Exploring Ensemble Learning Based Approaches for Multimodal Prediction of Insulin Dosage for Type-1 Diabetes Patients

Diabetes is an unfortunate ailment plaguing the life of 1 in every 10 adults. Diabetes is a gateway to more severe forms of ailments or health conditions. This disease is marked by the failure of human body to incorporate glucose from the bloodstream using the hormone insulin. Basal and bolus insulin dosages are calculated and prescribed to Type-1 diabetes patients but determining insulin dosage is a meticulous process since dosage values can vary from person to person throughout time and errors can cause adverse effects. Incorporating optimized ensemble machine learning techniques show acceptable results $\left(\mathbf{R}^{\mathbf{2}}=\mathbf{0. 8 6}\right)$ for basal and bolus dosages with Extreme Gradient Boosting giving best results.