抄録
Behavioral and psychological symptoms of dementia (BPSD) that develop in patients with dementia not only impose a heavy burden on caregivers, but also affect the quality of life of the patients themselves. If BPSD can be predicted in advance and symptoms can be dealt with, the burden on caregivers can be reduced. In a preliminary experiment, we predicted BPSD using machine learning based on environmental and vital sensor data collected from multiple nursing homes. However, the Average Precision of the PR curve is still low. In this study, we analyzed data to improve the accuracy of BPSD prediction.The data analysis confirms that certain symptoms have a 24-hour cycle. The results show the possibility of predicting the onset of BPSD using machine learning methods.