You are here:

Machine-Learning-Based Automatic Difficulty Estimation of Quizzes in Question-Posing Learning
PROCEEDING

, , Kanazawa Institute of Technology, Japan

E-Learn: World Conference on E-Learning in Corporate, Government, Healthcare, and Higher Education, in Las Vegas, NV, United States ISBN 978-1-939797-35-3 Publisher: Association for the Advancement of Computing in Education (AACE), San Diego, CA

Abstract

In this study, we developed a system that automatically estimates the difficulty level of quizzes created by learners in question-posing learning, a method that is considered to have a very beneficial educational effect. In the Fundamental Programming class of our department, a total of 265 quizzes were created by students over the past three years. We categorized them into three levels of difficulty and labeled them accordingly for use as training data. Three classification algorithms, Naïve Bayesian (NB), Support Vector Machine (SVM), and Neural Network (NN), were employed to estimate the difficulty of the quizzes. The accuracy for the three types of classifiers exceeded the results of a previous study of one-minute paper. The precision and recall rate of each algorithm was investigated, and SV and NN were found to produce nearly identical results, but NB was completely unable to identify quizzes of high difficulty.

Citation

Harayama, K. & Yamagishi, Y. (2018). Machine-Learning-Based Automatic Difficulty Estimation of Quizzes in Question-Posing Learning. In Proceedings of E-Learn: World Conference on E-Learning in Corporate, Government, Healthcare, and Higher Education (pp. 754-761). Las Vegas, NV, United States: Association for the Advancement of Computing in Education (AACE). Retrieved June 4, 2020 from .

References

View References & Citations Map

These references have been extracted automatically and may have some errors. Signed in users can suggest corrections to these mistakes.

Suggest Corrections to References