Obesity预测的机器学习与自学方法对比研究 - 面向广义与性别分组数据

Aug 6, 2026

Upload by Patrick

Obesity is a global health problem with long-term consequences, and it requires timely identification and intervention. This study applies multiple machine learning and self-learning approaches, including Random Forest, XGBoost, Decision Tree, k-Nearest Neighbors, Support Vector Machine, Logistic/Linear Regression, Naïve Bayes, and a Multilayer Perceptron neural network, on an obesity dataset to predict obesity and support reduction. Models are assessed using recall, accuracy, F1-score, and precision, comparing performance on generalized versus gender-segregated data to guide feature selection and early detection.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

BMI and WHO obesity standards

Country-level obesity motivation and gender focus

Literature Survey

Comparison of Different Machine Learning and Self-Learning Methods for Predicting Obesity on Generalized and Gender-Segregated Data 1Khushi Joshi, 2Neha Aher, 3Soha Arora, 4Varnika Mulay, 5Prof. Renuka Suryawanshi, 6Dr. Nitin Pise, 7Dr. Vitthal Gutte 1,2,3,4,5,6,7Department of Computer Science Engineering, MIT WPU, Pune, India [1](1gunnujoshi98@gmail.com)[gunnujoshi98@gmail.com](1gunnujoshi98@gmail.com), [2](2nehaaher16@gmail.com)[nehaaher16@gmail.com](2nehaaher16@gmail.com), [3](3 sohaarora12@gmail.com)[ sohaarora12@gmail.com](3 sohaarora12@gmail.com), [4](4varni.mulay@gmail.com)[varni.mulay@gmail.com](4varni.mulay@gmail.com), [5](5 renuka.suryawanshi@mitwpu.edu.in)[ renuka.suryawanshi@mitwpu.edu.in](5 renuka.suryawanshi@mitwpu.edu.in), [6](6 nitin.pise@mitwpu.edu.in)[ nitin.pise@mitwpu.edu.in](6 nitin.pise@mitwpu.edu.in), [7](7vitthalgutte2014@gmail.com)[vitthalgutte2014@gmail.com](7vitthalgutte2014@gmail.com) Abstract— Obesity is a global health concern with long-term implications. Our research applies numerous Machine Learning models consisting of Random Forest model, XGBT(Extreme Gradient Boosting) model, Decision Tree model, k-Nearest Neighbors technique, Support Vector Machine model, Linear Regression model, Naïve Bayes classifier and a neural network named Multilayer Perceptron on an obesity dataset so that we can predict obesity and reduce it. The models are evaluated on recall, accuracy, F1-score, and precision. The findings reveal the performance of the algorithms on generalised and gender-segregated data providing insights concerning feature selection and early obesity identification. This research aims to demonstrate the comparative study of obesity prediction for gender-neutral and genderspecific datasets. Keywords-Obesity, Decision Trees, Machine Learning, Support Vector Machine, Naïve Baye’s classifier, Random Forest. Abbreviations RFT : Random Forest XGBT : XGBoost DT : Decision Tree KNN : k-Nearest Neighbors SVM : Support Vector Machine LR : Logistic Regression NB : Naïve Bayes MLP: Multilayer Perceptron I. INTRODUCTION Unhealthy eating habits accompanied by a lethargic lifestyle lead to an unhealthful weight gain. Obesity is a wellness concern that causes immoderate weight gain in a human being. It is an intricate condition driven by various factors such as physical fitness, daily food and water intake, exercise regimen, age, height, etc. Being obese can be immensely harmful to any individual's health as it could result in a reduced life span and various lifethreatening diseases like cardiovascular diseases, diabetes, respiratory issues, kidney disease, skin infections, etc. Whether a person is obese or not is decided by a numerical value i.e the person's BMI-Body Mass Index. BMI formula [5, 6]: BMI = weight (kg) / height (m2) [5, 6] . The World Health Organization (WHO) has set some standards for categorizing obese people. Anybody having a BMI of 30 or more is considered obese. The WHO has also specified a few levels of obesity shown in Table 1 [5, 6] . Table-1: Health standards specified by WHO. | Sr. No. | Category | BMI | | --- | --- | --- | | 1. | Person is underweight | < 18.5 | | 2. | Person is healthy | 18.5 – 24.9 | | 3. | Person is overweight | 25.0 – 29.9 | | 4. | Person has obesity level 1 | 30.0 – 34.9 | | 5. | Person has obesity level 2 | 35.0 – 39.9 | | 6. | Person has obesity level 3 | > 40 | Our collection of data focuses on the data from sovereign nations of North and South America namely Colombia, Mexico and Peru. Hence, considering today’s situation, we discovered that according to the Global Obesity Observatory (GOO) [1-3]: Peru carries a probability of 7.5/10, denoting that every 7th person out of 10 is obese [1] . Mexico carries a probability of 8/10, denoting that every 8th person out of 10 is obese [2] . Colombia carries a probability of 7.5/10, denoting that every 7th person out of 10 is obese [3] . Looking at these figures, we reali

FAQs

该研究使用了哪些机器学习方法进行肥胖预测？

研究采用多种模型，包括随机森林、XGBoost、决策树、k近邻、支持向量机、回归模型、朴素贝叶斯以及多层感知机神经网络。

模型评估指标有哪些？

论文使用召回率、准确率、F1-score 和精确率对各算法表现进行评估。

研究为什么要对广义数据与性别分组数据进行比较？

研究通过比较两类数据上的效果，分析特征选择差异，并为早期识别肥胖提供依据，从而支撑性别中立与性别特定的数据对比研究。

Aug 6, 2026

Upload by Patrick