Empowering Decision Making in Healthcare - A Comparative Analysis of eXplainable AI Techniques Using SHAP and LIME for Cancer Patients Data

The present study undertook a comparative analysis of eXplainable AI techniques using SHAP and LIME for cancer patient data. The purpose of the present study is to highlight the lack of trustworthiness on machine learning algorithms. It also focused on the effectiveness of using an advanced technique as eXplainable Artificial Intelligence (XAI) to empower decision making in health care. The present study attempted to understand the predictive models when applied to cancer patient data from the National Cancer Institute's. The methodology used was SHAP (SHapley Additive Explanations) and LIME (Local Interpretable Model-Agnostic Explanations) to shed light on machine learning model. Initially, four machine algorithms were trained to get the best predictive model. The performance metrics were studied on different algorithm and accordingly all the result were compared. The Gini Importance Plot and Permutation Importance Plot were applied to assess how much each feature reduces impurity (like Gini impurity) across all trees. Further to get a deeper insights XAI technique SHAP and Lime along with a case study on a patient was performed through prediction formula that gave an accurate and trustworthy result. Partial Dependence Plot (PDP) was used to show the overall relationship between a feature and the predicted outcome, averaged over the dataset. The finding of the study showed Bagging algorithm performed the best in all aspect. It can be concluded that using eXplainable AI techniques makes the result more reliable and trustworthy. It was interpreted that the present study gave a promising AI-powered healthcare solutions for better and easier understanding of machine learning algorithm in health care domain. Based on the findings and result of the present study it can be suggested that eXplainable AI can be further used in other domains like finance, cyber security and law enforcement.