University entry selection framework using rule-based and back-propagation

Processing thousands of applications can be a challenging task, especially when the applicant does not consider the university requirements and their qualification. The selection officer will have to check the program requirements and calculate the merit score of the applicants. This process is base...

Full description

Saved in:
Bibliographic Details
Main Author: Maharani, Sitti Syarah
Format: Thesis
Language:English
Published: 2015
Subjects:
Online Access:http://psasir.upm.edu.my/id/eprint/65284/1/FSKTM%202015%2023%20upm%20ir.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Processing thousands of applications can be a challenging task, especially when the applicant does not consider the university requirements and their qualification. The selection officer will have to check the program requirements and calculate the merit score of the applicants. This process is based on rules determined by the Ministry of Education and the institution will have to select the qualified applicants among thousands of applications. In recent years, several student selection methods have been proposed using the fuzzy multiple decision making and decision trees. These approaches have produced high accuracy and good detection rates on closed domain university data. However, current selection procedure requires the admission officers to manually evaluate the applications and match the applicants’ qualifications with the program they applied. Because the selection process is tedious and very prone to mistakes, a comprehensive approach to detect and identify qualified applicants for university enrollment is highly desired. In this work, a student selection framework using rule-based and backpropagation neural network is presented. Two processes are involved in this work; the first phase known as pre-processing uses rule-based for checking the university requirements, merit calculation and data conversion to serve as input for the next phase. The second phase uses back-propagation neural network model to evaluate the qualified candidates for admission to particular programs. This means only selected data of the qualified applicants from the first phase will be sent to the next phase for further processing. The dataset consists of 3,790 datasets from Universiti Pendidikan Sultan Idris. The experiments have shown that the proposed method of ruled-based and back-propagation neural network produced better performance, where the framework has successfully been implemented and validated with the average performance of more than 95% accuracy for student selection across all sets of the test data.