DOI QR코드

DOI QR Code

Prediction of Correct Answer Rate and Identification of Significant Factors for CSAT English Test Based on Data Mining Techniques

데이터마이닝 기법을 활용한 대학수학능력시험 영어영역 정답률 예측 및 주요 요인 분석

  • 박희진 (서울과학기술대학교 IT정책전문대학원 산업정보시스템전공) ;
  • 장경애 (서울과학기술대학교 IT정책전문대학원 산업정보시스템전공) ;
  • 이윤호 (서울과학기술대학교 글로벌융합산업공학과) ;
  • 김우제 (서울과학기술대학교 글로벌융합산업공학과) ;
  • 강필성 (고려대학교 산업경영공학부)
  • Received : 2015.07.03
  • Accepted : 2015.08.30
  • Published : 2015.11.30

Abstract

College Scholastic Ability Test(CSAT) is a primary test to evaluate the study achievement of high-school students and used by most universities for admission decision in South Korea. Because its level of difficulty is a significant issue to both students and universities, the government makes a huge effort to have a consistent difficulty level every year. However, the actual levels of difficulty have significantly fluctuated, which causes many problems with university admission. In this paper, we build two types of data-driven prediction models to predict correct answer rate and to identify significant factors for CSAT English test through accumulated test data of CSAT, unlike traditional methods depending on experts' judgments. Initially, we derive candidate question-specific factors that can influence the correct answer rate, such as the position, EBS-relation, readability, from the annual CSAT practices and CSAT for 10 years. In addition, we drive context-specific factors by employing topic modeling which identify the underlying topics over the text. Then, the correct answer rate is predicted by multiple linear regression and level of difficulty is predicted by classification tree. The experimental results show that 90% of accuracy can be achieved by the level of difficulty (difficult/easy) classification model, whereas the error rate for correct answer rate is below 16%. Points and problem category are found to be critical to predict the correct answer rate. In addition, the correct answer rate is also influenced by some of the topics discovered by topic modeling. Based on our study, it will be possible to predict the range of expected correct answer rate for both question-level and entire test-level, which will help CSAT examiners to control the level of difficulties.

대학수학능력시험(수능)은 고등학교 3년간의 학업 성취도를 측정하는 대표적인 평가 도구로서 대한민국 대학 입시에 있어 매우 중요한 역할을 하는 시험이다. 응시생들의 학업 성취도를 효과적으로 평가하기 위해서는 수능의 난이도가 적절하게 조절되어야 하나 지금까지는 수능 난이도의 편차가 매우 크게 나타나 매 입시연도마다 여러 가지 문제점을 야기해왔다. 본 연구에서는 전문가의 판단에 의존한 기존 방식에서 벗어나 지금까지 시행된 모의고사 및 실제 시험을 통해 축적된 자료를 바탕으로 데이터마이닝 기법을 적용하여 영어영역 문제의 난이도를 예측하는 모델을 구축하고 난이도 예측에 영향을 미치는 요소를 판별하고자 한다. 이를 위해 각 문항의 특성을 판별할 수 있는 여러 지표와 함께 지문, 문제, 답안 등에 나타난 단어들의 특징을 토픽 모델링(topic modeling) 기법을 이용하여 정량화하고 이를 바탕으로 선형회귀분석 및 의사결정나무 기법을 이용하여 각 문항의 난이도를 예측하는 모델을 구축하였다. 구축된 예측 모델을 실제 문제에 적용한 결과 난이도의 상/하 구분에 대한 예측 정확도는 90% 수준으로 나타났으며, 실제 정답률 대비 오차 비율은 약 16% 이내인 것으로 나타났다. 또한 배점 및 문제 유형이 문제의 난이도에 큰 영향을 미치며 지문이 특정 주제에 관련된 경우에도 난이도에 영향을 미치는 것을 확인하였다. 본 연구에서 제시된 방법론을 이용하여 영어영역 각 문제들에 대한 기대 정답률의 범위를 추정할 수 있으며 이를 종합하여 영어영역 전체 문제에 대한 정답률 예측을 통해 적절한 난이도의 문제를 출제하는 데 기여할 수 있을 것으로 기대한다.

Keywords

References

  1. 2015 school year the CSAT questions headquarters, "2015 school year, the CSAT Press," in Proceedings 2015 school year the CSAT questions headquarter, 2014.
  2. Korea Institute for Curriculum and Evaluation, "2015 school year CSAT score results press release," in Proceedings Korea Institute for Curriculum and Evaluation, 2014.
  3. Korea Institute for Curriculum and Evaluation, "2015 school year CSAT plan," in Proceedings Korea Institute for Curriculum and Evaluation, 2014.
  4. T. C. Kang, "CSAT Improvement Study," Ministry of Education, pp.57-77, 2013.
  5. M. K. Kang and Y. M. Kim, "The internal analysis of the validation on item-types of Foreign (English) Language Domain of the current 2005 CSAT for designing the level-differentiated English tests of the 2014 CSAT," Journal of the Korea English Education Society, Vol.12, No2, pp.1-35, 2013.
  6. K. S. Lee, "The effects of th number of questions per passage, the length of passage, and the topic familiarity on multiple-choice English listening and reading comprehension tests," English Teaching, Vol.54, No.4, pp.327-351, 1999.
  7. N. B. Kim, "A corpus-based lexical analysis of the foreign language(English) test for the college scholastic ability test (CSAT)," English Language & Literature Teaching, Vol.14, No.4, pp.201-221, 2008.
  8. K. S. Chang, "A model of predicting item difficulty of the reading test of College Scholastic Ability Test," Foreign Languages Education, Vol.11, No.1, pp.111-130, 2004.
  9. Y. M. Sung, "Factor Analysis of English Test Scores in the College Scholastic Ability Test and Implications," Ph.D. dissertation, Inha University Graduate School, 2003.
  10. H. W. Lee and S. Y. Lee, "A study on the relationship between the scores of TOEFIC, TOEIC and TEPS, and college academic performance," English Language & Literature Teaching, Vol.9, No.1, pp.153-171, 2003.
  11. L. Breiman, J. Friedman, R. Olshen, and C. Stone, "Classification and Regression Trees," Wadsworth, 1984.
  12. D. Hand, H. Mannila, and P. Smyth, "Principles of Data Mining," A Bradford Book The MIT Press, 2001.
  13. F. Sebstiai, "Machine learning in automated text categorization," ACM Computing Surverys, Vol.34, No.1, 2002.
  14. J. H. Bae, J. E. Son, and M. Song, "Analysis of twitter for 2012 South Korea presidential election by text mining techniques," Journal of Intelligent Information Systems, Vol.19, No.3, pp.141-156, 2013.
  15. H. J. Lee and J. C. Park, "Probabilistic filtering for a biological knowledge discovery system with text mining and automatic inference," Journal of the Korea Society of Computer and Information, Vol.17, No.2, pp.139-147, 2012. https://doi.org/10.9708/jksci.2012.17.2.139
  16. D. Blei, "Probabilistic topic models," Communications of the ACM, Vol.55, No.4, pp.77-84, 2012. https://doi.org/10.1145/2133806.2133826
  17. S. R. Kang, "A Study on the Readability of English Textbooks: Middle School English 1 and 2 Based on the Revised 7th English National Curriculum," Master Dissertation, Inha University Graduate School, 2010.