Research Article
Patient Classification Based on Symptoms Using Machine Learning Algorithms Supporting Hospital Admission
@INPROCEEDINGS{10.1007/978-3-030-92942-8_4, author={Khoa Dang Dang Le and Huong Hoang Luong and Hai Thanh Nguyen}, title={Patient Classification Based on Symptoms Using Machine Learning Algorithms Supporting Hospital Admission}, proceedings={Nature of Computation and Communication. 7th EAI International Conference, ICTCC 2021, Virtual Event, October 28--29, 2021, Proceedings}, proceedings_a={ICTCC}, year={2022}, month={1}, keywords={Hospital admission Bag of words Explanation Clinic symptom Disease classification}, doi={10.1007/978-3-030-92942-8_4} }- Khoa Dang Dang Le
Huong Hoang Luong
Hai Thanh Nguyen
Year: 2022
Patient Classification Based on Symptoms Using Machine Learning Algorithms Supporting Hospital Admission
ICTCC
Springer
DOI: 10.1007/978-3-030-92942-8_4
Abstract
Overcrowding in receiving patients, medical examinations, and treatment for hospital admission is common at most hospitals in Vietnam. Receiving and classifying patients is the first step in a medical facility’s medical examination and treatment process. Therefore, overcrowding at the regular admission stage has become a complex problem to solve. This work proposes a patient classification scheme representing the text to speed up the patient input flow in hospital admission. First, the Bag of words approach has been built to represent the text as a vector exhibiting the frequency of words in the text. The data used for the evaluation were collected from March 2016 to March 2021 at My Tho City Medical Center - Tien Giang - Vietnam, including 230,479 clinic symptom samples from admissions and discharge office, outpatient department, accident, and Emergency Department. Among learning approaches used in the paper, Logistic Regression reached an accuracy of 79.1% for stratifying patients into ten common diseases in Vietnam. Besides, we have deployed a model explanation technique, Locally Interpretable Model-Agnostic Explanations (LIME), to provide valuable features in disease classification tasks. The experimental results are expected to suggest and classify the patient flow automatically in the hospital admission stage and discharge office to perform the patient flow in the clinics at the hospitals.
