A Dynamic Data Mining Model to Predict Student Graduation Level

Abstract

This study sought to develop an intelligent and dynamic model to predict early on a student's graduate level after graduation. For the purpose of early evaluation and to avoid graduating by a critical level, through a number of independent variables: some academic achievement scores in the high school diploma, the result of the first year and the score of the achievement test for some subjects for the first year of study, the student community for the scores of 2012, 2013 and 2014 at the Sudan University of Science and Technology, which numbered 326 Student. This study is based on three research questions Q1: It possible build a model to predict the student’s graduation level in a dynamic way that makes it easy to deploy and generalize? Q2: Is it possible to reprocess the model data set and choose the inputs automatically? Q3: can we identify courses that most influence the student’s graduation level through available academic data? In this research, we used the concept of data mining and machine learning (ML) algorithms and techniques to build a dynamic predictive model that can be used to perform the prediction process, where a feature selection technique was used to let the model select the best attributes automatically Which have high relationship and effect of the dependent variable, and then a linear regression algorithm was used to predict the student's rate upon graduation. We also used coefficient and correlation analysis and other statistical methods that were implemented in order to obtain the expected initial results and to know the possibility of building the model as a preliminary analytical study for the research. And we used R-squared method to evaluate the model. And we have reached through research that a model has been built and we can predict a student’s graduation level in a dynamic way that facilitates to deploy it and generalization. We also reached the ability to pre-process data and choose appropriate inputs for the model automatically, which made the model flexible, reliable and usable. The model was applied and tested on students who graduated for the years 2016, 2017 and 2018. The research concluded with a number of recommendations and future work that could become a continuation of this study. An application has also been designed that allows using the model and obtaining the results of student’s rate expecting at graduation. The application also allows uploading any data that is required to be analyzed and trained on the model through easy and simple user interfaces.