Evaluating the role of agri-innovators in enhancing ecological conservation and economic efficiency using deep learning

Under the dual carbon goals, it is crucial to rationally evaluate the performance of agricultural entrepreneurs’ strategies for ecological protection and economic development. Therefore, this paper addresses the static and lagging issues of traditional evaluation methods by constructing a coupled agricultural production ecological and performance evaluation system based on deep learning. In this study, we primarily collected data on five dimensions, including ecological production, economic policy, and technology, and used a random forest method for anomaly detection. For feature extraction, we used a CN network with a fused attention mechanism and a bidirectional, long-term, and short-term neural network to construct a coupled prediction model, enabling forecasts of ecological and economic performance indicators for the next 1–5 years. To enhance the model’s practicality, we trained a classification model to classify the feasibility of agricultural entrepreneurs’ strategies. We also conducted interpretable analysis experiments. The results show that the proposed model achieves a mean squared error of 0.063 for economic and 0.071 for ecological performance indicators. The strategy feasibility classification model achieved a classification accuracy of 85.6% (95% CI: [83.2%, 87.9%]) on the test set, with an F1 score of 0.84 and an AUC of 0.92, indicating that the model is robust in balancing precision and recall.This proposed model holds significant promise in helping agricultural entrepreneurs optimize agricultural ecosystems, economic performance, and ecological performance.