Abstract:
The development of information systems and technology in higher education
requires service evaluation that is fast, objective, and measurable. At UMSU, these
services are managed by BSTI and are intensively used by students, making student
feedback an important source for service quality improvement. However, open-
ended questionnaire responses are unstructured, so manual analysis tends to be
time-consuming and may lead to inconsistent interpretations. This study aims to
apply the Multinomial Naïve Bayes algorithm to classify student satisfaction
sentiment toward information systems and technology services at UMSU into three
categories: positive, neutral, and Negative. The data were collected from students’
written responses through Google Forms, with a minimum sample of 377
respondents selected using stratified proportionate sampling. The research stages
include manual Labeling, text preprocessing, TF-IDF feature extraction, and
sentiment classification. The model is evaluated using an 80:20 stratified train-test
split with accuracy, precision, recall, F1-score, and Confusion Matrix as evaluation
metrics. The output of this study is a simple system capable of managing feedback
data, classifying sentiment, and presenting the results in a concise Dashboard as a
basis for evaluating BSTI UMSU services.
