EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EARTHQUAKE FORECASTING MODEL IN AZERBAIJAN BASED ON A NEURAL NETWORK

Azad Bayramov

Modeling Control and Information Technologies · 11.11.2026 · (9) · 20

DOI: 10.31713/mcit.2026.002 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Modeling Control and Information Technologies, online erschienen am 11.11.2026. Verfasst von Azad Bayramov. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The paper presents a mathematical model for binary classification—"earthquake" vs. "no earthquake"—designed to estimate the probability of an earthquake occurring in Azerbaijan. The model is based on a fully connected neural network (multilayer perceptron, MLP). A sigmoid function strictly maps the network's output to the [0,1] range, mathematically allowing the result to be interpreted as the earthquake probability P. The work was conducted at the Republican Seismic Survey Center of the Azerbaijan National Academy of Sciences, based on data obtained from more than 80 seismic stations located across Azerbaijan.

Introduction

Earthquakes are among the most destructive natural disasters on Earth. These catastrophic events, capable of claiming hundreds of thousands of lives, occur in virtually all parts of the globe, including oceans and seas. Consequently, earthquake prediction and early warning constitute one of the primary challenges facing the science of seismology [1]. Numerous scientific studies have addressed this issue [2, 3, 4]. In recent years, artificial intelligence methods—specifically neural networks—have been increasingly employed to tackle the problem of earthquake prediction. [5,6]. This study attempts to model earthquake probability using a neural network. The work was conducted at the Republican Seismic Survey Center of the Azerbaijan National Academy of Sciences, based on data obtained from more than 80 seismic stations located across Azerbaijan [1,4]. Materials and Methods. Geophysical parameters. Mathematical method of neural networks. Matlab. Probability theory. Binary classification model. Sigmoid function. Structural diagram of the neural network. Weight matrices. MATLAB script for training a neural network. Seismic correlation prediction.

Results

A mathematical model based on a fully connected neural network (multilayer perceptron, MLP) has been developed and proposed. The number of neurons in the input layer corresponds to the dimensionality of vector x (9 neurons); two hidden layers with 32 and 16 neurons, respectively, are employed. The network is trained to minimize the binary cross-entropy function, resulting in the calculation of the earthquake probability P predicted by the network. Using MATLAB as the development environment, this type of neural network had written the code to train.

Conclusions

A sigmoid function strictly maps the network's output to the [0,1] range, mathematically allowing the result to be interpreted as the earthquake probability P. The work was conducted at the Republican Seismic Survey Center of the Azerbaijan National Academy of Sciences. Data collected from 80 seismic stations located within the territory of Azerbaijan will be used to perform calculations for estimating earthquake probability.

Zitieren

APA

Bayramov, A. (2026). EARTHQUAKE FORECASTING MODEL IN AZERBAIJAN BASED ON A NEURAL NETWORK. Modeling Control and Information Technologies, 20. https://doi.org/10.31713/mcit.2026.002

Vancouver

Bayramov A. EARTHQUAKE FORECASTING MODEL IN AZERBAIJAN BASED ON A NEURAL NETWORK. Modeling Control and Information Technologies. 2026:20. doi:10.31713/mcit.2026.002

BibTeX
@article{Bayramov2026,
  author = {Bayramov, Azad},
  title = {EARTHQUAKE FORECASTING MODEL IN AZERBAIJAN BASED ON A NEURAL NETWORK},
  journal = {Modeling Control and Information Technologies},
  year = {2026},
  number = {9},
  pages = {20},
  doi = {10.31713/mcit.2026.002}
}
Kostenlos starten →