Perbandingan Algoritma ETAS dan ETAS-FLP untuk Estimasi Model Space-Time Epidemic-Type Aftershock Sequence (Space-Time ETAS)

Rahman, Annisa Auliya (2024) Perbandingan Algoritma ETAS dan ETAS-FLP untuk Estimasi Model Space-Time Epidemic-Type Aftershock Sequence (Space-Time ETAS). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003221023-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Model space-time Epidemic-Type Aftershock Sequence (ETAS) adalah model populer untuk analisis gempa bumi. Model space-time ETAS memiliki fungsi kompleks, sehingga perlu metode numerik atau pendekatan spesifik untuk menyelesaikannya. Algoritma berbasis Forward Likelihood Predictive (FLP), yaitu ETAS-FLP, menggunakan metode numerik Newton dan pemilihan bandwidth berdasarkan informasi prediktif. Namun, kajian estimasi parameter dengan ETAS-FLP belum dibahas secara detail. Selain itu, terdapat algoritma ETAS yang dikembangkan berdasarkan metode numerik berbasis gradien (Davidon-Fletcher-Powell (DFP)) dan adaptive bandwidth berdasarkan jarak nearest neighbor. Kedua algoritma dikembangkan untuk model space-time ETAS, namun belum ada penelitian yang membandingkan keduanya secara khusus, sehingga sulit menentukan preferensi penggunaannya. Penelitian ini mengkaji tahapan estimasi parameter model space-time ETAS dengan algoritma ETAS-FLP serta membandingkan performanya dengan algoritma ETAS. Estimasi parameter terdiri dari dua bagian, yaitu: estimasi parametrik dengan Maximum Likelihood Estimation (MLE) serta metode numerik Newton, dan estimasi nonparametrik menggunakan estimator kernel dengan dua fixed bandwidth yang dipilih berdasarkan pendekatan FLP. Estimasi tersebut dilakukan secara iteratif hingga konvergen. Perbandingan performa kedua algoritma dilakukan pada studi terapan dengan beberapa skenario menggunakan katalog gempa bumi Sumatra. Hasil penelitian menunjukkan bahwa algoritma ETAS dan ETAS-FLP memiliki nilai Akaike Information Criterion (AIC) yang sebanding. Sementara itu, algoritma ETAS-FLP lebih unggul dalam hal waktu komputasi dibanding algoritma ETAS, dengan waktu komputasi 4,96% dan 18,65% lebih cepat. Algoritma ETAS-FLP lebih disarankan untuk aplikasi pada dataset yang besar. Kasus gempa bumi Sumatra menunjukkan bahwa efek temporal lebih dominan dalam memicu gempa susulan dibandingkan efek spasial. Intensitas gempa tertinggi di sekitar Pulau Nias dan Kepulauan Mentawai.
============================================================
The space-time Epidemic-Type Aftershock Sequence (ETAS) model is a widely used for earthquake catalog analysis. Due to its complex functions, a numerical method or specific approach is required. The Forward Likelihood Predictive (FLP) based algorithm, ETAS-FLP, employs the standard Newton-type numerical method and bandwidth selection based on predictive information. Nevertheless, parameter estimation studies with ETAS-FLP have not been discussed in detail. Furthermore, ETAS algorithm has been developed based on gradient-based numerical method (Davidon-Fletcher-Powell (DFP)) and adaptive bandwidth based on nearest neighbour distance. Although both algorithms were developed for the ETAS space-time model, no research has specifically compared the two algorithms, making it difficult to determine the preference for their use. Therefore, this study will examine the parameter estimation stages of the ETAS space-time model using the ETAS-FLP algorithm and compare its performance with the ETAS algorithm. Parameter estimation is comprised of two parts: parametric estimation with Maximum Likelihood Estimation (MLE) and the standard Newton-type numerical method, and nonparametric estimation using a kernel estimator with two fixed bandwidths selected based on the FLP approach. The estimation is performed iteratively until convergence. A comparison of the two algorithms was conducted in an applied study with several scenarios using the Sumatra earthquake catalogue. The results demonstrate that the ETAS and ETAS-FLP algorithms exhibit comparable performance in terms of Akaike Information Criterion (AIC). The ETAS-FLP algorithm demonstrates superior performance in terms of computation time compared to the traditional ETAS algorithm, showing a 4.96% to 18.65% reduction in computation time. The ETAS-FLP algorithm is preferable for applications on large datasets. Sumatra earthquakes show that temporal effects are more dominant in triggering aftershocks than spatial effects. The highest earthquake intensity is observed around Nias Island and the Mentawai Islands.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Earthquake, Forward Likelihood Predictive, Natural disasters, Space-time ETAS, Bencana alam, Forward Likelihood Predictive, Gempa bumi, Space-time ETAS.
Subjects:	H Social Sciences > HA Statistics > HA30.6 Spatial analysis
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Annisa Auliya Rahman
Date Deposited:	07 Aug 2024 22:34
Last Modified:	07 Aug 2024 22:34
URI:	http://repository.its.ac.id/id/eprint/114639

Actions (login required)

View Item