Analyzing the Perception of Immigrants in South Africa: A Machine Learning Approach to Aggregate Twitter Sentiment Data

In South Africa, home to over 4 million immigrants, capturing public sentiment towards immigration is crucial for policy and social cohesion. Employing BERT within machine learning, this study analyzed sentiments in 18,000 geotagged tweets, revealing predominant negative sentiments and regional disparities. Temporal analysis correlated sentiment shifts with significant events, while spatial analysis highlighted urban concentrations of xenophobic sentiment. Negative sentiments spiked during politically charged periods, contrasting with low positive sentiments. The study underscores the need for diverse data sources for comprehensive sentiment understanding, with implications for sectors like tourism. These insights exemplify computational social science's potential to inform immigration policy-making and societal harmony initiatives, affirming machine learning's role in nuanced public discourse analysis.