Explainable AI for Cancer Prediction
Aswin Kumar Govindan, Sukhpal Singh Gill · 2024
Cancer has quickly risen to become one of the world&s;s most serious health problems and the leading cause of death. Artificial intelligence (AI)-based solutions have been created recently to aid clinicians in making decisions that will lower morbidity and death. This contemporary research does, however, frequently suffer from explanations of its results. In other words, the end user is not informed of the internal reasoning behind the forecasts. These models cannot be used in clinical practice because of their black-box nature, which makes them challenging for doctors to understand. To encounter the current problem explainable artificial intelligence (XAI) can be implemented to check the trustworthiness of the model by interpreting the model to obtain the process of how the model decides to confirm a subject&s;s cancer. DeepSHAP is a technique that calculates the SHAP values for machine learning or deep learning models to increase model interpretability by assigning importance scores to the features processed in the model. Experimental results demonstrate that the XGBoost classifier has the highest accuracy rate of 98.58%. The result obtained from the summary plot using DeepSHAP indicates which feature interprets the model to decide if the patient has cancer or not. The suggested model can also advance medical understanding of cancer detection by offering broad insights into how disease risk changes both globally and locally.