Can social media predict demand in humanitarian crises? A case study of the 2023 Türkiye earthquake

Abstract

During sudden onset disasters, the main objective of humanitarian supply chains is to efficiently attend to the immediate needs and demands of the affected people. One of the main challenges on their way is the accurate estimation and prediction of demand, especially when communication with the affected areas is limited due to the critical situation. In recent years, social networks have become crucial communication channels during disasters, particularly for real-time access to information. This study explores the role of social media, specifically platform X, in improving the efficiency of humanitarian supply chains by bridging the gap between the supply and demand of relief items. We aim to extract and analyze the spatial distribution of demand for relief supplies, as posted on platform X during the events following the February 6th, 2023, Türkiye Earthquakes, and to compare these findings with reports from traditional news channels. We propose a novel framework that leverages machine learning approaches such as BERTopic to extract key demand categories and named entity recognition (NER) to identify the geographical locations of expressed demand in X posts. By combining these techniques, the research seeks to offer a solution to improve the coordination and delivery of relief supplies in disaster-stricken areas, enhancing the overall responsiveness of humanitarian efforts. By comparing the extracted needs from platform X with official government announcements and traditional media communications, our findings show that social media plays a critical role in informing individual donors about the evolving needs of disaster victims.