Machine Learning-Based Approach to Analyze Sentiments on Moroccan Higher Education Through Twitter

In today's rapidly evolving world, assessing the quality of education systems has become important to meet the rising demand for high-quality education. Traditional evaluation methods like exams and surveys have limitations in providing comprehensive insights into system performance. To complement these approaches, sentiment analysis has emerged as a valuable tool, extracting sentiments from social media users, particularly on Twitter, using machine learning algorithms and natural language processing. This study focuses on sentiment analysis of the Moroccan higher education system, utilizing a dataset of 5608 English tweets collected between June 29, 2007, and July 07, 2023, through snscrape. We used a multinomial Naive Bayes classifier (MNB) and compared its performance with other models like decision tree, random forest, logistic regression, and support vector machine (SVM). The results reveal predominantly positive opinions regarding the Moroccan higher educational system, evident through positive sentiments across various key areas, including teaching, research and innovation, administration, partnerships, infrastructure, and overall quality. The multinomial Naive Bayes classifier achieves the highest accuracy of 93% among the models evaluated. These findings serve as a quality indicator, guiding educational institutions in aligning their curriculum and teaching methodologies with current goals and societal demands, while also identifying areas for improvement.