Comparative Analysis of Naive Bayes and GPT-2 for Twitter-Based Sentiment Analysis in Bank Reputation Management
Year: 2024
Type: article
Abstract: This study employs sentiment analysis to assess bank reputation during system failures, using Twitter data from 959 tweets collected between 2023 and 2024.We compare the performance of Naive Bayes and GPT-2 algorithms in analyzing customer complaints about banking application issues.The research focuses on three variables: problem features, initial customer sentiment, and sentiment after bank response.Naive Bayes achieved accuracies of 76.75%, 62.23%, and 87.78% for these variables respectively, while GPT-2 recorded 82.02%, 67.33%, and 61.39%.The study reveals that GPT-2 outperforms Naive Bayes in initial sentiment classification, likely due to its superior ability to understand context.These findings provide valuable insights for banks to improve their reputation management strategies and customer service responses during system failures.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4407163469