%0 Thesis
%9 S1
%A SUMBOGO, RIFQI PUTRAWAN
%B Biro Perpustakaan
%D 2025
%F umbprints:96594
%I Universitas Mercu Buana Jakarta
%K Danantara, Sentimen, Naive Bayes, LSTM, Twitter
%P 59
%T KOMPARASI KINERJA NAIVE BAYES DAN LSTM DALAM ANALISIS SENTIMEN TERHADAP DANANTARA
%U http://repository.mercubuana.ac.id/96594/
%X This study aims to analyze the sentiment of the Indonesian public toward  Danantara, a national investment management agency, using the Naive Bayes and  Long Short-Term Memory (LSTM) algorithms. Data were collected from the social  media platform X (formerly Twitter) during the period from November 1, 2024, to  June 30, 2025, resulting in 2,562 tweets. After a manual relevance filtering process,  924 tweets were identified as directly related to the topic of Danantara. Sentiment  labeling was then carried out automatically using the IndoBERT model, classifying  the tweets into three categories: positive, negative, and neutral. The labeling results  showed a dominance of negative sentiment at 73.2%, followed by positive sentiment  at 26.7%, and neutral sentiment at 0.1%. The Naive Bayes and LSTM models were  subsequently used to compare sentiment classification performance on the labeled  tweets. Based on model performance evaluation, the Naive Bayes algorithm yielded  the best results with an accuracy of 78.91%, precision of 79.59%, recall of 78.91%,  and an F1-score of 74.87%. In comparison, the LSTM model achieved an accuracy  of 74.73%, precision of 73.53%, recall of 74.73%, and an F1-score of 73.98%.  Keywords: Danantara, Sentiment, Naive Bayes, LSTM, Twitter.    Penelitian ini bertujuan untuk menganalisis sentimen masyarakat Indonesia  terhadap Danantara, sebuah badan pengelola investasi nasional, dengan  menggunakan algoritma Naive Bayes dan Long Short-Term Memory (LSTM).  Data dikumpulkan melalui media sosial X (sebelumnya Twitter) selama periode 1  November 2024 hingga 30 Juni 2025, menghasilkan 2.562 tweet. Setelah dilakukan  tahap seleksi relevansi secara manual, diperoleh 924 tweet yang berkaitan langsung  dengan topik Danantara. Selanjutnya, proses pelabelan sentimen dilakukan secara  otomatis menggunakan model IndoBERT ke dalam tiga kategori: positif, negatif,  dan netral. Hasil pelabelan menunjukkan dominasi sentimen negatif sebesar 73,2%,  diikuti oleh sentimen positif sebesar 26,7%, dan netral sebesar 0,1%. Model Naive  Bayes dan LSTM kemudian digunakan untuk membandingkan kinerja klasifikasi  sentimen terhadap tweet yang telah dilabeli. Berdasarkan evaluasi performa model,  algoritma Naive Bayes menunjukkan hasil terbaik dengan akurasi sebesar 78,91%,  precision 79,59%, recall 78,91%, dan F1-score 74,87%. Sementara itu, model  LSTM mencatatkan akurasi 74,73%, precision 73,53%, recall 74,73%, dan F1-  score 73,98%.  Kata kunci: Danantara, Sentimen, Naive Bayes, LSTM, Twitter