An Neural Collaborative Filtering (NCF) based Recommender System for Personalized Rehabilitation Exercises

Rehabilitation exercises should be prescribed and performed by carefully selecting the type of exercise and intensity for each patient. If rehabilitation exercises are prescribed incorrectly, they may be ineffective or cause dangerous situations such as secondary injuries. In this study, we developed a personalized rehabilitation exercise recommender system based on data from a clinical trial of rehabilitation exercise intervention conducted at Chungnam National University Sejong Hospital. We generated a training dataset from the clinical trial data of rehabilitation exercise intervention for patients with shoulder adhesive capsulitis, rotator cuff injury, and back pain, and developed a rehabilitation exercise recommender system based on NCF's NeuMF model to verify its performance. The results show that the developed rehabilitation exercise recommender model outperforms the methods based on GMF and MLP models only. The personalized rehabilitation exercise recommender system in this study is expected to help rehabilitation patients recover their functions, improve their health, and return to society quickly and safely.