<![CDATA[Delays in student completion are a critical issue in higher education because they impact academic efficiency, program accreditation, and graduate quality. This study aims to develop a machine-learning-based model for predicting student graduation using an integrated pipeline. This pipeline encompasses data processing, model building, and hyperparameter optimisation. The dataset was obtained from eight semesters of student academic data, totalling 146 credits. This dataset includes both numeric and categorical variables, such as GPA, number of credits passed per semester, study load, and student background characteristics. Preprocessing was performed using ColumnTransformer, which combined StandardScaler for numeric features and OneHotEncoder for categorical features. A classification model was developed using the Random Forest algorithm and optimised with GridSearchCV to identify the optimal hyperparameter settings. Model evaluation was performed using accuracy metrics, confusion matrices, and classification reports. The findings of this study indicate that the model achieves an accuracy of 81%, suggesting a strong ability to classify students as on-time or late graduates. Feature analysis shows that the average Grade Point Average (GPA), the number of Semester Credit Units completed each semester, and consistency in study load are the main factors influencing the timeliness of study completion. The implementation of an integrated channel has proven effective in maintaining preprocessing consistency and reducing the possibility of data leakage. The developed model can be implemented as an early warning system to support data-driven academic decision-making.]]>
