Seismic Precursor Data Completion Using Bayesian Augmented Tensor Factorization

Seismic precursor data is crucial for understanding and monitoring earthquake phenomena. However, due to limitations in observational capabilities and data noise, such data tendd to be incomplete, requiring careful preprocessing before analysis. As a type of multi-dimensional spatiotemporal data, imputing seismic precursor data is of great significance. This paper utilizes a seismic precursor data completion method based on Bayesian Augmented Tensor Factorization (BATF). By integrating temporal correlations and probabilistic modeling, the method effectively reconstructs missing data and quantifies the uncertainty of predictions. Validation on real-world seismic records shows that BATF outperforms traditional interpolation methods and tensor factorization techniques. The findings indicate that BATF is an effective tool for enhancing the reliability and interpretability of seismic precursor datasets, offering new perspectives for improving earthquake prediction models.