In this study, we evaluated the classification model for people in a care facility using integrated data from a robot and the environment. Focusing on the posture, walking distance and walking speed as indicators for classification of caregivers and elderly people, we prototyped the system which collected these indicators from sensors installed in a robot and the environment. In measuring the activities of caregivers and elderly people in a care facility by using the system, we found differences in posture angle and walking speed between them. Therefore, we prototyped four classification models between caregivers and elderly people based on these indicators. The two models are individual models trained on each indicator obtained by the sensor of the robot or environment, respectively. Another model is an ensemble trained model by combining these individual models. The other model is an integrated model trained on dataset which integrated posture angle and walking speed. As the evaluation of these models, we used the area under the Receiver Operating Characteristic (ROC) curve (AUC). The results showed that the ensembled trained model had the highest classification performance with an AUC of 0.94. This suggests the possibility of creating a high classification performance model through ensemble learning with individual models trained on each indicator according to the situation.
MURANO et al. (Wed,) studied this question.