Abstract
To explore the application of neural networks to estimate geomagnetic field disturbances, this study pays particular attention to K-index classification. The primary goal remains to provide a robust and efficient method for classifying the different levels of geomagnetic activity via neural networks. Data preprocessing, model architecture optimization, and classification performance evaluation are involved in this study at large. It proposes a new neural network method that attends to specific complexities of geomagnetic data and compares its merits against conventional methods. Neural networks, Long Short-Term Memory (LSTM) models in particular, revealed much better performance than traditional techniques with attained classification accuracy reaching 98%. The findings suggest that there is a promise for the applications of neural networks within geomagnetic studies and this will pave the way towards Artificial Intelligence models for forecasting or for the further inclusion of this approach to its articulation in discussions surrounding space weather research.