Enhancing Earthquake Prediction through Strategic Feature Selection and Machine Learning Algorithms
Dataset preservation is crucial in the field of machine learning used for earthquake detection and prediction. This work improves earthquake pre-diction and hazard prevention by doing a comprehensive analysis of seismic data using cutting-edge machine learning algorithms. An extensive dataset with a wide range of seismic parameters is assembled from many sources throughout several decades and geographic locations. The initial step involves comparing each dataset separately to find any shared characteristics. Furthermore, the approach employed feature selection approaches and entailed a thorough data preprocessing phase before dimensionality reduction using Principal Component Analysis (PCA). Following refinement, numerous machine learning models were trained using the dataset under varied preprocessing conditions. In terms of improving model performance, the results showed that feature selection performs noticeably better than PCA. The model with the highest accuracy and recall rates among the examined models was a KNN with specific salient feature selection. The proposed method highlights the crucial importance of meticulous data preprocessing, strategic feature selection, and appropriate model choice in developing high-performing predictive models. The research findings contribute significantly to earthquake prediction, offering a refined approach that improves prediction accuracy and supports disaster management and mitigation efforts globally.