(英) |
Self-efficacy is the degree of confidence in one's ability to perform a behavior, and has been attempted to apply to predicting behavior. In this paper,
for aiming to measure this self-efficacy from anticipatory gaze, gaze data while performing two types of tasks (puzzle task and maze task) using a mouse was collected. Four gaze features (saccade distance, percentage of time which looking 100 px around the cursor, the transition time difference to Area of Interest (AOI) between the fixation point and the object, and angle between the gaze point and the cursor vector when clicked) were proposed to be used as anticipatory gaze, and were compared them with subjective self-efficacy. As a result, the best correlation coefficient value was 0.57. In order to improve the predictive performance of self-efficacy, linear regression, support vector regression, and random forest regression models were introduced. As a result, for the puzzle task, a coefficient of determination of 0.34 was obtained by random forest regression, and for the maze task, a coefficient of determination of 0.465 was obtained by support vector regression. |