Enhancing Early Diabetes Risk Prediction: Optimization and Application of Logistic Regression Models

Abstract

Diabetes has become a significant challenge in global public health, with its incidence steadily rising and increasingly affecting younger populations. Statistics indicate that over 422 million people worldwide are currently living with diabetes. This chronic metabolic disease severely impacts patients’ quality of life and imposes substantial economic pressure on healthcare systems. Therefore, timely identification of high-risk groups and the implementation of effective preventive strategies are crucial. However, the multifactorial nature of diabetes presents limitations in existing prediction methods. This study utilizes logistic regression analysis to highlight the key roles of factors such as gender, age, frequent urination, excessive thirst, fatigue, localized paralysis, genital fungal infections, emotional instability, and increased appetite in assessing diabetes risk. The model achieved prediction accuracies close to 93% in both the training and test sets. Additionally, the p-values for frequent urination, localized movement disorders, excessive drinking, mood fluctuations, weakness, genital fungal infections, and increased appetite were significantly below 0.01, strongly indicating a close association between these characteristics and the risk of developing diabetes. The findings of this study lay a critical foundation for the early warning system of diabetes and emphasize the need for future research to incorporate richer clinical data and innovative technologies to optimize prediction models.