METHOD OF PREDICTIVE MODELING OF DRY EYE DISEASE BASED ON THE ANALYSIS OF BEHAVIORAL AND PHYSIOLOGICAL FACTORS

Abstract

This article presents a method for predictive modeling of dry eye disease based on the analysis of physiological and behavioral factors using machine learning techniques. The growing prevalence of dry eye syndrome, especially among individuals with prolonged screen time, underscores the importance of developing effective early diagnostic tools. A review of current research indicates that conventional diagnostic methods often rely heavily on clinical indicators, overlooking the role of behavioral and lifestyle-related risk factors, which are essential for comprehensive risk assessment. 
The proposed method leverages the XGBoost algorithm and includes several stages: data preprocessing, feature selection and engineering, risk prediction, and generation of personalized preventive recommendations. An object-oriented implementation ensures modularity, scalability, and maintainability of the system. Data normalization, handling of missing values, encoding of categorical variables, and the use of clinically interpretable features are integrated into the system’s architecture. 
The model was trained and tested on a dataset comprising more than 20,000 records, and its performance was evaluated using five-fold cross-validation. The achieved accuracy was 89%, with a precision of 87%, recall of 86%, and F1-score of 87%, confirming the reliability of the approach. An analysis of feature importance revealed that screen time, blinking frequency, age, sleep quality, and stress level were the most influential risk factors. The model also demonstrated a clear separation between affected and healthy individuals based on predicted probabilities, supporting its diagnostic value.
The proposed system can be effectively used in ophthalmic practice for early identification of individuals at high risk of dry eye disease. By integrating behavioral insights with physiological data, it supports more personalized and preventive healthcare approaches, ultimately improving the quality of patient care and reducing the burden of this increasingly common condition.