Pemilu serentak 2024 di Indonesia menggunakan aplikasi Sirekap 2024 untuk memastikan transparansi perhitungan suara. Penelitian ini menganalisis sentimen ulasan aplikasi Sirekap 2024 di Google Play Store menggunakan metode Recurrent Neural Network (RNN). Data dikumpulkan melalui web scrapping, mengumpulkan 17.705 ulasan dari 1 Februari 2024 hingga 30 Maret 2024. Tahapan preprocessing meliputi text cleansing, normalisasi Slangword, tokenizing, stopword removal, dan stemming. Sentimen dilabeli menggunakan kamus Inset lexicon, menghasilkan 10.651 ulasan negatif, 3.129 positif, dan 1.710 netral. Pembobotan kata menggunakan model Word2vec CBOW dengan ukuran vektor 50. Data dibagi menjadi set pelatihan (80%), validasi (10%), dan pengujian (10%). Model RNN dilatih dan dievaluasi dengan confusion matrix, menghasilkan akurasi 89%. Hasil evaluasi menunjukkan precision, recall, dan f1-score yang baik: kelas netral (precision 0.67, recall 0.72, f1-score 0.69), kelas positif (precision 0.83, recall 0.72, f1-score 0.82), dan kelas negatif (precision 0.94, recall 0.93, f1-score 0.93). Penelitian ini menunjukkan efektivitas model RNN dalam menganalisis sentimen ulasan aplikasi Sirekap 2024.

The 2024 simultaneous elections in Indonesia use the Sirekap 2024 application to ensure transparency of vote counting. This research analyzes the sentiment of Sirekap 2024 app reviews on the Google Play Store using the Recurrent Neural Network (RNN) method. Data was collected through web scrapping, collecting 17,705 reviews from February 1, 2024 to March 30, 2024. Preprocessing stages include text cleansing, Slangword normalization, tokenizing, stopword removal, and stemming. Sentiment was labeled using the Inset lexicon dictionary, resulting in 10,651 negative, 3,129 positive, and 1,710 neutral reviews. Word weighting used the Word2vec CBOW model with a vector size of 50. Data was divided into training (80%), validation (10%), and testing (10%) sets. The RNN model was trained and evaluated with confusion matrix, resulting in 89% accuracy. The evaluation results showed good precision, recall, and f1-score: neutral class (precision 0.67, recall 0.72, f1-score 0.69), positive class (precision 0.83, recall 0.72, f1 score 0.82), and negative class (precision 0.94, recall 0.93, f1-score 0.93). This research demonstrates the effectiveness of the RNN model in analyzing the sentiment of Sirekap 2024 app reviews.