Seismic activity prediction of the northern part of Pakistan from novel machine learning technique
| Autor: | Adeel Zafar, Bilal Aslam, Umer Khalil, Umar Azam |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
010504 meteorology & atmospheric sciences
Artificial neural network Computer science Earthquake prediction Particle swarm optimization Magnitude (mathematics) 010502 geochemistry & geophysics computer.software_genre 01 natural sciences Field (computer science) Support vector machine Hybrid neural network Geophysics Binary classification Geochemistry and Petrology Data mining computer Seismology 0105 earth and related environmental sciences |
| Zdroj: | Journal of Seismology. 25:639-652 |
| ISSN: | 1573-157X 1383-4649 |
| DOI: | 10.1007/s10950-021-09982-3 |
| Popis: | The prediction of the earthquake has been a testing investigation field, where a prediction of the impending incidence of destructive calamity is made. In this research, eight seismic features are processed by utilizing seismological notions, such as seismic quiescence, the eminent geophysical specifics of Gutenberg–Richter’s inverse law, and dissemination of typical earthquake extents for earthquake prediction. A classification system based on support vector regressor (SVR) along with hybrid neural network (HNN) is formed to attain the predictions of earthquakes for the Hindukush region. The challenge is expressed as a binary classification undertaking, and for earthquakes of magnitude equal to or more than 5.5, the predictions are generated for 1 month. HNN is a step-by-step amalgamation of three diverse neural networks, and enhanced particle swarm optimization (EPSO) is used to extend weight optimization at an individual layer, thus enhancing the performance of HNN. In amalgamation with the SVR-HNN prediction system, the freshly processed seismic aspects are applied to the Hindukush region. For analyzing the results, another considered performance measure is accuracy. Comparative to earlier prediction investigations, the achieved numerical outcomes demonstrate enhanced prediction implementation for the considered region. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink