Sentiment Analysis of X Application Users on Bitcoin Using the Naïve Bayes Method Optimized with Particle Swarm Optimization (PSO)

Abstract

Advancements in technology and social media have significantly transformed the way individuals express their opinions—one of which is toward decentralized digital currencies that utilize blockchain technology to enable peer-to-peer transactions, such as Bitcoin. This study aims to evaluate user sentiment toward Bitcoin by implementing the Naïve Bayes method optimized with Particle Swarm Optimization (PSO), using data gathered from the X application (formerly Twitter). The data were collected through web scraping of user posts containing the keyword “Bitcoin.” Text preprocessing was performed to enhance data quality, followed by feature extraction using the Term Frequency–Inverse Document Frequency (TF-IDF) approach to convert textual data into numerical representations. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. Initial results show that the Naïve Bayes classifier performs well in sentiment classification. The integration of PSO as an optimization method improved classification performance from 66.14% to 69.14%. This study contributes to a deeper understanding of public opinion on Bitcoin and demonstrates the effectiveness of combining Naïve Bayes and PSO in text-based sentiment analysis.

Keywords

• sentiment analysis
• bitcoin
• naïve bayes
• tf-idf
• pso