STUDENTS DATA CLASSIFICATION MODEL

In this project, research is conducted based on data sets of undergraduates at varsity level
to classify student performance data. The objective of the project is to develop a system
that utilizes various intelligent techniques with targeted accuracy being at a minimal level
of88%.
The system is designed to predict students' CGPA upon graduation. Any further actions
that can be taken to avoid students' dismissals, or to strengthen their area of interest or
expertise can be derived from the outcome of this intelligent system.
The project is implemented using data sets Iris and Student. Techniques used to support
classification are separated into two different subprojects: (1) Back propagation feed
forward neural network using Bayes probability to initialize weights, and (2) Fuzzy
system. The proposed optimization of neural network and Bayes Theorem returns
92.55% level of accuracy for the student data. Further improvements can be performed
on areas such as the individual variations of each technique and the combination of all
three techniques to optimize accuracy.
The project contributes in customizing a grading system for Universiti Teknologi
PETRONAS. This system structure is generally relevant to many universities in Malaysia
as they adopt a fairly similar approach in grading