Predicting the Rehospitalization Times of Diabetic Patients Using Machine Learning Models

The prediction of rehospitalization times for diabetic patients holds significant potential in enhancing healthcare management and reducing associated costs. In this study, machine learning models are applied to predict the rehospitalization times of diabetic patients, considering a variety of clinical, physical, and behavioral factors. The dataset comprises electronic health records of diabetic patients, including features such as race, gender, sex, medication history, previous hospitalizations, and laboratory results. Various machine learning algorithms, including Random Forest, XGBoost, Decision Tree, Artificial Neural Network, Deep Neural Network and K-NN are evaluated in terms of prediction accuracy, interpretability, and robustness. Feature engineering techniques, such as label encoding, feature selections, and handling of imbalanced classes, are also explored to optimize model performance. The results demonstrate that the proposed models can predict rehospitalization times with high accuracy, highlighting the most influential factors contributing to readmissions. The findings suggest that predictive models can serve as effective tools for early intervention strategies, improving patient outcomes, and enabling more proactive healthcare management for diabetic patients.