A knowledge graph-based recommender system for dementia care: Design and evaluation study.

ABSTRACT

BACKGROUND: Caring for people with dementia is perceived as one of the most challenging caring roles, so effective and practical support is essential. One such innovative approach to internet-based tailored health intervention is the use of recommender system. OBJECTIVE: This study develops a dementia care intelligent recommender system (DCIRS) that can provide personalized and timely care recommendations for caregivers when they encounter difficult various care problems in people with dementia. METHODS: The development process was divided into 3 stages. In stage 1, we complete the construction of the domain knowledge graph of dementia care. In stage 2, the established domain knowledge graph of dementia care was introduced into the recommendation model by the way of graph embedding to form a recommendation model composed of graph embedding module and recommendation module. In stage 3, on the basis of the application of knowledge graph and recommendation mode, DCIRS was developed, for practical use. In addition, DCIRS has been validated for accuracy for assessing the correctness of the profiling and recommendation, by enrolling 56 caregivers. RESULTS: The proposed DCIRS has functions of knowledge graph management and dementia care decision support. Experiments on 56 caregivers in single class recommendation task; the value of accuracy is equals to 98.92% and indicates the high capability of DCIRS. CONCLUSIONS: This study was a pioneering research to develop a more comprehensive DCIRS for caregivers of people with dementia. According to the evaluation results, our DCIRS showing high specificity and accuracy. This system can provide a novel perspective for patient-centered and needs-based support of caregivers of people with dementia.