SENTIMENT ANALYSIS ON INDONESIAN SOCIAL MEDIA USING INTEGRATION OF INSET-LEXICON AND MACHINE LEARNING

This thesis explores sentiment analysis on Indonesian social media, with a focus on Twitter, to gauge public opinions on the relocation of Indonesia's capital (IKN). Indonesia's diverse linguistic landscape and the frequent use of informal language on social media platforms present considerable challenges for traditional sentiment analysis techniques. To address these challenges, the research combines the InSet lexicon, which is specifically tailored for the Indonesian language, with advanced machine learning models, including Support Vector Machine (SVM), Random Forest, and deep learning models like BERT, IndoBERT, and IndoBERTweet.
The findings indicate that integrating lexicon-based approaches with machine learning significantly improves the accuracy and reliability of sentiment analysis. Among the models tested, BERT-based models, especially IndoBERTweet, exhibited the highest accuracy. This highlights the effectiveness of transformer-based models in understanding and interpreting the nuanced sentiments expressed in Indonesian social media content. The research contributes to the advancement of sentiment analysis methodologies and provides practical insights for effectively monitoring public sentiment on critical national issues such as the IKN relocation.
In summary, the study demonstrates the potential of combining lexicon-based methods with state-of-the-art machine learning models to enhance the quality of sentiment analysis, particularly in the context of Indonesia's complex linguistic environment. This approach not only improves the analytical accuracy but also offers valuable tools for policymakers and stakeholders to better understand public sentiment, thereby aiding in more informed decision-making processes.

Keywords: Sentiment Analysis, Lexicon, Machine Learning, IKN (The New Capital of Indonesia)