Implementation Of the Decision Tree Algorithm in Predicting Student Graduation Rate Based on Academic and Absence Data at A Vocational High School in Bandung

Abstract

Education is a fundamental component of human resource development, directly impacting a nation's advancement. Graduation serves as a benchmark for the efficacy of this educational process. A vocational high school in Bandung is experiencing difficulties with pupils failing to graduate on schedule, resulting in low graduation rates. The reactive management of this condition complicates early intervention. This study advocates for the application of a Decision Tree algorithm as a predictive mechanism to enhance student graduation rates. The research employs a quantitative methodology, utilizing a predictive model grounded in academic data from students at a vocational high school in Bandung. The applied methodology is Prototype, which includes steps such as needs analysis, design, prototype model development, testing, assessment, and refining. In the data preparation step, missing values are imputed, features are derived, irrelevant columns are removed, and categorical variables are encoded. The dataset is subsequently split into 80% for training and 20% for testing. 

The Decision Tree Classifier model is constructed and trained utilizing the processed training data. Model performance evaluation uses metrics such as accuracy, precision, recall, and F1-score, along with confusion matrix analysis and feature importance scores. Test findings demonstrate that in four tests conducted under diverse conditions, the model attains above 90% accuracy in experiments 1, 2, and 3, but accuracy declines to approximately 80% in experiment 4. This indicates that, across varying conditions, the model consistently yields high forecasts and is suitable for deployment as an early warning system for potential graduation delays. This research demonstrates that machine learning can enhance the optimization of student graduation rates.