On-demand classes are currently in use, but they have the disadvantage of making it difficult to concentrate on the class, and there is a need for concentration assessment. The purpose of this study is to propose a method for assessing concentration during on-demand classes using physiological information that can be easily obtained on a daily basis. In the experiment, the participants took on-demand classes in a concentrated or non-concentrated state according to instructions. A dataset was created using blood volume pulse (BVP), skin conductance (SC), and pupil diameter data collected during class participation. Machine learning was used to classify concentration and non-concentration states and evaluate accuracy. The results showed that the model using all participants' data had an accuracy of 0.5, while the model using individual participants' data had an average accuracy of 0.6. This suggests that models using individual participants' data are more suitable for evaluating concentration levels.
ABE et al. (Wed,) studied this question.