Analisis Sentimen Ulasan Mahasiswa terhadap Kinerja Dosen Menggunakan Multinomial Naive Bayes Berbasis Web

Chua, Daniel (2026) Analisis Sentimen Ulasan Mahasiswa terhadap Kinerja Dosen Menggunakan Multinomial Naive Bayes Berbasis Web. S1 Teknik Informatika thesis, STMIK Widya Cipta Dharma.

	Text 2243010-S1-Teknik Informatika.pdf Restricted to Repository staff only Download (1MB) | Request a copy
	Text 2243010-S1-Jurnal.pdf Download (422kB)

Abstract

ABSTRACT Student reviews provide qualitative evidence for evaluating lecturer performance, but manual processing is time-consuming, subjective, and difficult when the text contains informal expressions. This study develops a web-based sentiment analysis system that classifies student reviews into positive, neutral, and negative classes while addressing the class imbalance problem. The research follows the Cross-Industry Standard Process for Data Mining (CRISP-DM), covering business understanding, data understanding, data preparation, modeling, evaluation, and deployment. A total of 755 raw responses were collected from 151 students, and manual cleaning produced 413 valid reviews consisting of 217 positive, 174 negative, and 22 neutral reviews. Text preparation included case folding, cleaning, tokenization, slang normalization, stopword removal, and stemming. The data were split into 330 training and 83 testing records. Features were extracted using Term Frequency-Inverse Document Frequency with an N-Gram range of one to three, while Random Over-Sampling was applied only to the training matrix. The optimal Multinomial Naive Bayes model used a Laplace smoothing parameter of 0.1 and achieved 80.72% accuracy, 72.51% macro precision, 79.80% macro recall, and 74.89% macro F1-score. The model was deployed using Flask, React, and MySQL in an interactive dashboard for the Quality Assurance Unit and study programs. Functional black-box testing showed that all tested student and administrator features operated according to the expected results. Keywords: Sentiment Analysis, Multinomial Naive Bayes, TF-IDF, Random Over-Sampling, CRISP-DM =================================== ABSTRAK Ulasan mahasiswa memberikan informasi kualitatif yang penting untuk mengevaluasi kinerja dosen, tetapi pengolahan secara manual membutuhkan waktu, rentan terhadap subjektivitas, dan sulit dilakukan ketika teks menggunakan bahasa tidak baku. Penelitian ini mengembangkan sistem analisis sentimen berbasis web untuk mengklasifikasikan ulasan mahasiswa ke dalam kelas positif, netral, dan negatif sekaligus menangani ketidakseimbangan kelas. Tahapan penelitian menggunakan Cross-Industry Standard Process for Data Mining (CRISP-DM), yaitu pemahaman bisnis, pemahaman data, persiapan data, pemodelan, evaluasi, dan penerapan. Sebanyak 755 ulasan mentah diperoleh dari 151 mahasiswa. Setelah pembersihan manual, diperoleh 413 ulasan valid yang terdiri atas 217 ulasan positif, 174 negatif, dan 22 netral. Persiapan teks meliputi case folding, cleaning, tokenizing, normalisasi kata tidak baku, stopword removal, dan stemming. Data dibagi menjadi 330 data latih dan 83 data uji. Ekstraksi fitur menggunakan Term Frequency-Inverse Document Frequency dengan rentang N-Gram satu sampai tiga, sedangkan Random Over Sampling hanya diterapkan pada matriks data latih. Model Multinomial Naive Bayes terbaik menggunakan parameter Laplace Smoothing 0,1 dan menghasilkan accuracy 80,72%, macro precision 72,51%, macro recall 79,80%, serta macro F1-score 74,89%. Model diimplementasikan menggunakan Flask, React, dan MySQL dalam dashboard interaktif untuk Unit Penjaminan Mutu dan program studi. Pengujian black box menunjukkan seluruh fungsi mahasiswa dan administrator yang diuji berjalan sesuai dengan hasil yang diharapkan. Kata Kunci: Analisis Sentimen, Multinomial Naive Bayes, TF-IDF, Random Over Sampling, CRISP-DM

Item Type:	Thesis (S1 Teknik Informatika)
Additional Information:	Hanifah Ekawati, S.Pd., M.Pd. Wahyuni, S.Kom., M.Kom.
Uncontrolled Keywords:	Analisis Sentimen, Multinomial Naive Bayes, Evaluasi Kinerja Dosen, CRISP-DM, Random Over Sampling
Subjects:	Q Science > QA Mathematics > QA76 Computer software
Divisions:	Teknik Informatika
Depositing User:	Mr Daniel Chua
Date Deposited:	10 Aug 2026 06:36
Last Modified:	10 Aug 2026 06:36
URI:	http://repository.wicida.ac.id/id/eprint/6456

Actions (login required)

View Item