A study of obesity class prediction built on neural networks.

Objective To use neural networks and optimization algorithms, establish an obesity level prediction model to assess obesity risk. Methods Perform correlation analysis on 2 111 recorded data collected from participants aged between 14 and 61 years old in Mexico, Peru, and Colombia, and establish a BP neural network obesity level prediction model. At the same time, optimize the number of hidden nodes and transfer function of the model through pruning to find the optimal network structure. In addition, the genetic algorithm and the simulated annealing algorithm were used to optimize the weights and thresholds of the model, ultimately establishing a high-precision and practical GASA-BP neural network obesity level prediction model. Results The R2 of the prediction model was 0.975 1, and the MAE was 0.352, indicating high prediction accuracy and strong practicality. In the process of predicting obesity levels in the model, weight index was the most important, with a correlation of 0.913 with obesity levels. The correlation between overweight members in the family was also relatively strong, with a correlation of 0.505. Conclusion The GASA-BP neural network prediction model performs better than other models in predicting obesity levels, and can make the most accurate prediction of obesity levels, providing guidance and reference for personalized obesity assessments and subsequent prevention and control measures. [ABSTRACT FROM AUTHOR]