Thesis

My undergraduate thesis at Universitas Pelita Harapan Medan asked whether a short survey can predict which students will graduate late.

Comparison of Naive Bayes and Support Vector Machine algorithms for predicting on-time student graduation

Perbandingan Algoritma Naïve Bayes dan Support Vector Machine dalam Memprediksi Kelulusan Mahasiswa Tepat Waktu (Studi Kasus: Universitas Pelita Harapan Kampus Medan)

95%Naive Bayes accuracy
90%SVM accuracy
2023Year

Method

  1. A questionnaire collected responses from graduates of UPH Medan Campus, covering study habits, attendance, organisation activity and family support.
  2. Categorical fields were one-hot encoded, then checked against a Pearson correlation pass before training.
  3. Gaussian Naive Bayes and a Support Vector Machine were tuned with GridSearchCV, validated with leave-one-out cross-validation, and compared on accuracy, precision, recall and F1.

Results

MetricNaive BayesSupport Vector Machine
Accuracy95%90%

Abstract

Timely graduation matters to students, who can apply for work and promotions sooner, and to universities, whose quality and reputation depend on it. This research compares Naive Bayes and Support Vector Machine in predicting on-time graduation, in order to identify the more suitable model for planning student support. The data was collected through a questionnaire distributed to graduates of Universitas Pelita Harapan Medan Campus. Both models were trained and evaluated on the same data, and their results were compared: Naive Bayes reached an overall accuracy of 95%, while the Support Vector Machine reached 90%. Naive Bayes is therefore the more suitable model for this prediction.

Keywords: Naive Bayes, Support Vector Machine, graduation on time, classification

Full text

The full thesis is in the vault.

Open the vault

Institution
Universitas Pelita Harapan, Medan Campus
Programme
Informatics
Year
2023
Language
Indonesian, with an English abstract