Fadlurohman, Refy (2026) IMPLEMENTASI METODE HYBRID DEEP LEARNING (CNN-BILSTM) UNTUK ANALISIS SENTIMEN ISU AKSI KAMISAN TERHADAP PENEGAKAN HAM DI MEDIA SOSIAL X. S1 / D3 thesis, Universitas Kuningan.
ABSTRAK.pdf
Download (1MB)
BAB I.pdf
Download (305kB)
BAB II.pdf
Restricted to Repository staff only
Download (393kB) | Request a copy
BAB III.pdf
Restricted to Repository staff only
Download (484kB) | Request a copy
BAB IV.pdf
Restricted to Repository staff only
Download (4MB) | Request a copy
BAB V.pdf
Download (217kB)
DAFTAR PUSTAKA.pdf
Download (215kB)
LAMPIRAN.pdf
Restricted to Repository staff only
Download (2MB) | Request a copy
Abstract
Media sosial X telah menjadi ruang publik bagi masyarakat untuk menyuarakan opini dan aspirasi politik, salah satunya terkait isu pelanggaran Hak Asasi Manusia (HAM) melalui gerakan Aksi Kamisan. Dinamika teks opini yang masif dan tidak terstruktur ini perlu dianalisis secara komputasi guna memetakan persepsi masyarakat terhadap penegakan hukum oleh pemerintah. Penelitian ini bertujuan untuk mengimplementasikan dan mengevaluasi kinerja metode Hybrid Deep Learning dalam mengklasifikasikan sentimen publik ke dalam kategori positif, negatif, dan netral. Data yang digunakan bersumber dari 4.938 cuitan di media sosial X yang dikumpulkan melalui proses crawling. Tahapan pra-pemrosesan teks meliputi pembersihan data, penyeragaman huruf, normalisasi slang word, penghapusan kata hubung, dan stemming. Teks kemudian diubah menjadi representasi vektor menggunakan Word2Vec, sementara ketidakseimbangan kelas pada dataset ditangani secara efektif dengan teknik SMOTE. Arsitektur model yang diusulkan menggabungkan Convolutional Neural Network (CNN) untuk mengekstraksi fitur lokal dan Bidirectional Long Short-Term Memory (BiLSTM) untuk menangkap konteks kalimat secara dua arah. Hasil pengujian menggunakan Confusion Matrix menunjukkan bahwa model hybrid CNN-BiLSTM mampu mencapai tingkat akurasi sebesar 70% pada data riil. Berdasarkan hasil klasifikasi, sentimen publik didominasi oleh kelas positif (50,8%) dan netral (43,7%), sedangkan sentimen negatif hanya mencakup sebagian kecil (5,5%). Penelitian ini menyimpulkan bahwa pendekatan hybrid cukup memadai dan layak ( feasible ) dalam membedah sentimen teks bahasa Indonesia yang kompleks, sekaligus mengonfirmasi tingginya dukungan digital masyarakat terhadap gerakan Aksi Kamisan.
Social media X has become a public space for the community to voice their political opinions and aspirations, one of which is related to the issue of Human Rights violations through the Aksi Kamisan movement. The dynamics of these massive and unstructured opinion texts need to be computationally analyzed to map public perception regarding law enforcement by the government. This study aims to implement and evaluate the performance of a Hybrid Deep Learning method in classifying public sentiment into positive, negative, and neutral categories. The data used originates from 4,938 tweets on social media X collected through a crawling process. The text preprocessing stages include data cleaning, case folding, slang word normalization, stopword removal, and stemming. The text is then converted into vector representations using Word2Vec, while class imbalance in the dataset is effectively handled using the SMOTE technique. The proposed model architecture combines a Convolutional Neural Network (CNN) to extract local features and a Bidirectional Long Short-Term Memory (BiLSTM) to capture sentence context bidirectionally. Testing resultsing a Confusion Matrix show that the CNN-BiLSTM hybrid model is able to achieve an accuracy rate of 70% on real data. Based on the classification results, public sentiment is dominated by positive (50.8%) and neutral (43.7%) classes, while negative sentiment only covers a small portion (5.5%). This study concludes that the hybrid approach is adequate and feasible in dissecting complex Indonesian text sentiment, while simultaneously confirming the high digital support of the community for the Aksi Kamisan.
| Item Type: | Thesis (S1 / D3) |
|---|---|
| Uncontrolled Keywords: | Kata Kunci: Aksi Kamisan, Analisis Sentimen, CNN-BiLSTM, Deep Learning, Word2Vec. Keywords: Aksi Kamisan, CNN-BiLSTM, Deep Learning, Sentiment Analysis, Word2Vec. |
| Subjects: | H Social Sciences > H Social Sciences (General) J Political Science > JA Political science (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > T Technology (General) |
| Divisions: | Fakultas Ilmu Komputer > S1 Sistem Informasi |
| Depositing User: | S.Kom Refy Fadlurohman |
| Date Deposited: | 20 Aug 2026 06:45 |
| Last Modified: | 20 Aug 2026 06:45 |
| URI: | https://rama.uniku.ac.id/id/eprint/6323 |

