설명가능 인공지능을 활용한 라이프로그 기반 치매 위험도 산정 방법에 관한 연구

Abstract

This paper presents a novel method for accurately assessing dementia risk by utilizing explainable artificial intelligence (eXplainable AI) and lifelog data, which plays a crucial role in the early diagnosis of dementia. This study focuses on calculating dementia risk scores that not only derive the accuracy of the predictive model but also interpret the model's prediction results using SHAP values. This enables us to provide patients with clearer and more specific follow-up plans. The significant contribution of this study is that, based on the calculated scores, individuals with a high risk of dementia are promptly guided to undergo cognitive screening tests (CIST), allowing dementia treatment to commence at the optimal stage. By individually explaining the impact of each feature on the prediction results, SHAP values assist medical professionals in better understanding and utilizing the model's predictions.