PREDICTING HIGHER EDUCATION STUDENTS’ ACADEMIC PERFORMANCE USING MACHINE LEARNING

Abstract

The primary goal of every educational institution is to deliver the best educational experience and knowledge to students. Achieving this goal involves recognizing students in need of extra support and implementing measures to enhance their academic performance. This study investigates five machine learning algorithms to construct a classification model that is capable of predicting students’ academic performance. The machine learning algorithms utilized in this study include Naïve Bayes, Decision Tree, Logistic Regression, Random Forest, and Linear Regression. The model is constructed using three distinct machine-learning platforms: WEKA, RapidMiner Studio, and Python. The dataset for constructing the models are gathered directly from students via the questionnaire data collection method. Initially, data was collected from 3,620 students. After preprocessing, the dataset was reduced to 3,001 participants, comprising 916 females and 2,085 males. The key stages of data preprocessing applied in this study include data cleaning, data reduction, and data transformation. Subsequently, the dataset was divided, allocating 80% for training purposes and 20% for testing. The study adopts an experimental research methodology, constructing a model with chosen machine-learning algorithms and tools. It is developed on a specific training dataset and evaluated based on precision, recall, and accuracy metrics. The experimental results indicate that the random forest algorithm, implemented using Python programming tools, achieved promising outcomes with an accuracy of 95.00%, precision of 95.03%, and recall of 95.01%. The findings of this study are promising and could potentially act as a springboard for additional investigation within this area of research. The study identified a clear link between academic ranking and various factors such as socio-demographic characteristics, economic background, and educational practices. These factors encompass the student's place of origin (be it urban, rural, or emerging regions), family background (including parents' education and economic standing), previous academic performance, time allocated for studying, materials used for examination preparation, and hours spent with peers. This research utilized exclusively student data gathered from Haramaya University. Therefore, it is recommended that future researchers strive to develop a generic model by collecting data from a diverse range of Ethiopian universities.