Analisis Hubungan Sentimen Publik di Media Sosial X dan SPI Terhadap Pemerintah Provinsi di Indonesia Menggunakan Algoritma GRU dan LSTM

Rafli, Andi Muhammad (2024) Analisis Hubungan Sentimen Publik di Media Sosial X dan SPI Terhadap Pemerintah Provinsi di Indonesia Menggunakan Algoritma GRU dan LSTM. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025201089-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (7MB) | Request a copy

Abstract

Media sosial, khususnya Twitter, telah menjadi platform utama bagi masyarakat untuk mengekspresikan pandangan dan opini mereka terhadap berbagai isu, termasuk kinerja pemerintah provinsi di Indonesia. Penelitian ini bertujuan untuk menganalisis sentimen publik di media sosial X (Twitter) terkait pemerintah provinsi dengan menggunakan algoritma Gated Recurrent Unit (GRU) dan Long Short-Term Memory (LSTM). Tahap pertama penelitian melibatkan pengumpulan dan pra-pemrosesan data teks, termasuk case folding, penghapusan tautan, dan lainnya. Setelah itu, data dilabeli secara manual dan dengan model pra-terlatih IndoBERT untuk klasifikasi ke dalam label positif, negatif, dan netral. Metode oversampling dan parameter tuning diterapkan untuk menghasilkan model LSTM dengan akurasi 0,9313, F1-Score 0,9352, dan loss 0,3269.
Analisis topik dengan metode LDA menunjukkan dominasi sentimen negatif pada topik seperti "Persepsi Publik tentang Program dan Kebijakan Anies Baswedan," "Proyek Infrastruktur dan Pengelolaan Lahan," dan "Kinerja Pemerintah Provinsi dalam Sektor Transportasi." Ini menunjukkan kecenderungan negatif masyarakat terhadap isu-isu pemerintah provinsi.
Penelitian ini juga membandingkan analisis sentimen di lima provinsi dengan hasil Survei Penilaian Integritas (SPI). Perhitungan korelasi Pearson menunjukkan nilai sentimen negatif dan positif terhadap SPI masing-masing sebesar 0,53 dan -0,53. Hasil ini mengindikasikan bahwa tingginya nilai SPI tidak selalu berhubungan langsung dengan tingginya atau rendahnya sentimen positif dan negatif, menunjukkan bahwa faktor lain mungkin mempengaruhi bagaimana masyarakat memandang kinerja pemerintah provinsi.
============================================================
Social media, especially Twitter, has become a major platform for people to express their views and opinions on various issues, including the performance of provincial governments in Indonesia. This study aims to analyze public sentiment on social media X (Twitter) regarding provincial governments using Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) algorithms. The first stage of the research involved collecting and preprocessing text data, including case folding, removing links, usernames, punctuation, stopwords, double spaces, duplicate data, emoticons, and tweets with fewer than four words. After that, the data was manually labeled and classified using a pre-trained IndoBERT model into positive, negative, and neutral labels. Data imbalance was addressed using oversampling and parameter tuning methods, resulting in an LSTM model with an accuracy of 0.9313, an F1-Score of 0.9352, and a loss of 0.3269.
Topic analysis using the Latent Dirichlet Allocation (LDA) method shows a dominance of negative sentiment on topics such as "Public Perception of Anies Baswedan's Programs and Policies," "Infrastructure Projects and Land Management by the Provincial Government," and "Public Perception of the Provincial Government's Performance in the Transportation Sector." This indicates a negative tendency in public sentiment towards provincial government issues.
The study also compares sentiment analysis across five provinces with the Integrity Assessment Survey (SPI). Pearson correlation calculations show that the correlations between negative and positive sentiment scores and SPI are 0.53 and -0.53, respectively. This suggests that a high SPI value does not necessarily correlate with high or low positive and negative sentiment, indicating that other factors may influence public perceptions of provincial government performance.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Sentimen, Media Sosial X, Pemerintah Provinsi, GRU, LSTM, SPI, Sentiment Analysis, Social Media X, Provincial Government, GRU, LSTM, SPI
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions:	Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Andi Muhammad Rafli
Date Deposited:	01 Aug 2024 22:26
Last Modified:	01 Aug 2024 22:26
URI:	http://repository.its.ac.id/id/eprint/111297

Actions (login required)

View Item