Use of Logistic Regression Method in Predicting Obesity Levels with Machine Learning Method

Abstract

Obesity is a worldwide health issue due to excessive fat accumulation, especially prevalent in developing countries. It increases risks for diabetes, heart disease, and cancer, affecting multiple body systems. In 2016, 1.9 billion people were overweight, with 650 million classified as obese, emphasizing its global impact on public health. Both rich and developing nations are seeing sharp increases in their obesity rates, while low- and middle-income nations are seeing the biggest increases. This emphasizes how critical it is to create international plans for the administration and avoidance of obesity. This thorough analysis demonstrates the substantial effects of obesity on public health, health systems, and individual health. Public health policy are thus greatly influenced by studies on the causes, effects, and practical management techniques of obesity. The aim of this study is to derive classification metrics for a machine learning(ML) model suitable for classifying obesity levels of individuals and to present the corresponding accurate classification performance metric. Using the logistic regression model, the following classification performance metrics for predicting obesity levels were calculated: Area under ROC curve (AUC) is 0.980, Classification accuracy (CA) is 0.909, F1-Score is 0.911, Precision (Prec) is 0.909, Recall is 0.860, Matthews correlation coefficient (MCC) is 0.992, and Specificity (Spec) is 0.992. Notably, the classification accuracy (CA) of 90.9% indicates a significant achievement in correctly classifying the levels of obesity.This evaluation demonstrates the efficacy of the logistic regression model in distinguishing between different obesity levels, with high values in various performance metrics such as AUC and MCC underscoring the model's robustness and reliability in medical.