ANALISIS SENTIMEN GLOBAL TERHADAP TAGAR BRICS INDONESIA MENGGUNAKAN MODEL MULTINOMIAL NAIVE BAYES DENGAN LAPLACE SMOOTHING DAN SMOTE

ANANDA RIZKY SANDIKA, . (2025) ANALISIS SENTIMEN GLOBAL TERHADAP TAGAR BRICS INDONESIA MENGGUNAKAN MODEL MULTINOMIAL NAIVE BAYES DENGAN LAPLACE SMOOTHING DAN SMOTE. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text Cover.pdf Download (2MB)
	Text BAB I.pdf Download (539kB)
	Text BAB II.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB III.pdf Restricted to Registered users only Download (975kB) | Request a copy
	Text BAB IV.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text DAFTAR PUSTAKA.pdf Download (397kB)
	Text LAMPIRAN.pdf Restricted to Registered users only Download (2MB) | Request a copy
	Text BAB V.pdf Restricted to Registered users only Download (444kB) | Request a copy

Abstract

Penelitian ini bertujuan untuk mengevaluasi kinerja algoritma Multinomial Naïve Bayes dalam analisis sentimen terhadap topik #BRICS Indonesia, dengan membandingkan beberapa pendekatan pengujian. Empat metode diuji, yaitu: (1) Multinomial Naïve Bayes tanpa modifikasi, (2) Multinomial Naïve Bayes dengan Laplace smoothing, (3) Multinomial Naïve Bayes dengan Synthetic Minority Over-sampling Technique (SMOTE), dan (4) kombinasi antara Laplace smoothing dan SMOTE. Hasil pengujian menunjukkan bahwa metode Multinomial Naïve Bayes dengan Laplace smoothing memberikan performa terbaik, dengan accuracy 74%, precision 86%, recall 57%, dan f1-score 60%. Sementara itu, Multinomial Naïve Bayes tanpa modifikasi menghasilkan accuracy 51%, precision 56%, recall 56%, dan f1-score 48%. Metode dengan SMOTE menghasilkan accuracy 46%, precision 55%, recall 55%, dan f1-score 44%. Adapun kombinasi Laplace smoothing dan SMOTE memberikan accuracy 71%, precision 68%, recall 75%, dan f1-score 68%. Berdasarkan hasil tersebut, dapat disimpulkan bahwa penggunaan algoritma Multinomial Naïve Bayes dengan Laplace smoothing mampu memberikan peningkatan kinerja yang paling signifikan dalam klasifikasi sentimen menggunakan Multinomial Naïve Bayes. Temuan ini diharapkan dapat menjadi referensi bagi penelitian selanjutnya yang berkaitan dengan analisis sentimen dan penerapan algoritma pembelajaran mesin pada isu-isu global. ***** This research aims to evaluate the performance of the Multinomial Naïve Bayes algorithm in sentiment analysis on the topic of #BRICS Indonesia, by comparing several testing approaches. Four methods were tested: (1) Multinomial Naïve Bayes without modification, (2) Multinomial Naïve Bayes with Laplace smoothing, (3) Multinomial Naïve Bayes with Synthetic Minority Over-sampling Technique (SMOTE), and (4) a combination of Laplace smoothing and SMOTE. The test results show that the Multinomial Naïve Bayes method with Laplace smoothing provides the best performance, with an accuracy of 74%, precision of 86%, recall of 57%, and f1-score of 60%. Meanwhile, the Multinomial Naïve Bayes without modification yields an accuracy of 51%, precision of 56%, recall of 56%, and f1-score of 48%. The method with SMOTE results in an accuracy of 46%, precision of 55%, recall of 55%, and f1-score of 44%. The combination of Laplace smoothing and SMOTE provides an accuracy of 71%, precision of 68%, recall of 75%, and f1-score of 68%. Based on these results, it can be concluded that the use of the Multinomial Naïve Bayes algorithm with Laplace smoothing significantly improves performance in sentiment classification using Multinomial Naïve Bayes. These findings are expected to serve as a reference for future research related to sentiment analysis and the application of machine learning algorithms on global issues.

Item Type:	Thesis (Sarjana)
Additional Information:	1). Dr. Widodo, M.Kom. ; 2). Murien Nugraheni, M.Cs.
Subjects:	Sains > Matematika > Ilmu Komputer Sains > Statistika Teknologi dan Ilmu Terapan > Teknik Komputer
Divisions:	FT > S1 Pendidikan Teknik Informatika Komputer
Depositing User:	Users 28004 not found.
Date Deposited:	04 Aug 2025 07:44
Last Modified:	04 Aug 2025 07:44
URI:	http://repository.unj.ac.id/id/eprint/57664

Actions (login required)

View Item