CatBoost Model for Enhanced Treatment Prediction in Type 2 Diabetes Patients

Abstract

Customized therapeutic approaches are essential for enhancing outcomes and patient satisfaction in the management of type 2 diabetes. This paper presents a predictive model utilizing CatBoost, a gradient boosting algorithm adept at managing categorical data, to offer appropriate prescription regimens for diabetes patients. This research utilizes a dataset of 51,000 anonymized patient records supplied by the Palestinian Ministry of Health, which includes extensive demographic, clinical, and pharmaceutical information. Our methodology prioritizes rigorous preprocessing techniques, encompassing feature selection and the management of missing values, to guarantee data quality. CatBoost surpassed baseline models, such as Random Forest and Support Vector Machine, attaining a prediction accuracy of 97%. Critical factors affecting treatment efficacy comprised HbA1c levels, adherence to medication, and concomitant conditions. This study shows how effective machine learning is in offering personalized care, especially in places with limited resources, and helps improve smart decision support systems for managing chronic diseases.