Comparison of Deep Learning Methods and Model Proposal for Earthquake Intensity Estimation

Rabia Uyar; Durmuş Özdemir

Makale detayları

Comparison of Deep Learning Methods and Model Proposal for Earthquake Intensity Estimation

Yazarlar:

Rabia Uyar | Durmuş Özdemir

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Ampirik Yazılım Mühendisliği | Bilgisayar Sistem Yazılımı | Bilgisayar Yazılımı | Yazılım Mimarisi | Jeoloji (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.25 Issue 3, pp.522-534

Belge Tipi:

Makale

Yayın Tarihi:

10/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001348287

pISSN:

2149-3367

Anahtar kelimeler:

Özet:

 Earthquake is one of the natural disasters that shows that the soil, which is considered immobile by people, will also move and that all the structures on it can be damaged and collapsed with the loss of life. Earthquake disasters cause serious losses in many countries, including our country, because of the fundamental problems in urbanization and construction, due to the lack of adequate precautions. To prevent such losses, researchers carry out studies to predict the earthquake intensity and the place where the earthquake occurred. To predict the time and intensity of the earthquake, useful earthquake prediction studies are carried out by analyzing the data set created by combining data such as the date, time, depth, latitude, longitude, and intensity of previous earthquakes. In this context, it is aimed to estimate the intensities of possible earthquakes based on only the earthquake intensity data of the large and small earthquakes that occurred in our country between the years 1965-2022. In line with these determined purposes, two different estimation models, namely Earthquake Intensity Prediction Model with LSTM Long short-term memory and Earthquake Intensity Prediction Model with ANN, were developed by using earthquake catalog data obtained from the United States Geological Survey Board to predict the intensity of earthquakes. The prediction results of these two proposed models are compared with both the real earthquake data and the prediction results of the Automatic Regressive Integrated Moving Average ARIMA model, and the success of the developed models is presented.

 Earthquake is one of the natural disasters that shows that the soil, which is considered immobile by people, will also move and that all the structures on it can be damaged and collapsed with the loss of life. Earthquake disasters cause serious losses in many countries, including our country, because of the fundamental problems in urbanization and construction, due to the lack of adequate precautions. To prevent such losses, researchers carry out studies to predict the earthquake intensity and the place where the earthquake occurred. To predict the time and intensity of the earthquake, useful earthquake prediction studies are carried out by analyzing the data set created by combining data such as the date, time, depth, latitude, longitude, and intensity of previous earthquakes. In this context, it is aimed to estimate the intensities of possible earthquakes based on only the earthquake intensity data of the large and small earthquakes that occurred in our country between the years 1965-2022. In line with these determined purposes, two different estimation models, namely Earthquake Intensity Prediction Model with LSTM Long short-term memory and Earthquake Intensity Prediction Model with ANN, were developed by using earthquake catalog data obtained from the United States Geological Survey Board to predict the intensity of earthquakes. The prediction results of these two proposed models are compared with both the real earthquake data and the prediction results of the Automatic Regressive Integrated Moving Average ARIMA model, and the success of the developed models is presented.

 Earthquake is one of the natural disasters that shows that the soil, which is considered immobile by people, will also move and that all the structures on it can be damaged and collapsed with the loss of life. Earthquake disasters cause serious losses in many countries, including our country, because of the fundamental problems in urbanization and construction, due to the lack of adequate precautions. To prevent such losses, researchers carry out studies to predict the earthquake intensity and the place where the earthquake occurred. To predict the time and intensity of the earthquake, useful earthquake prediction studies are carried out by analyzing the data set created by combining data such as the date, time, depth, latitude, longitude, and intensity of previous earthquakes. In this context, it is aimed to estimate the intensities of possible earthquakes based on only the earthquake intensity data of the large and small earthquakes that occurred in our country between the years 1965-2022. In line with these determined purposes, two different estimation models, namely Earthquake Intensity Prediction Model with LSTM Long short-term memory and Earthquake Intensity Prediction Model with ANN, were developed by using earthquake catalog data obtained from the United States Geological Survey Board to predict the intensity of earthquakes. The prediction results of these two proposed models are compared with both the real earthquake data and the prediction results of the Automatic Regressive Integrated Moving Average ARIMA model, and the success of the developed models is presented.

