Exploring Explainable Machine Learning in Healthcare: Closing the Predictive Accuracy and Clinical Interpretability Gap

Machine learning has emerged as a powerful tool for healthcare, aiding in patient outcome predictions and clinical decision-making (Bhardwaj et al. (2017 IEEE 41st annual computer software and applications conference (COMPSAC). IEEE, pp. 236–24 2017)). However, the opacity of many models has hindered their clinical adoption, where interpretability and transparency are vital. This study explores “Explainable Machine Learning for Healthcare,” aiming to bridge the gap between clinical interpretability and prediction accuracy. It underscores the importance of interpretability in healthcare. In this study, we have developed two models (Naive Bayes and Logistic Regression) to accurately classify 24 diseases based on patient descriptions written in everyday English, with a 0.99 accuracy rate. A LIME explainer is attached to each model to explain the predictions they made. This study utilizes the “Symptom2Disease” dataset, sourced from Kaggle, containing 1200 data points, with 50 data points allocated to each disease. The study also suggests future research priorities, such as developing innovative algorithms and visualization methods that balance interpretability and prediction accuracy in healthcare. In conclusion, this research advocates for models that excel in both predictive accuracy and providing healthcare professionals with transparency for informed clinical decision-making.