물질의 상태 분류에 대한 과학교사와 머신러닝 모델의 분류 결과의 비교 분석

Abstract

Objectives This study compares and analyzes state classification results of matter between developed machine learning model and research participants who are elementary and middle school science teachers, identifies the cause of the situation in which classification results are inconsistent. And based on the results, we tried to find educational implications that can help learn state classification of matter.
 Methods For 31 elementary and middle school science teachers enrolled in the Graduate School of Education at the College of Education located in the central region, matter classification activities were performed and a decision tree algorithm was applied to the machine learning model. And the effectiveness of the program was confirmed through model performance evaluation such as accuracy, F1-score, precision, recall.
 Results The classification accuracy of developed machine learning model for classifying state of matter was 0.820, the F1-score was 0.820, the precision was 0.826 and the recall was 0.820. In addition, the degree of discrepancy between the classification results of science teachers and the classification results of the decision tree algorithm was larger in heterogeneous mixtures than in pure matters or homogeneous mixtures. This discrepancy was analyzed as a phenomenon that occurs because science teachers do not consistently apply the classification criteria from the macroscopic and microscopic perspectives or do have the concept that a specific matter is a specific state in advance when classifying the state of matter. Based on the results of these studies, the cause of confusion revealed in the process of classifying the state of matter pointed out in previous studies was found.
 Conclusions Based on the research results, it was possible to find the cause of confusion revealed in the process of classifying the state of matter of students pointed out in previous studies, and since machine learning can be an effective tool for diagnosing learning conditions, it is suggested that teacher training is needed to utilize it.