Social Network Sentiment Analysis for Natural Disaster Response

Social media platforms, such as X/Twitter and Facebook, have become vital tools for disseminating time-sensitive information during natural disasters. These platforms provide real-time data that can assist governments and related agencies in managing disaster response more effectively. This study evaluates the use of social media data for disaster response by conducting sentiment analysis using traditional machine learning models and deep learning models. The analysis focuses on classifying sentiments related to natural disasters using an existing dataset of tweets. We compare traditional machine learning algorithms, such as Logistic Regression (LR), Support Vector Machines (SVM), Random Forests (RF) and Stochastic Gradient Descent (SGD), with deep learning models, mainly transformer-based models, such as BERT, BigBird, ALBERT, and ELECTRA, in the context of sentiment classification. The results indicate that the transformer-based ALBERT model achieves the highest accuracy, with an average weighted F1 score of 89%. This work provides benchmark results for sentiment classification models and demonstrates their potential in improving decision-making for disaster response efforts.