Using Machine Learning to Explore the Crucial Factors of Assistive Technology Assessments: Cases of Wheelchairs.

Abstract

The global population is gradually entering an aging society; chronic diseases and functional disabilities have increased, thereby increasing the number of people with limitations. Therefore, the demand for assistive devices has increased substantially. Due to numerous and complex types of assistive devices, an assessment by a professional therapist is required to help the individual find a suitable assistive device. According to actual site data, the assessment needs of "wheelchairs" accounted for most of the cases. Therefore, this study identified five key evaluation characteristics (head condition, age, pelvic condition, cognitive ability, and judgment) for "transit wheelchairs" and "reclining and tilting wheelchairs" from the diagnostic records of "wheelchairs" using the classification and regression trees (CART) decision tree algorithm. Furthermore, the study established an evaluation model through the Naïve Bayes classification method and obtained an accuracy rate of 72.0% after a 10-fold cross-validation. Finally, the study considered users' convenience and combined it with a LINE BOT to allow the user or the user's family to engage in self-evaluation. Preliminary suggestions for wheelchair types were given through the assessment model so that evaluators could not only determine a case's situation in advance and reduce the time required for fixed-point or home assessments, but also help cases find the appropriate wheelchair type more easily and quickly.