ニューラルネットワークを利用した観点別評価における学校数学問題分類システムの開発

抄録

We have developed the system that classifies the problems in school mathematics for criterion-referenced evaluation into each viewpoint of learning situation by using neural network. We classified the school mathematics problems in four viewpoints for the criterion-referenced evaluation by using this system. The following matters were clarified; (1) As for the specialists of mathematics education, the average of the identification rate in this system is 84.6% and the system is highly accurate. (2) As for the novice of mathematics education, the average of the identification rate when using this system is higher 11.4% than not using this system. (3) The average of the identification rate using this system is 36.4% higher than using a multiple discriminant analysis. As a result, by using this system, the teacher was able to classify the problems of school mathematics into four viewpoints of the evaluation more accurately.