In response to the heavy burden of metabolic dysfunction-associated steatosis liver disease (NAFLD) in China, limited existing screening methods, unsuitability of foreign models for Chinese people, small sample size of domestic models, and insufficient predictive performance, the present work constructs a high-precision NAFLD progression risk assessment model for employed laborers and farmers. Firstly, based on the multi-source health check-up data of 68573 employed laborers, 14 key indicators including age, gender, blood pressure, blood glucose, blood lipids, and hepatic biochemical function were selected as features to construct a CNN-LSTM hybrid model fusion predictive framework. The performance of the model was comprehensively evaluated using AUC, classification accuracy, precision, recall, F1 value, and calibration curve. The experimental results show that the proposed CNN-LSTM hybrid model intelligent learning algorithm based NAFLD progression risk assessment model has an AUC index value of 0.9861 on the training dataset and 0.9336 on the validation dataset, with an classification accuracy index value of 0.8505, an classification accuracy index value of 0.8363, and an F1 index value of 0.8570. All indicators are superior to the comparison model, indicating that the proposed CNN-LSTM hybrid model has high classification accuracy, strong generalization and stability in predicting NAFLD progression risk, and can provide technical support for clinical precision intervention and population health management prevention and control.
