PERBANDINGAN AKURASI METODE NAÏVE BAYES, DECISION TREE (C.45), DAN RANDOM FOREST DALAM MELAKUKAN PREDIKSI MASA TUNGGU KERJA ALUMNI BERDASARKAN DATA TRACER STUDY PADA FAKULTAS PSIKOLOGI UIN JAKARTA

Abstrak

Universities today face a major challenge in preparing graduates who are relevant to the
needs of the global job market. One way to measure this relevance is through tracer studies,
which track alumni to evaluate their job waiting period. However, tracer study data analysis
is still often done manually, which requires a more effective approach to obtain more
accurate and informative results. This study aims to compare the accuracy of three data
mining methods-Naïve Bayes, Decision Tree (C.45), and Random Forest in predicting the job
waiting period of alumni of the Faculty of Psychology UIN Jakarta using alumni tracer study
data in 2020-2022. From 300 raw data, 170 clean data were selected for analysis. A webbased
information system was developed using the Rapid Application Development (RAD)
method and Unified Modeling Language (UML) modeling. Three data mining algorithms
were applied to classify the data and predict the waiting period of alumni employment. The
evaluation results show Naïve Bayes has the highest accuracy of 53.96%, followed by
Random Forest (50.79%) and Decision Tree (46.03%). Although the accuracy is still relatively
low, this web-based system has succeeded in facilitating analysis and reducing dependence
on manual analysis. In conclusion, despite limitations in accuracy, this system has the
potential to assist universities in making strategic decisions related to the development of
graduate quality and suitability for the labor market. This research provides a basis for
further development of tracer study data analysis using data mining methods