Multiple Agents based Disaster Prediction for Public Environments using Data Mining Techniques

Abstract

Real-time data on natural disasters are collected, explained, analysed, predicted, and shown in the disaster management system. The development of GIS-based informational understanding has been documented (GIS). Using GIS and geographic data mining, the disaster management approach can pinpoint the epicentre of an occurrence and direct relief workers along the safest possible paths to the scene. The precise geological state and geographical placement of many areas makes them vulnerable to a wide range of natural disasters, including earthquakes, floods, land debris, landslides, cloud bursts, and human casualties. An efficient real-time system for predicting natural occurrences and locations is necessary to minimise damages and suffering. This research presents a unique methodology for predicting the location of disasters using density-based spatiotemporal clustering and global positioning system data. Before implementing clustering and feature selection, the process of data cleansing removes redundant, irrelevant, and inconsistent information from the news databases based on natural events. Areas prone to natural disasters like earthquakes, floods, landslides, and so on will be culled using a spatiotemporal clustering technique. The clustered data is then sorted by terms associated with natural catastrophes, and features are selected accordingly. In order to aid event detectors and location estimators, extracted features are supplied to a decision tree, which then categorises the data into both positive and negative classes.