Analisis Sentimen Pengguna X (Twitter) Terhadap Gelombang Pemutusan Hubungan Kerja (PHK) Di Indonesia Dengan Metode Regresi Logistik Biner, Naïve Bayes, Dan Support Vector Machine (SVM)

Azkadini, Putri (2026) Analisis Sentimen Pengguna X (Twitter) Terhadap Gelombang Pemutusan Hubungan Kerja (PHK) Di Indonesia Dengan Metode Regresi Logistik Biner, Naïve Bayes, Dan Support Vector Machine (SVM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221139-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB)

Abstract

Stabilitas ketenagakerjaan merupakan salah satu pilar penting bagi stabilitas ekonomi nasional. Namun, ketidakpastian ekonomi global turut memicu peningkatan kasus Pemutusan Hubungan Kerja (PHK) di Indonesia, terutama pada periode 2022–2025. Isu ini memunculkan respons dan opini publik yang luas di media sosial X (Twitter), sehingga diperlukan analisis yang objektif untuk memetakan sentimen masyarakat dan memperoleh gambaran reaksi sosial terhadap fenomena PHK. Penelitian ini bertujuan untuk menganalisis karakteristik sentimen masyarakat serta membandingkan performa tiga metode klasifikasi, yaitu Regresi Logistik Biner, Naïve Bayes, dan Support Vector Machine (SVM). Data yang digunakan berupa cuitan berbahasa Indonesia hasil scraping pada periode 2022 hingga 2025. Tahapan pra-pemrosesan teks meliputi cleaning, case folding, normalisasi, stemming, stopword removal, dan tokenisasi. Pelabelan sentimen dilakukan secara otomatis menggunakan model pra-latih IndoBERT, sementara ekstraksi fitur dilakukan menggunakan Term Frequency–Inverse Document Frequency (TF-IDF). Evaluasi kinerja model dilakukan menggunakan metrik akurasi, presisi, recall, F1-score, dan Area Under the Curve (AUC) untuk mengantisipasi potensi ketidakseimbangan kelas (imbalanced data). Hasil penelitian menunjukkan bahwa sentimen masyarakat terhadap isu PHK didominasi sentimen negatif sebesar 87,60%. Penerapan SMOTE terbukti meningkatkan performa seluruh model secara substansial dengan rata-rata peningkatan AUC sebesar 0,142 dan sensitivity sebesar 65,64 poin persentase. Berdasarkan performa klasifikasi, model terbaik adalah SVM dengan kernel Radial Basis Function (RBF) dan SMOTE dengan akurasi 97,41%, AUC 0,9946, dan sensitivity 96,64%. Sementara itu, berdasarkan kriteria interpretabilitas, Regresi Logistik Biner dengan SMOTE menjadi model terbaik karena menghasilkan estimasi koefisien dan odds ratio yang eksplisit untuk menjelaskan kontribusi tiap kata terhadap sentimen, berbeda dengan SVM yang bersifat black box.
=======================================================================================================================================
Employment stability is one of the key pillars of national economic stability. However, global economic uncertainty has contributed to a rise in layoff in Indonesia, particularly during the period 2022–2025. This issue has generated widespread public responses and opinions on social media platform X (Twitter), necessitating an objective analysis to map public sentiment and gain insight into social reactions toward the layoff phenomenon. This study aims to analyze the characteristics of public sentiment and compare the performance of three classification methods, Binary Logistic Regression, Naïve Bayes, and Support Vector Machine (SVM). The data used consist of Indonesian-language tweets collected through web scraping over the period 2022–2025. Text preprocessing stages include cleaning, case folding, normalization, stemming, stopword removal, and tokenization. Sentiment labeling was performed automatically using the pre-trained IndoBERT model, while feature extraction was carried out using Term Frequency–Inverse Document Frequency (TF-IDF). Model performance was evaluated using accuracy, precision, recall, F1-score, and Area Under the Curve (AUC) to address potential class imbalance. The results show that public sentiment toward the layoff issue is dominated by negative sentiment at 87,60%. The application of SMOTE substantially improved the performance of all models, with an average AUC increase of 0,142 and an average sensitivity increase of 65,64 percentage points. Based on classification performance, the best model is SVM with a Radial Basis Function (RBF) kernel combined with SMOTE, achieving an accuracy of 97,41%, AUC of 0,9946, and sensitivity of 96,64%. Meanwhile, based on interpretability criteria, Binary Logistic Regression with SMOTE emerged as the best model as it produces explicit coefficient estimates and odds ratios to explain each word’s contribution to sentiment, unlike SVM, which operates as a black box.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Sentimen, PHK, Naïve Bayes, Regresi Logistik Biner, Support Vector Machine (SVM) Sentiment Analysis, Layoffs, Binary Logistic Regression, Naïve Bayes, Support Vector Machine (SVM)
Subjects:	H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. H Social Sciences > HA Statistics > HA31.7 Estimation Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Putri Azkadini
Date Deposited:	31 Jul 2026 03:37
Last Modified:	31 Jul 2026 03:37
URI:	http://repository.its.ac.id/id/eprint/140656

Actions (login required)

View Item