Klasifikasi Sentimen dan Pemodelan Topik MBKM Menggunakan Ensemble Learning dan Latent Dirichlet Allocation pada Media Sosial X

Abstract

Sentiment is an opinion or information conveyed by individuals to assess a particular topic or product. Sentiments are usually conveyed orally or in writing. Since the development of technology, many people have conveyed their sentiments or opinions through social media such as X, Instagram, Facebook, and others. One of the social media that is widely used to convey sentiments and ideas is X. The ease of conveying opinions through social media With so many opinions, we need a way to classify opinions quickly, such as using machine learning. MBKM is a policy of the Ministry of Education and Culture, Research and Technology of the Republic of Indonesia (Ministry of Education, Culture, Research and Technology of the Republic of Indonesia) which aims to support the preparation of student competencies to meet the needs of the times in line with the rapidly developing world of work and technology. Public opinions regarding the MBKM program published via social media X are very diverse, ranging from positive, negative and neutral opinions. These differences in opinion lead to the need for further research on MBKM sentiment. Then the researcher will conduct research on public opinion sentiment regarding the MBKM program using Ensemble Learning Voting with a combination of Random Forest, Naïve Bayes, and Support Vector Machine algorithms to determine the accuracy results of Ensemble Learning Voting and Latent Dirichlet Allocation to find various topics that arise based on opinions. This research obtained the results of the Merdeka Belajar Kampus Merdeka sentiment classification using the ensemble learning voting method which was proven to get better effectiveness results than a single model. Then in modeling MBKM sentiment topics based on keywords, the number of topics that were successfully identified from each keyword based on the type of sentiment was a maximum of 4 topics and a minimum of 2 topics