Understanding AI: Interpretability and Transparency in Machine Learning Models
Ramesh Dugyala, Sandeep Kumar Singh, Mohammed Saleh Al Ansari, C. Gunasundari, Kilaru Aswini, G. Sandhya · 2023
Artificial intelligence (AI) is recognized as a valuable tool in various healthcare uses for diagnosing and therapeutic decision-making. Due to the tremendous rise in accessible data and processing capacity, machine learning (ML) models have performed well or greater than doctors in numerous activities. The AI platform needs to be transparent, resilient, and interpretable to adhere to the principles of trusted AI. Present ML systems are alluded to as black boxes because of the absence of understanding of the mechanics related to the decision-making procedures. As a result, before ML can be implemented into ordinary healthcare processes, its transparency and interpretability must be understood. To address this issue, this study presents a method for understanding the transparency and interpretability of ML suggestion systems. We particularly adapt the suggested technique to a chronic condition that is frequent in seniors: heart disease. The suggested approach illustrates the fundamental cause for these suggestions and increases patient trust and interpretability of ML models by assessing the influence of various patient features on the suggestions.