Utilising sentiment analysis of COVID-19 tweets through explainable artificial intelligence to derive business insights

The COVID-19 pandemic has undeniably transformed societies across the globe in profound ways particularly in the Arab region. This extensive transformation has prompted a range of essential control measures that have sought to mitigate the spread of the virus. Among these measures are lockdowns, curfews, and travel restrictions, which have collectively shaped daily life and public interaction. A dataset of approximately 264,000 tweets was collected using keywords such as #COVID-19, #coronavirus, and #lockdown. After rigorous preprocessing, the data was represented using TFIDF vectorisation. Three machine learning algorithms including random forest, logistic regression, and support vector machine were employed for sentiment classification, with model interpretability enhanced via LIME and SHAP. Logistic regression, combined with explainable AI (XAI) and TF-IDF, achieved the highest accuracy of 82% compared to other algorithms. This research highlights the potential of sentiment analysis in informing business and entrepreneurship strategies during times of the pandemic crisis.