Advancing Earthquake Prediction Accuracy Through Machine Learning: A GBM-Based Study on Diagnostic Performance and Reliability

In the study of seismic data, earthquake prediction has proved to be an important issue because of the sophisticated, non-linear, character of the information. In this paper, we discuss the effectiveness of Gradient Boosting Machines (GBM), a strong ensemble learning algorithm, in classifying earthquake events over a well-organized seismic data set. We used stratified data partitions to train and validate our model, which would result in the balancing and confirm appropriate representation of classes. This analysis was done by using various performance measurements and illustrations in order to carry out an effective evaluation. The Confusion matrix indicates a high level of classification of the model, with true positive number of 358 and true negative number of 288, and false negative 34, and false positive 85. The combined accuracy is at 84 percent implying a balanced sensitivity on the model with regard to earthquake and non-earthquake cases. Also, the precision of non-earthquake and earthquake are 0.89 and 0.81 respectively in the classification report whereas recall is high in earthquake cases (0.91) which means high sensitivity and less chance of misses. The high values of the F1-score (0.83 and 0.86, respectively) indicate overall high balance between the precision and recall in classes 0 and 1. The Discriminative strength of the model is further supported by Receiver Operating Characteristic (ROC) curve which indicates a good ability to discriminate between the 2 classes with Area under the curve (AUC) score of 0.90. In the learning curve results, training and validation scores improve steadily and converge at the end of the learning curve, and threshold analysis proves that the model is optimized by avoiding either false positive or false negative. This method which is based on GBM shows quite high level of accuracy and strength in earthquake prediction. Due to its capacity to run complicated seismic characteristics, it is being utilized as an invaluable instrument in disaster preparedness, in combination with early warning systems, which portrays a data-based approach to the kind of reliability that results in seismic event predictions.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.