New Perspectives in Science Education

Edition 14

Accepted Abstracts

Analysis of the Influence of Learning State before University Admission to College Dropout Using Hierarchical Bayesian Model

Naruhiko Shiratori, Kaetsu University (Japan)

Shintaro Tairi, Yokohama College of Commerce (Japan)

Tetsuya Oishi, Tokyo Institute of Technology (Japan)

Masao Mori, Tokyo Institute of Technology (Japan)

Masao Murota, Tokyo Institute of Technology (Japan)

Abstract

In this research, we express the probability of college students dropping out using a hierarchical Bayesian model and derive to what extent the variables before admission influence college dropout. In Japan, about 5% of students entered university, but they drop out without graduation. Many studies have shown that the proportion of dropout students varies from university to university and correlates with university scale and deviation value. Many researches have been made to forecast dropouts in advance, but studies have not been made to express how the dropout probability dynamically changes depending on the situation before admission. It is assumed that the change in the dropout probability varies according to the variables before enrollment (number of absences at high school, type of high school etc). Variables used in the model consist of pre-admission variables and post-admission variables, and the post-admission variables are the number of units per semester and GPA. We created a hierarchical Bayesian model using pre-admission variables and post-admission variables to derive the drop-out probability for that term. It was found that the probability of dropping out depends on the variables before entering the school, especially the type of high school and the number of absences. The probability of withdrawal is increased because the type of high school is communication system and the number of days absent at high school is large. Furthermore, we found that the change in the probability of dropout for each term differs for each variable before enrollment.

Keywords: Hierarchical Bayesian Model, Logistic Regression, Dropout;

 

Back to the list

REGISTER NOW

Reserved area


Media Partners:

Click BrownWalker Press logo for the International Academic and Industry Conference Event Calendar announcing scientific, academic and industry gatherings, online events, call for papers and journal articles
Pixel - Via Luigi Lanzi 12 - 50134 Firenze (FI) - VAT IT 05118710481
    Copyright © 2024 - All rights reserved

Privacy Policy

Webmaster: Pinzani.it