TY  - JOUR
T1  - Personalized Exercise Recommender Systems for Rehabilitation Using Graph Neural Networks
AU  - Kim, Sooyon 
AU  - Oh, Jeong-Heon 
AU  - Oh, Dagun 
AU  - Seo, Changwon 
AU  - Shin, Won-Yong 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2022
DA  - 2022/1/15
DO  - 10.7840/kics.2022.47.4.644
KW  - 재활 서비스
KW  - 운동 프로그램
KW  - 그래프 신경망
KW  - 추천 시스템
KW  - Deep learning
KW  - Recommender system
KW  - Graph neural network
KW  - Exercise program
KW  - Rehabilitation service
KW  - 딥러닝
AB  - Recently, in the field of rehabilitation treatment, applications of deep learning, artificial intelligence, and big data technology to smart rehabilitation treatment have been actively studied. In this paper, we propose a personalized exercise recommender system for rehabilitation suitable for each patient’s individual characteristics using a graph neural network (GNN), which captures the graph structure more effectively. Since it is impossible to use actual patient data due to the legal issue for data collection, this paper synthetically generates attribute data of knee disease patients, selects various rehabilitation programs, and generates empirical patient-exercise connectivity data using cross-validation. The performance of proposed model is evaluated in terms of precision, recall, nDCG in the top-N recommendation scenario, recommending the top-N items among all the predicted preferences for each user. When the generated dataset is used, it is empirically demonstrated that the proposed method outperforms the conventional recommendation algorithm in terms of all performance metrics.