Integrated support vector regressor and hybrid neural network techniques for earthquake prediction along Chaman fault, Baluchistan

Prediction of an extreme seismic event in any area has been a perplexing study area. Baluchistan is the largest province of Pakistan, and the whole province lies in a seismically active region which makes it vulnerable to earthquakes. In the current study, different aspects of seismology of Baluchistan, for instance, frequency content, recurrence time, Gutenberg-Richter law, seismic energy release, seismic rate changes, and other seismic features, are calculated. Furthermore, the Maximum Relevance and Minimum Redundancy (mRMR) principle is used to select the appropriate features. In order to predict the occurrence of future earthquakes, a Support Vector Regressor (SVR) and Hybrid Neural Network (HNN) centered prediction model (SVR-HNN) is formulated. An Enhanced Particle Swarm Optimization (EPSO) algorithm is incorporated in HNN for weight optimization at each layer. After testing the stability performance, the proposed model with systematically selected seismic features is applied for earthquake prediction along the Chaman fault in Baluchistan province. The accuracy of the prediction model is evaluated by computing prominent performance measures. The sensitivity and specificity for the earthquake prediction for the studied region were found to be 68.3% and 87.4%, respectively, with an accuracy of 81.2%. The accuracy of the prediction model is evaluated by computing prominent performance measures. The results of the study revealed a close relationship with the existing historical earthquake data as per the evaluation indices and can be considered acceptable. The proposed methodology can be equally useful for other areas for the prediction of a scenario earthquake. The results of the proposed methodology can help the decision and policymakers to plan cities and houses such that the damage can be minimized. Moreover, the decision-makers and future researchers can propose proper disaster mitigation strategies for averting the damage of future earthquakes.