Research Article
Applying the Shapley Value Method to Predict Mortality in Liver Cancer Based on Explainable AI
@INPROCEEDINGS{10.1007/978-3-031-31275-5_14, author={Lun-Ping Hung and Chong-Huai Xu and Ching-Sheng Wang and Chien-Liang Chen}, title={Applying the Shapley Value Method to Predict Mortality in Liver Cancer Based on Explainable AI}, proceedings={Smart Grid and Internet of Things. 6th EAI International Conference, SGIoT 2022, TaiChung, Taiwan, November 19-20, 2022, Proceedings}, proceedings_a={SGIOT}, year={2023}, month={5}, keywords={Machine learning Hepatocellular carcinoma Risk factors SHapley Additive exPlanations (SHAP) Extreme gradient boosting (XGBoost)}, doi={10.1007/978-3-031-31275-5_14} }- Lun-Ping Hung
Chong-Huai Xu
Ching-Sheng Wang
Chien-Liang Chen
Year: 2023
Applying the Shapley Value Method to Predict Mortality in Liver Cancer Based on Explainable AI
SGIOT
Springer
DOI: 10.1007/978-3-031-31275-5_14
Abstract
Hepatocellular carcinoma (HCC) is the sixth-leading cause of death worldwide and has the highest mortality rate among all types of cancers. In most cases, the patient has entered the terminal phase of a cancer disease when hepatocellular carcinoma occurs. Therefore, if the cause of cancer can be identified, disease deterioration can be prevented. With the rise of artificial intelligence (A.I.) technology in recent years, many scholars have used machine learning technology to predict the probability of dying from hepatocellular carcinoma and have obtained good results. However, the studies lack interpretability and do not facilitate the further analyses of medical experts. Therefore, this study proposes a deep learning model based on XGBoost and uses the data evaluation method of Shapley value to study the characteristics of machine learning and verify the results using the hepatocellular carcinoma dataset. The proposed model delivered strong prediction performance, with an accuracy of 92.68%, and accurately interpreted the dataset features, supporting analyses by medical experts.
