Analysis of Predictive Factors for High School Teacher’s Instructional Improvement Activities Using Random Forest

This study utilized random forest (RF) regression and classification models to predict instructional improvement activities among high school teachers and explore factors classifying teachers with low engagement in such activities. It employed data from the 9th wave of the Government Employees Pension Service (GEPS), with 3,271 teachers as the analysis subjects. A total of 168 items were employed as explanatory factors. The main research results are as follows. First, the RF regression model analysis showed that teachers’ collaboration on teaching and learning methods and teaching efficacy were ranked 1st and 2nd. Additionally, four factors related to school average decision-making authority, such as budget allocation, were ranked within the top 10. Second, the classification model analysis found that three factors related to school average decision-making authority, including class formation (ranked 1st) were ranked within the top 10. Additionally, teachers’ average cooperation time on school administrative tasks was ranked 2nd. Common key explanatory factors identified by the RF model included school average decision-making authority related to class formation, average satisfaction with streamlining school approval processes, learner-centered teaching that gathers students’ opinions, and principal leadership encouraging the application of new teaching methods. The implications based on this are as follows: expanding teachers’ decision-making authority, creating environments conducive to collaborative activities among teachers, improving teachers’ school satisfaction related to relationships with school personnel, administrative workload reduction, and promoting the spread of student-centered teaching.