 Deprem, tektonik kuvvetlerin veya volkan faaliyetlerinin etkisiyle yer kabuğunun kırılması sonucunda ortaya çıkan enerjinin sismik dalgalar hâlinde yayılarak geçtikleri ortamları ve yeryüzünü kuvvetle sarsması olayıdır. Deprem afeti ülkemizde dahil olmak üzere birçok ülkede şehirleşme ve yapılaşmadaki temel sorunlardan dolayı yeterince önlem alınamadığı için ciddi ölçüde kayıplar meydana getirmektedir. Bu tür kayıpların önüne geçebilmek için araştırmacılar, deprem şiddeti ve depremin meydana geldiği yeri önceden tahmin etmek için çalışmalar yürütmektedirler. Depremin meydana geleceği zamanı ve şiddetini tahmin edebilmek için, önceden meydana gelen depremlerin oluştuğu tarih, zaman, derinlik, enlem, boylam ve şiddet gibi verilerin bir araya getirilerek oluşturulan veri setinin çok iyi analiz edilerek faydalı deprem tahmin çalışmaları yürütülmektedir. Bu kapsamda, 1965-2022 yılları arasında ülkemizde meydana gelen büyük ve küçük şiddetteki depremlerin sadece deprem şiddeti verileri esas alınarak, meydana gelmesi muhtemel depremlerin şiddetlerinin tahmin edilmesi amaçlanmıştır. Belirlenen bu amaçlar doğrultusunda, depremlerin şiddetini tahmin edebilmek için, Amerika Birleşik Devletleri Jeolojik Araştırma Kurulundan elde edilen deprem katalog verileri kullanılarak, LSTM Uzun-Kısa Vadeli Bellek ile Deprem Şiddeti Tahmin Modeli ve YSA ile Deprem Şiddeti Tahmin Modeli olmak üzere iki farklı tahmin modeli geliştirilmiştir. Önerilen bu iki modelin tahmin sonuçları hem gerçek deprem verileriyle hem de Otomatik Regresif Entegre Hareketli Ortalama ARIMA modelinin tahmin sonuçları ile karşılaştırılmıştır. Önerilen iki modelden LSTM ile DŞTM Deprem Şiddeti Tahmin Modeli; 0,561 R-kare değeri ile, YSA ile DŞTMne 0,3 R-kare göre daha yüksek bir başarı göstermiştir.

Makale detayları

Comparison of Deep Learning Methods and Model Proposal for Earthquake Intensity Estimation

Yazarlar:

Rabia Uyar | Durmuş Özdemir

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Ampirik Yazılım Mühendisliği | Bilgisayar Sistem Yazılımı | Bilgisayar Yazılımı | Yazılım Mimarisi | Jeoloji (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.25 Issue 3, pp.522-534

Belge Tipi:

Makale

Yayın Tarihi:

10/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001348287

pISSN:

2149-3367

Deep learning

Artificial neural networks

ARIMA

LSTMDerin öğrenme

Yapay sinir ağları

ARIMA

LSTM

Sayı

Makale detayları

Yazarlar: Rabia Uyar | Durmuş Özdemir

Yayıncı: Afyon Kocatepe University

Ana Başlıklar: Ampirik Yazılım Mühendisliği | Bilgisayar Sistem Yazılımı | Bilgisayar Yazılımı | Yazılım Mimarisi | Jeoloji (Diğer)

Dergi: Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.25 Issue 3, pp.522-534

Belge Tipi: Makale

Yayın Tarihi: 10/2025

Belge Dili: Türkçe

Document ID (Org No.): 2001348287

pISSN: 2149-3367

Sayı

Yazarlar:

Rabia Uyar | Durmuş Özdemir

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Ampirik Yazılım Mühendisliği | Bilgisayar Sistem Yazılımı | Bilgisayar Yazılımı | Yazılım Mimarisi | Jeoloji (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi Fen Ve Mühendislik Bilimleri Dergisi Vol.25 Issue 3, pp.522-534

Belge Tipi:

Makale

Yayın Tarihi:

10/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001348287

pISSN:

2149-3367