Artificial Intelligence-Powered Cyclone Classification Framework Using Mobilenetv1 and Goose Optimizer: Climate-Resilient Farming

Abstract

The intensifying frequency of cyclones owing to climate change presents a substantial threat to agricultural productivity, particularly in coastal and climate-sensitive regions. The present work proposes an artificial intelligence (AI)-powered framework for climate-resilient farming. The proposed AI framework consists of a pre-trained convolutional neural network (CNN), specifically the MobileNetV1 architecture. This architecture enables efficient feature extraction, making it well suited for real-time applications in resource-constrained agricultural systems. MobileNetV1 was fine-tuned using hyper-parameters, such as the number of epochs, learning rate, and batch size. A dataset of 1,600 satellite images was created with four distinct cyclone classes: tropical depression (class 1), tropical storm (class 2), severe tropical storm (class 3), and typhoon (class 4). The model's performance was evaluated for various optimizers, and the goose optimizer emerged as the most effective. By leveraging adaptive gradient adjustments, the goose optimizer enhances the training process, achieving a classification accuracy of 98.33% and an area under the curve (AUC) of 100%. These results were further validated using a confusion matrix and receiver operating characteristic (ROC) curves. In addition, the superior performance of the proposed AI-powered framework was compared with other pre-trained CNN models, such as VGG16, VGG19, and ResNet, for different optimizers. The proposed AI framework offers a promising solution for empowering farmers with predictive intelligence, thereby enhancing their resilience in cyclone-affected agricultural systems.