Twitter Sentiment Analysis of Thai University Central Admission System (TCAS)

Abstract

The aim of this study was to analyze sentiments expressed in students' Twitter messages regarding the Thai University Central Admission System (TCAS) from 2018 to 2023, spanning six years. To accomplish this, we employed machine learning techniques, utilizing supervised learning for sentiment classification and message categorization. The research process encompassed four key stages. Firstly, we utilized web scraping techniques through the Apify platform to collect data by searching for keywords, both with and without the hashtag symbol (#). Secondly, data preparation procedures were conducted, involving tokenization, filtering, and the removal of stop words, ensuring that the text data was suitably prepared for analysis. Thirdly, we conducted modeling by constructing machine learning models using sentiment analysis techniques, and subsequently, the text data was classified using a supervised learning approach. The fourth step involved evaluation, where we integrated the prediction results into the model evaluation process to measure accuracy while considering the trade-off between accuracy and recall. The analysis findings demonstrate that the model exhibits robust predictive capability, as indicated by its high F1 score, which ranged from 0.60 to 0.98. Furthermore, an examination of student tweets concerning TCAS over the six-year period (2018-2023) reveals a prevailing negative sentiment. Specifically, out of the 2,145 messages analyzed, 2,098 messages (97.80%) conveyed negative sentiment, while 47 messages (2.20%) expressed positive sentiment.