COMPARING MACHINE LEARNING MODELS FOR E-LEARNING SENTIMENT ANALYSIS

Abo-Rashed Knaan, Polshchikov Konstantin Alexandrovich, Golovko Iaroslav Iurevich

Sentiment analysis plays a crucial role in understanding student experiences in e-learning environments. This study examines the comparative effectiveness of four distinct machine learning models - Naive Bayes, Support Vector Machines, Decision Trees, and Random Forests - in sentiment analysis of 3000 tweets related to e-learning. The models' evaluations are based on parameters such as accuracy, recall, specificity, and the F1 score. The ideal model selection should balance factors like computational resources, interpretability, and adaptability. These results offer valuable insights for e-learning platform administrators and teachers, potentially guiding improvements in course content, its delivery, and user experience. The study acknowledges potential limitations related to model bias and the complexity of generalizing conclusions to other e-learning platforms. Future research directions include mitigating bias and enhancing the applicability of these models to various platforms and data sources. This work contributes significantly to the evolving field of applying machine learning in sentiment analysis in e-learning.

Keywords:Sentiment Analysis, E-learning, Machine Learning, Naive Bayes, Support Vector Machines, Decision Trees, Random Forest