Identification of best classifier method for Aphasia classification

Abstract

Aphasia is a disorder, causing someone to lose communication abilities which is resulted from brain damage on the part for linguistic abilities. Speech therapy is the primary treatment for individuals with aphasia, but access to therapy can be limited by a number of factors, including distance, cost, and availability of healthcare professionals. Telerehabilitation has emerged as an approach to delivering speech therapy services to individuals with aphasia, overcoming barriers associated with in-person therapy. Speech therapy telerehabilitation based on the Western Aphasia Battery (WAB) was designed and developed as mobile applications system in this study. The mobile apps tests four of the user communication skills, which are the ability to read, write, understanding the communication, and naming things, as advised by the speech therapy specialist. This study focused on the usage of developed mobile application for Malay- speaking people with aphasia and classifying aphasia using various supervised machine learning to identify best classifier methods for it. Five healthy participants and twenty aphasia patients participated in this study. Sequential machine learning methods consisted of decision tree, k-nearest neighbor, logistic regression, sequential minimal optimization for support vector classifier and the machine learning ensemble methods which were AdaBoost, and random forest, are used for analysis of the data collected from the developed mobile apps system. The study focused on evaluating the performance of various supervised machine learning methods in classifying healthy individuals and aphasia patients based on their scores and time taken for completing the exercise and identifying the best machine learning methods. Machine learning ensemble methods are very proficient in classification between healthy individual and aphasia people.