Explainable AI for Healthcare: A Study for Interpreting Diabetes Prediction

Abstract

With extensive use of machine learning in field of healthcare, explainable AI becomes a vital part of machine learning process by helping healthcare practitioners for understanding critical decision-making process. Machine learning has greatly impacted medical field by identification of factors responsible for vulnerable health risks and factors affecting them. Hence, interpretation of machine learning model is extremely critical in the healthcare sector. Healthcare practitioners must understand factors that affect decision-making process for diseases like diabetes made by machine learning model. Explainable artificial intelligence helps clinical practitioners in interpretation of black-box models and their decision-making process to verify why a particular decision was taken by machine learning model that is extremely important in medical field. Interpretability in machine learning methods can help healthcare practitioners to understand the features that affect the decision-making process for detection of diseases like diabetes. In the past few years, researchers have been successful in detecting diabetes using machine learning models. Explainable artificial intelligence is exceptionally useful in turning black-box into glass box machine learning model and explains their respective decision-making process in healthcare. The paper presents various interpretable machine learning methods for understanding factors affecting decision-making in case of diabetes prediction that could be explained using model agnostic methods.