Using Logistic Regression Models and Artificial Neural Networks to Study the Factors Affecting the Academic Achievement of University Student

Abstract

This study aimed to identify the most important factors affecting the academic achievement of a university student and then build a predictive model using these factors to predict the student's academic status. The ordinal logistic regression method was used to identify the influencing factors, and neural network technique to build the prediction model. It was applied to a sample of 188 students at Al-Yamamah University in the Kingdom of Saudi Arabia. The analyzed data were collected using a questionnaire of three main axes: representing student related factors, family, and academic factors. The results of the logistic regression model showed that 10 variables had a significant effect on the student's academic achievement: age, cumulative average at high school, the distance of residence, work besides studying, daily study hours, frequent absences, fear of exams, the economic level of the family, parents divorce, and family follow-up. As for the MLP network model with architecture (18-28-3), whose inputs were the significant variables chosen by the logistic model, the values of all performance indicators were supportive of the model's quality and predictive ability, where the correct classification rate values in the training data reached 96%, validity 93% and testing 90%. The results of the classification table were also high, with the correct classification of the three categories of the target variable: 100% good, 91.95% fair, 82.35% poor. The total percentage of correct classification was 94.68%. All these indicators confirm the validity and feasibility of using this model as an initial tool for predicting the academic status of a university student.