[3Xin4] Poster session 1

[3Xin4-72] Correct Answer Rate Prediction using Features Extracted from Test Questions

〇Kaoru Hiramatsu¹ (1.Saitama University)

Keywords:academic achievement test, feature, test

In tests designed to assess student achievement levels, there must be appropriate variation in the difficulty of the questions, and the distribution of scores must be close to the variation in student understanding. To achieve this goal, the adjustment of question representations has been done by experienced question creators with reference to past example questions and their knowledge. To help their adjustment, we created a regression model using LightGBM to predict the correct answer rate from the linguistic features of the question sentence and the attributes of questions, and verified the performance of the regression model.