Machine Learning Performance Analysis for Classification of Medical Specialties

Abstract

Recently supervised machine learning studies have become increasingly significant due to the increasing availability of electronic documents from different sources. Text classification can be defined as a task that automatically categorizes a group of documents into one or more predefined classes according to the subject. The main purpose of text classification is to assist users in extracting information from textual sources and dealing with processes such as retrieval, classification, and machine learning techniques together to classify different patterns. Based on machine learning algorithms, the text classification system includes four processes, that is text preprocessing, text representation, classification, and evaluation. In this paper, a text classification system model for medical specialties is designed. This research aims to assist nurses in diagnosing patients more quickly through the text of the patient’s medical record so that it can be recommended to a more appropriate medical specialty. In the classification section, we separately select and compare five different methods, that is multi-layer Perceptron, Logistic Regression, Random Forest, K-Nearest Neighbor, and Support Vector Machine as our classification algorithm. The best method for 4,998 data and 40 classes is Random Forest and Multi-Layer Perceptron with an accuracy of 47.53%. Then, we tested and analyzed the classifier model to conclude. The experimental conclusion shows that the medical specialization text classification system still obtains results that are not optimal based on machine learning algorithms.