Research on the Evaluation of Financial Performance of Agricultural Listed Enterprises Based on the ABC-LSTM Model

Evaluating the financial performance of agricultural listed companies is crucial for ensuring food security. However, conventional methods often fail to capture the nonlinear and temporal characteristics inherent in their operational data. This study proposes a hybrid model combining Factor Analysis (FA) with an Artificial Bee Colony-optimized Long Short-Term Memory network (ABC-LSTM). Using financial data from 42 agricultural listed enterprises (2021-2023), we first extract four core factors (including profitability and solvency) through FA, then optimize LSTM hyperparameters using the ABC algorithm. Empirical results demonstrate that the ABC-LSTM model achieves outstanding performance in predicting comprehensive enterprise performance scores, with a Mean Squared Error (MSE) of 0.0030 (82.46% lower than traditional LSTM) and a coefficient of determination ($\mathrm{R}^{2}$) of 0.9903. Both fitting curves and residual plots confirm the model’s high accuracy and stability, showing strong consistency between predicted and actual values for most enterprises. This approach provides a novel methodology for dynamic performance evaluation in the agricultural sector and offers valuable insights for time-series modeling across other industries.