PENERAPAN ALGORITMA CERDAR BIDIRECTIONAL ENCODER REPRESENTATIONS FORM TRANSFORMERS (BERT) DALAM MENGANALISIS OPINI PUBLIK TERHADAP PRODUK YANG MENGALAMI BOIKOT

In this modern era, social media has become one of the primary platforms for people to express opinions and participate in public discussions. Instagram, as one of the most popular social media platforms in the world, serves as a medium to express views and opinions on various social, economic, and political issues. One of the topics that has gained traction on social media is product boycotting. Boycotting a product can significantly impact brand image and sales. Renowned brands such as McDonald’s, KFC, Starbucks, Burger King, and Pizza Hut have certain characteristics that make them prime targets for boycott actions. This research utilizes 1,750 datasets derived from comments on the Instagram accounts of these related products. The data is divided into two labels, Positive and Negative, based on automatic labeling from transformers and manual labeling. Sentiment analysis results show that McDonald’s has 41.43% Positive sentiment and 58.57% Negative, KFC with 85.14% Positive and 14.86% Negative, Starbucks with 97.71% Positive and 2.29% Negative, Burger King with 50% Positive and Negative, and Pizza Hut with 80.57% Positive and 19.43% Negative. From the modeling results using the pre-trained Bidirectional Encoder Representation From Transformers (BERT) method from Bert-Base-Uncased, the accuracy results obtained for McDonald’s products are 84.14%, KFC products 95%, Starbucks products 94.16%, Burger King products 91.42%, and Pizza Hut products 93.80%.

Keywords: McDonald’s, KFC, Starbucks, Burger King, Pizza Hut, Bidirectional Encoder Representations From Transformers (BERT), Bert-Base-Uncased, Sentiment Analysis