Comparative Analysis of Predicting Hospitalization Time for Diabetes Patients Using Gradient Boosting and Random Forest Algorithms

Accurate prediction of hospitalization time for diabetes patients is essential for effective healthcare management and resource allocation. This study employs and compares two advanced machine learning algorithms, Gradient Boosting and Random Forest, comparing the performance after converting categorical variables to numerical values to predict the hospitalization duration using the UCI Diabetes dataset. Both methods were implemented to develop predictive models, and these models were evaluated based on accuracy and predictive power, providing insights into the most effective approach for predicting hospitalization outcomes. For the model evaluations, the common evaluation metrics method will be used. Random Forest achieves a MSE of 5.45, MAE of 1.76, RMSE of 2.33 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.39, whereas Gradient Boosting achieves a MSE of 5.51, MAE of 1.76, RMSE of 2.35 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.38, using the UCI diabetes dataset that include 101766 data and 50 features. The results demonstrate the comparative strengths and limitations of each algorithm, providing insights into their respective predictive capabilities. This analysis aims to inform the selection of the most suitable model for clinical applications, ultimately contributing to enhanced diabetes patient care and optimized hospital resource management.