Analysis of Multi-Class Weather Classification using deep learning models and machine learning classifiers

Abstract

Extreme weather detection in huge datasets is a difficult task for researchers studying climate change. Current algorithms for detecting severe weather are reliant on human experience in classifying occurrences using arbitrary physical thresholds. On the same dataset, numerous competing approaches frequently yield wildly dissimilar findings. Understanding the trends and potential effects of such weather conditions depends on accurate categorization of severe events in climate simulations and observational data archives. In this paper, deep learning techniques are used as an alternate tool for identifying extreme weather occurrences. From labelled data, deep neural networks can develop high-level representations of a wide range of patterns. In this work, we have created a deep convolutional neural network (CNN) classification system. Our deep CNN system detects extreme events with VGG19 model and logistic regression classifier with 98.5% accuracy.