Leveraging Social Media Streams for Disaster Prediction: A Machine and Deep Learning Approach

Natural disasters have an impact on the lives of people and animals worldwide. Twitter, which is a social networking tool that enables individuals to share news, opinions, and even their experiences, has emerged as a valuable source of real-time information. Since real-time data is easily accessed, many service providers make an effort to analyze it on a daily basis for purposes of detecting emergencies and lowering risk factors. As for people, a person cannot sit and analyze thousands of records and, at the same time, recognize threats in real time. To address this challenge, machine learning (ML) and deep learning (DL) techniques have become essential tools for automating disaster detection and classification. The article examines the disaster prediction capabilities of Twitter text data and evaluates the power of diverse supervised machine learning models (logistic regression, naive Bayes, support vector machines (SVM), k-nearest neighbors (KNN) and XGBoost). Also, a deep learning model based on Long Short-Term Memory (LSTM) is tested to see how it performs on sequential data in tweets. The main goal of the work is investigating the impact of the usage of different data modes, including text, visual data (including pictures and satellite data), audio data (including sounds of emergency calls or acoustic patterns), and blog or news posts, on effective disaster prediction and classification. Although the nature of Twitter text data (real-time, short, and publicly available) has been prioritized in this research, the property of multimodal data sources to improve situational awareness is also recognized as their increasing role. Through the analysis of these varied streams of information with both a traditional machine learning approach and an advanced deep learning approach, the comprehensive study will attempt to create a more holistic and responsive disaster detection framework that could serve as an aid to decision-makers and emergency responders in their attempt to mitigate losses and improve the response strategy.