Performance Analysis Of Machine Learning-Based Models For Early Diagnosis Of Obesity Using Blood Test Parameters

Authors : Sare Nur Cuhadar, Gül Karaduman, Ahmet Uyanik, Habibe Durmaz

Pages : 117-128

View : 68 | Download : 102

Publication Date : 2023-12-31

Article Type : Research

Abstract :Obesity is a global health issue that continues to grow, with projections indicating further increases in obesity rates. The World Health Organization defines overweight and obesity as the abnormal or excessive accumulation of fat, posing risks to overall health. Obesity is not only a significant condition itself but is also directly linked to various diseases such as type 2 diabetes, coronary heart disease, hypertension, and certain types of cancer. The rising prevalence of obesity presents significant health complications and risks for individuals of all ages, particularly children and adolescents. Obese or overweight children face an increased likelihood of developing severe health problems in adulthood, potentially enduring the same physical condition throughout their lives. Urgent action is necessary to mitigate this global health concern. In this study, we aimed to predict obesity risk through the use of machine learning algorithms. Our research gathered 367 data from individuals of different age groups, classified as either obese or non-obese, based on their blood test results. We employed nine machine learning algorithms, including BayesNet, Naïve Bayes, SMO, Simple Logistic, IBk, Kstar, J48, Random Forest, and Random Tree algorithms. Our analysis successfully determined the obesity status of individuals based on internal results, the Simple Logistic algorithm achieved the highest accuracy rate at 98.6395. On the other hand, the Simple Logistic and Kstar algorithm demonstrated the highest accuracy rate of 100% for the extenal set. Our model provides valuable insights for further research and interventions for analyzing the blood test values associated with obesity.
Keywords : Obesity, machine learning, feature selection, classification, model performance.