Selection of the optimal machine learning technique for the development of a personalized insulin infusion algorithm to control “hybrid closed-loop” artificial pancreas

Abstract

Background and Aims. Training a personalized control algorithm is the key component for an artificial pancreas (AP) solution. Most of documented applications of machine learning are for classification algorithms not for AP control. In this article, it is explained and proved that machine learning (ML) is a valid technology to produce an accurate regression control algorithm as lowcost solution to control a hybrid closed loop AP system.