Zulfa, Alrivia (2026) Analisis Temporal Sentimen Publik Terhadap Tapera Menggunakan LSTM Untuk Mengidentifikasi Pola Opini. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5002211069_Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5002211069_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Kebijakan Tabungan Perumahan Rakyat (Tapera) pasca-pengumuman pada Mei 2024 memicu opini publik yang beragam di media sosial, terutama kritik terhadap pemotongan gaji wajib. Penelitian ini bertujuan menganalisis sentimen temporal dan mengidentifikasi pola opini masyarakat terhadap Tapera menggunakan komentar TikTok, untuk memahami dinamika opini terhadap kebijakan ekonomi sensitif. Metode meliputi web scraping data pasca-pengumuman, preprocessing teks (cleaning, case folding, tokenization, normalisasi slang, stemming, word embedding, labeling lexicon-based dengan InSet, word embedding FastText), klasifikasi sentimen menggunakan LSTM Bidirectional, serta clustering TF-IDF dengan K-Means. Analisis temporal dilakukan secara mingguan dan bulanan. Hasil testing menunjukkan model LSTM Bidirectional mencapai akurasi 95.68 persen dan F1-score macro average 0.96, dengan validation loss akhir 0.224434 – menandakan generalisasi tinggi dan keseimbangan prediksi di ketiga kelas sentimen. Analisis temporal mengungkap sentimen negatif dominan di awal periode (proporsi rata-rata 75 persen), lonjakan hingga 95 persen pada pertengahan (Oktober–November 2024), dan penurunan hingga 45 persen di akhir (Desember 2024–Maret 2025), disertai peningkatan gabungan netral-positif hingga 55 persen. Clustering dengan K=5 mengidentifikasi lima pola utama, di mana tema kritik (66 persen keseluruhan) mendominasi awal dan menurun seiring waktu, sementara tema dukungan terhadap tujuan perumahan (11 persen keseluruhan) meningkat hingga 25 persen di akhir periode. Penelitian ini menyimpulkan bahwa opini publik terhadap Tapera bersifat dinamis: resistensi kuat awal berbasis kekhawatiran ekonomi, diikuti amplifikasi negatif, dan adaptasi bertahap dengan munculnya dukungan manfaat hunian. Temuan ini memberikan wawasan bagi pembuat kebijakan untuk sosialisasi dini dan respons cepat terhadap spike opini negatif, meskipun tantangan noise lexicon dan bahasa informal tetap ada.
====================================================================================================================================
The announcement of the People's Housing Savings (Tapera) policy in May 2024 triggered diverse public opinions on social media, particularly criticism regarding mandatory salary deductions. This study aims to analyze temporal sentiment and identify opinion patterns toward Tapera using TikTok comments, to understand the dynamics of public opinion on sensitive economic policies. The methods include web scraping post-announcement data, text preprocessing (cleaning, case folding, tokenization, slang normalization, stopword removal, stemming, lexicon-based labeling with InSet, FastText word embedding), sentiment classification using Bidirectional LSTM, and TF-IDF clustering with K-Means. Temporal analysis was conducted weekly and monthly.Testing results show the Bidirectional LSTM model achieved a test accuracy of 95.68% and F1-score macro average of 0.96, with a final validation loss of 0.224434—indicating high generalization and balanced prediction across all three sentiment classes. Temporal analysis revealed dominant negative sentiment in the early period (average proportion 75%), a spike up to 95% in the middle (October–November 2024), and a decline to 45% at the end (December 2024–March 2025), accompanied by an increase in combined neutral-positive sentiment to 55%. Clustering with K=5 identified five main patterns, where criticism themes (66% overall) dominated the beginning and decreased over time, while support for housing goals (11% overall) rose to 25% in the final period.This study concludes that public opinion toward Tapera is dynamic: strong initial resistance based on economic concerns, followed by negative amplification, and gradual adaptation with emerging support for housing benefits. These findings provide insights for policymakers on early socialization and rapid response to negative opinion spikes, despite challenges from lexicon noise and informal language variations.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Analisis Sentimen, LSTM, Tapera, TikTok, Sentiment Analysis, LSTM, Tapera, Tiktok |
| Subjects: | Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA278 Cluster Analysis. Multivariate analysis. Correspondence analysis (Statistics) T Technology > T Technology (General) > T385 Visualization--Technique T Technology > T Technology (General) > T57.5 Data Processing |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis |
| Depositing User: | Alrivia Zulfa |
| Date Deposited: | 28 Jan 2026 03:08 |
| Last Modified: | 28 Jan 2026 03:08 |
| URI: | http://repository.its.ac.id/id/eprint/130668 |
Actions (login required)
 |
View Item |