Perbandingan Pelabelan Data dalam Analisis Sentimen Kebijakan Perubahan Syarat Kelulusan Mahasiswa di Platform TikTok

Mia kamayani; Anissya Agsani Pratiwi;

Perbandingan Pelabelan Data dalam Analisis Sentimen Kebijakan
Perubahan Syarat Kelulusan Mahasiswa di Platform TikTok

Anissya Agsani Pratiwi

Penelitian ini fokus pada analisis sentimen mahasiswa terhadap perubahan syarat kelulusan
mahasiswa yang menghilangkan kewajiban skripsi. Data sentimen diekspresikan melalui platform
media sosial TikTok, dan algoritma Naïve Bayes digunakan untuk mengklasifikasikan sentimen
menjadi positif atau negatif. Proses penelitian mencakup, pengambilan data, pembersihan data,
preprocessing data, pelabelan data, hingga klasifikasi menggunakan algoritma Naive Bayes.
Penelitian ini melibatkan dua tahap pelabelan dalam 913 data: pelabelan pertama manual
menghasilkan 510 sentimen positif dan 403 negatif, sementara pelabelan kedua otomatis dengan
RapidMiner menghasilkan 415 sentimen positif dan 498 negatif. Beberapa mahasiswa memberikan
ulasan positif menganggap hal ini sebagai langkah inovatif untuk persiapan di dunia kerja. Meskipun
beberapa merasa khawatir dengan tingkat kesulitan yang lebih tinggi. Hasil penelitian menunjukkan
mayoritas tanggapan positif terhadap perubahan syarat kelulusan mahasiswa , dengan nilai pelabelan
manual mencapai accuracy 93.98%, precision 100%, recall 87.99%. Sedangkan pelabelan otomatis
dengan Rapidminer memperoleh nilai accuracy 79.63%, precision 83.61%, recall 74.62%.
Kata Kunci: Analisis Sentimen, Algoritma Naïve Bayes, Skripsi, Proyek, TikTok

ABSTRACT

Comparison of Data Labeling in Sentiment Analysis of Policy Changes in Student

Graduation Requirements on the TikTok Platform

Anissya Agsani Pratiwi

This research focuses on analyzing students' sentiments towards the changes in graduation
requirements that eliminate the mandatory thesis. Sentiment data are expressed through the TikTok
social media platform, and the Naïve Bayes algorithm is employed to classify sentiments as positive
or negative. The research process includes data collection, data cleaning, data preprocessing, data
labeling, and classification using the Naive Bayes algorithm. The study involves two labeling stages
with 913 data points: the first manual labeling produces 510 positive sentiments and 403 negative
sentiments, while the second automatic labeling with RapidMiner results in 415 positive sentiments
and 498 negative sentiments. Some students provide positive reviews, considering it an innovative
step for preparation in the professional world, although some express concerns about the higher

level of difficulty. The research findings indicate a majority of positive responses to the project-
based curriculum, with manual labeling achieving an accuracy of 93.98%, precision of 100%, and

recall of 89.22%. Meanwhile, automatic labeling with RapidMiner obtains an accuracy of 79.63%,
precision of 83.61%, and recall of 74.62%.
Keywords: Sentiment Analysis, Algorithm Naïve Bayes, Thesis, Project, Tik Tok.