Stormsense: A Multi-Model Cloudburst Prediction System Using ML And Deep Image Processing

Abstract

Multiple hazards exist due to cloudburst events be- cause of their abrupt surge and destructive nature. The prediction accuracy faces obstacles because of the multiple atmospheric variables that exist in the environment. The proposed system utilizes Random Forest (RF) and Support Vector Machine (SVM) for numerical meteorological data analysis while Convolutional Neural Networks (CNNs) using VGG16 performs satellite image processing for cloudburst prediction purposes. The system en- ables users to obtain automated weather predictions through a web-based application by either providing input data parameters or uploading satellite images. The system gives real-time alerts through the Pushbullet API to notify users when cloudburst predictions are detected. The combination of numerical inputs with satellite imagery analysis in the evaluation proved successful in developing a better and safer prediction system for disaster preparedness.