McU Postgraduate Student Develops A Model for Early Detection of Student Dropouts

A Postgraduate Student of McPherson University has Developed A Model for Early Detection of Student Dropouts

Aina Daniel Ayodele, a postgraduate student from the Department of Computer Science, College of Computing, recently defended his dissertation titled “A Machine Learning Model for Early Detection of Students at Risk of Dropping Out Using Psycho-social Data.” The project was presented in fulfillment of the requirements for the award of a Master’s degree in Computer Science.

Ayodele’s research focused on tackling a growing concern in the educational sector: student dropouts. Recognizing that academic performance is critical to students’ progress, he noted that traditional detection methods often fall short in identifying students at risk. His study explored how psycho-social factors, those beyond mere grades could be leveraged to predict and prevent dropout cases effectively.

The study employed a survey-based approach, gathering responses from 1,744 undergraduate students at McPherson University. Data collected included academic records, demographic details, socioeconomic status, behavioral traits, family background, social relationships, and perceptions of the school environment. This comprehensive dataset allowed for a multifaceted analysis of the factors influencing student retention.

Four machine learning algorithms—Support Vector Machine (SVM), Decision Tree (DT), Logistic Regression (LR), and Random Forest (RF)—were tested on the dataset. These models were evaluated using five performance metrics: accuracy, precision, recall, F1 score, and the ROC AUC score, to determine which could best predict student dropout risk.

The results of the evaluation showed that the Random Forest model outperformed the other algorithms with an AUC score of 0.7849, making it the most accurate predictor among those tested. It was followed by SVM with a score of 0.7012, logistic regression at 0.6317, and decision tree at 0.6291. The study recommended Random Forest as the optimal model for classifying and predicting students at risk of dropping out based on psycho-social data.

Ayodele concluded that institutions could adopt the Random Forest model as a reliable early warning system to guide timely interventions and improve student retention. He emphasized that using machine learning models like RF can significantly improve decision making in educational management through data driven insights.

The dissertation defense was examined by a distinguished panel, with Prof. Shade Kuyoro, a Professor of Computer Science from Babcock University, serving as the External Examiner. Dr. Emmanuel Ibam chaired the examination panel, while Dr. Kayode Oladapo and Dr. Oladapo Adeduro served as the Supervisor and Co-Supervisor respectively.

Recommended Articles: