Sentiment Classification of Student Opinions on AI Utilization Using Naive Bayes Algorithm

Abstract

The rapid advancement of technology has encouraged the use of Artificial Intelligence in various sectors, including higher education. Among students, Artificial Intelligence was used to support learning activities, although it posed ethical concerns such as technological dependency and the risk of plagiarism. This study aimed to examine students’ sentiment toward the use of Artificial Intelligence in academic contexts. Data were collected through an online survey with a total of 498 responses, consisting of 347 positive sentiments and 151 negative sentiments. The classification process employed the Multinomial Naïve Bayes (MNB) algorithm using training data proportions of 60%, 70%, 80%, and 90%, along with combinations of unigram, bigram, and trigram features. The highest accuracy of 0.84 was obtained using 90% training data with the combination of all three features. As a comparison, the Random Forest algorithm achieved its highest accuracy of 0.86 using unigram features with the same training proportion. The results showed that both algorithms performed well in sentiment classification, with Random Forest slightly outperforming. Furthermore, the findings revealed variations in students’ adherence to academic ethics regarding the use of Artificial Intelligence.