Analyzing Factors Influencing Diabetes and Predicting Its Occurrence Using Logistic Regression

Résumé

The current work seeks to examine the determinants affecting diabetes and to create a predictive
model grounded in those determinants utilizing Logistic Regression. The dataset utilized for this study is
obtained from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and comprises
data on 768 female patients. The main goal is to find the most important factors that lead to diabetes and to
see how effectively the model can tell the difference between those who are diabetic and those who are not.
Logistic regression, a prevalent statistical technique for binary classification, is utilized to estimate the chance
of diabetes occurrence based on many independent variables, including glucose level, blood pressure, body
mass index (BMI), insulin level, skin thickness, and age. The Hosmer and Lemeshow goodness-of-fit test is
used to check how well the model works and how accurate it is. The results show that the logistic regression
model fits the data well and may be used to make predictions. The investigation revealed that skin thickness,
insulin levels, and age are negligible predictors, whereas glucose levels, body mass index (BMI), and blood
pressure significantly influence diabetes prediction. This discovery is consistent with medical knowledge that
obesity and high glucose levels are significant risk factors for diabetes. While other studies have investigated
machine learning techniques for diabetes prediction, this work underscores a statistical modeling approach
utilizing logistic regression. The model not only predicts diabetes risk but also helps us understand the main
elements that affect it in the examined population.