time series data was preprocessed to be entirely complete and accurate after which time sliced to split the data into training and testing sets. It was trained multiple models including Random Forest, XGBoost, LSTM, and CatBoost and evaluated. Bayesian hyperparameter tuning was used in order to optimize model parameters. We have also experimented with the fusion of three high performance machine learning models including Random Forest, XGBoost, and CatBoost with the ensemble performance of their combination with the LSTM model. The experiments show that LSTM model has good predictive support for behavior of asset allocation and risk management in the environment, and as a result, it has a solid theoretical foundation for making decisions. The future work should dedicate to hybrid framework that integrate deep learning and traditional model to achieve the correct manner to accommodate the precision versus computational efficiency.