Human health status data is characterized by high dimensionality, complex indicators, and interactive relationships, and the traditional single prediction model faces the problems of insufficient accuracy and robustness. In this paper, we propose an intelligent management method of human health status based on Stacking integrated learning, which constructs a feature system from five dimensions: physical health, mental health, lifestyle and behavioral health, social adaptation and environmental health, and disease prevention and health management. The study firstly screened 20 key feature variables by Lasso regression and stepwise regression, and then designed a dynamic weight estimation algorithm based on the Breiman method, combining long-term historical data and short-term neighboring data to optimize the weight configuration. The experimental results show that compared with a single model, the proposed Stacking integration model performs well in a number of metrics such as AUC, accuracy, and F1-Score, with an AUC value of 0.97287 and an accuracy of 0.93149. Through 10-fold crossvalidation for 10 consecutive tests, the model demonstrates less volatility than the individual base classifiers, which verifies that the method is highly stable in the high stability in human health status prediction. The Stacking model in integrated learning significantly improves the accuracy and generalization ability of the prediction results by effectively integrating the advantages of various learners, provides reliable technical support for the intelligent assessment and management of health conditions, and is of great practical value in promoting personalized health management and precision medicine.
