랜덤 포레스트를 활용한 4년제 대학 중도탈락률 예측 요인 탐색: 대학 수준 결정요인을 중심으로

초록

This study aims to provide implications for reducing the dropout rate of 4-year universities from the perspectives of an institutional level. To this end, this study conducted a trend analysis of the dropout rates from 2016 to 2018 using higher education public disclosure information and developed a random forest model predicting the dropout rate of 4- year universities. In this prediction model, 14 variables consisting of 6 factors(e.g. university characteristics, social reputation, faculty research, educational investment, class condition and educational environment) were applied. The percentage of explained variance in the model turned out to be 63.4%. It was found that the dropout rate of 4-year universities remain stable for 3 years from 2016. All 14 variables applied to the prediction model exceeded the minimum standard value for the interpretation of predictive power. The top 5 determinants predicting the dropout rate were research fund benefits per faculty member, the number of research papers published per faculty member, acceptance rate, the number of students, educational expenditure per student. The results of this study demonstrate the importance of planning strategic policies based on periodic management of educational performance index at the institutional level for improving the dropout rate of 4-year universities.

키워드