Use of Spatio-temporal Features for Earthquake Forecasting of imbalanced Data

With improvement in instrumentation to precisely record seismic activities, the quality of seismic data is improving day by day, leading to more informative data sets. These data sets possess temporal and geospatial patterns that can be extracted by feature engineering of temporal and geospatial factors. However, the less frequent large-magnitude earthquakes often create an imbalance in earthquake data. In this study, we propose three machine learning-based algorithm-level techniques to transform time series earthquake data into an equivalent data set with temporal and geospatial features to treat the magnitude class imbalance. Results from several study regions including the Himalayas, Central Java, Sulawesi, Sumatra, and Southeast Asia are compared to discuss the efficacy of the proposed algorithms. Accuracy, precision, and F1 score are used as evaluation metrics. Therefore, the present work has provided a formulation to use machine learning algorithms for imbalanced data in earthquake forecasting.