A Review on Earth Quake Prediction Using CapsNet

Earthquake prediction remains one of the most challenging tasks in geoscience owing to the nonlinear, chaotic nature of seismic signals. Recent advancements in Artificial Intelligence (AI) and Deep Learning (DL) have enabled researchers to extract meaningful patterns from large volumes of seismic waveform data. Traditional models, especially CNNs, often lose spatial information due to max-pooling. Capsule Networks (CapsNet), proposed by Sabour et al., preserve hierarchical spatial relationships through vector capsules and dynamic routing, making them suitable for complex seismic feature modelling. This review presents a detailed analysis of existing approaches, performance, limitations, and research gaps in CapsNet-based earthquake prediction. Additionally, the paper highlights various deep-learning-driven seismic detection methods and emphasizes the superiority of CapsNet in capturing spatial-temporal correlations crucial for early earthquake detection.