Machine Learning‐assisted Interpretation of Scan‐Rate‐dependent Glucose Sensing on Nickel Cobalt Oxide/Porous Carbon Modified Screen‐printed Electrodes

ABSTRACT

Artificial intelligence (AI) is being increasingly applied in diabetes care, particularly in glucose monitoring and management. However, the reliability of AI‐driven glucose sensing is often limited by sensor inaccuracy due to the plethora of electrocatalytic materials available. In this regard, the present work explores the glucose‐sensing capability of nickel cobalt oxide/porous carbon hybrid electrode by integrating experimentally acquired cyclic voltammetry (CV) data with machine learning (ML). Key electrochemical descriptors, such as oxidation current, oxidation potential and the area under the oxidation curve, served as inputs for the predictive models, which were evaluated through a blind concentration interpolation strategy. The results reveal that predictive accuracy is governed not by algorithmic complexity, but rather by the scan rate and its associated electrochemical behaviour. Simple models such as linear regression and ridge regression exhibited strong agreement with the underlying electrochemical dynamics, highlighting the feasibility of combining regression‐based ML with electrochemical sensing. Rather than treating AI as a black box, this integrated approach offers a transparent and structured way to interpret CV data, uncover stable predictive relationships, and advance the reliability of glucose detection systems.