Enhancing model transparency: Integrating local interpretable model agnostic explanations and SHapley additive exPlanations for explainable artificial intelligence in Juvenile onset diabetes prediction
P. Urjitha, N. R. Shamanth Showri, Sathvik V. Koushik, C.R. Shreya, C.D. Divya · 2025
Ensuring transparency remains pressing issue when integrating machine learning models, especially in healthcare settings where trust and comprehension are paramount. This study introduces an innovative strategy to tackle this challenge by merging two state-of-the-art interpretability techniques: Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive 1 exPlanations (SHAP). Our approach aims to harness the strengths of both methods to offer more effective and holistic explanations of opaque models. Through extensive testing, including a case study on Type 1 Diabetes diagnosis, we illustrate the efficacy of our combined approach in improving model transparency and interpretability. By bridging the divide between global and local interpretability, our method facilitates informed decision-making and cultivates confidence in ML applications, particularly in critical areas like healthcare.