Deep Learning을 活用한 學生 이탈율 방지 방안 硏究

Abstract

Currently, there are many problems due to the decline in school-age population. Moreover, Korea has the largest number of universities compared to the population, and the university enrollment rate is also the highest in the world. As a result, the minimum student retention rate required for the survival of each university is becoming increasingly important. The purpose of this study is to find out how the number of freshmen in Jeju Island from 2011 to 2014, and 8,000 students who have already selected the school, The basic direction for managing students' retention rate, which is consistently maintained from admission to graduation, is based on the data collected by gender, departure report, area of ​​origin, grades, graduation, etc. I want to know if it is. Based on the optimal input parameters, the association analysis is performed using the apriori algorithm based on the optimal input parameters, and the most suitable training data is collected for maintenance rate management. Based on this, Deep Learning We will make the module as a basic data for development. A total of 8891 students' data were separated into training data for building a deep learning module and testing data for evaluating the model. It is shown that the students who graduated from the specialization college and graduated from college are more likely to abandon the school in the middle of the year. These results indicate that the specialization high school is more difficult in terms of academic achievement and academic continuity, And seems to need attention. Deep learning consisted of three hidden layers and initialized the weight by using Xavier initialization module with learning rate of 0.5, and maintained accuracy of 80%. In order to show accuracy more than 90%, it is necessary to acquire more various training data by region and university and to apply the module layer more widely.