[2E1-GS-13] AI application: Medical application (2)

[2E1-GS-13-04] Evaluation of Rule-based Classifiers that Predict Low Blood Pressure Caused by Polypharmacy

〇Takeshi Ikeda¹, Yoshitaka Kameya², Tomohiro Mizuno³, Fumihiro Mizokami⁴ (1. Graduate School of Science and Technology, Meijo University, 2. Faculty of Science and Technology, Meijo University, 3. Faculty of Pharmacy, Meijo University, 4. Department of Pharmacy, The National Center for Geriatrics and Gerontology)

Keywords:Polypharmacy, Rule-based Classifier, Discriminative Pattern Mining

In recent years, problems caused by polypharmacy have been paid attention in medical care for elderly people. In this paper, we attempt to build and evaluate interpretable, rule-based classifiers that predict low blood pressure of an elderly patient from the patient's history of drug administration. These rule-based classifiers were built based on discriminative patterns extracted from a class-imbalanced dataset comprised of symbolic, numeric, and missing values, and we evaluated them from both predictive and clinical perspectives, in comparison with XGBoost, C4.5, and RIPPER.