Analisis Sentimen Terhadap Kebijakan Efisisensi Anggaran Pemerintah Menggunakan Metode Support Vector Machine (SVM) Dan Vector Space Model (VSM)

Dwi Fanti Utina (Universitas Muhammadiyah Gorontalo)
Alter Lasarudin
Wahyudin Hasyim (Universitas Muhammadiyah Gorontalo)

Jurnal Nasional Komputasi dan Teknologi Informasi
Vol. 9 No. 2 (2026): April, 2026

Abstrak - Media sosial saat ini menjadi wadah utama bagi masyarakat dalam menyuarakan aspirasi terhadap kebijakan pemerintah secara terbuka, termasuk isu kebijakan efisiensi anggaran. Penelitian ini bertujuan menganalisis kecenderungan sentimen publik terhadap kebijakan pemerintah menggunakan algoritma Support Vector Machine (SVM) yang diintegrasikan dengan Vector Space Model (VSM). Data dikumpulkan melalui teknik crawling pada platform X (Twitter), kemudian diproses melalui tahapan cleaning, case folding, tokenizing, filtering, dan stemming. Penelitian ini juga menerapkan pendekatan Lexicon-Based untuk pelabelan sentimen. Seluruh implementasi sistem dikembangkan menggunakan bahasa Python pada Google Colaboratory. Hasil pengujian dengan confusion matrix menunjukkan bahwa model mampu mengklasifikasikan data berlabel mandiri dengan akurasi 71%, dan mencapai 96% pada data berlabel Lexicon-Based Approach. Hal ini membuktikan integrasi VSM sangat membantu SVM dalam menyajikan representasi fitur dan mengenali pola teks secara akurat. Penelitian ini menyimpulkan bahwa persepsi masyarakat dapat dipetakan secara objektif melalui analisis sentimen, sehingga dapat menjadi bahan evaluasi dan referensi strategis bagi pemerintah dalam mengambil kebijakan di masa depan. Kata Kunci: Analisis Sentimen; Support Vector Machine; Efisiensi Anggaran; X (Twitter); Vector Space Model;   Abstract - Social media is currently the primary platform for the public to openly voice their concerns about government policies, including issues of budget efficiency. This study aims to analyze public sentiment trends toward government policies using the Support Vector Machine (SVM) algorithm integrated with the Vector Space Model (VSM). Data was collected through crawling techniques on the X (Twitter) platform, then processed through cleaning, case folding, tokenizing, filtering, and stemming. This study also applied a Lexicon-Based approach for sentiment labeling. The entire system implementation was developed using Python on Google Collaboratory. Test results using a confusion matrix showed that the model was able to classify self-labeled data with 71% accuracy and achieved 96% accuracy on data labeled with the Lexicon-Based Approach. This demonstrates that the VSM integration significantly assists the SVM in presenting feature representations and accurately recognizing text patterns. This study concludes that public perception can be objectively mapped through sentiment analysis, thus serving as evaluation material and a strategic reference for the government in future policymaking.                                                                                                     Keywords: Sentiment Analysis; Support Vector Machine; Budget Efficiency; X (Twitter); Vector Space Model;