DETEKSI BANJIR BERBASIS LLM (LARGE LANGUAGE MODELS) MENGGUNAKAN DATA TWITTER/X VIA CHATBOT WHATSAPP

Abstract

This research focuses on developing an innovative artificial intelligence-based flood detection system, leveraging the power of Large Language Models (LLMs) and real-time data from the Twitter/X social media platform. Motivated by the need for faster and more effective early warning systems compared to conventional methods, this study aims to integrate Twitter/X data analysis—where communities often report flood incidents directly—with the advanced natural language understanding capabilities of LLMs. Through stages such as Twitter/X data collection and pre-processing, feature extraction using the TF-IDF algorithm, development of an LLM model for flood tweet classification, and the construction of a WhatsApp chatbot, this system is designed for automatic flood information detection. The targeted output is a prototype flood detection system via a WhatsApp chatbot capable of providing accurate and rapid notifications to users. Key contributions of this research include technological innovation by combining LLMs and TF-IDF, enhancing the speed of flood information response, and improving information accessibility through the popular WhatsApp platform. Collectively, these advancements hold significant potential in supporting disaster mitigation efforts and reducing flood-related losses in Indonesia. The overall accuracy obtained was 96%, which is very good in classifying flood tweets