INN: Experimental Evaluation of Machine Learning Based Obesity and Overweight Prediction Using Intensive Neural Networks

Obesity and overweight are significant public health concerns worldwide, contributing to a range of chronic diseases, including cardiovascular disease, diabetes, and certain cancers. This study presents an advanced predictive model for obesity and overweight detection using an Intensive Neural Network (INN) integrated with Support Vector Machine (SVM). The model is designed to address the complex, non-linear relationships between various factors contributing to obesity, such as age, gender, height, weight, BMI, physical activity levels, dietary habits, and genetic factors. A dataset consisting of 28,765 manually collected data points was used to train and validate the model. Through meticulous data preprocessing, including handling missing values, noise reduction, feature scaling, and feature selection, the data was prepared for optimal model performance. The proposed INN + SVM model achieved an outstanding accuracy of 94.76%, outperforming nine existing models, including Convolutional Neural Network (90.47%) and Deep Neural Network (89.12%). Additional metrics such as Precision (93.22%), Recall (92.88%), and F1-Score (93.05%) further validate the model's superiority. The integration of SVM with INN enhances the model's ability to classify difficult cases, resulting in a robust and reliable tool for predicting obesity and overweight conditions. The results suggest that this hybrid approach holds significant potential for improving early detection and intervention strategies in healthcare.