Obtaining Elderly Patients’ Lifestyle Information from Unstructured Text Sources

Abstract

In this work, we made many simulations to take information from free-text notes belonging to the patient that indicates his or her habits. Detailed information about a patient's habits will allow the nurse to create a personalized daily schedule. Due to the fact that each patient has a different routine. The information is separated into five categories: dietary habits, drinking/smoking habits, excretion habits/toilet style, fashionable/colour preference/footwear, and favourite music/radio/TV shows. To realize this, we use six machine learning models, there are Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbours (KNN), Multinomial Logistic Regression (MLR), and Gradient Boosting (GB). As a result, all models have more than 90% accuracy on both train sets and test sets.In this study, we focus on SVM with Train Set Accuracy score 93.9% and Test Set Accuracy score 94.1%, which is more resistant to overfitting issues.