Perancangan Generator Data Sintetis Berbasis Model Block Bootstrap Untuk Augmentasi Data Prekursor Gempa Bumi

Ramadhan, Edo Dwi Yusriza (2026) Perancangan Generator Data Sintetis Berbasis Model Block Bootstrap Untuk Augmentasi Data Prekursor Gempa Bumi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009201159-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Data prekursor gempa bumi yang terbatas yang memiliki karakteristik deret waktu yang saling bergantung menjadi tantangan dari pengembangan model prediksi gempa. Penelitian ini memiliki tujuan untuk merancang generator data sintetis berbasis metode block bootstrap sebagai teknik augmentasi data untuk data prekursor gempa bumi guna menanggulangi keterbatasan data yang sangat minim. Metode yang digunakan meliputi Moving Block Bootstrap (MBB), Circular Block Bootstrap (CBB), dan Stationary Block Bootstrap (SBB) yang diaplikasikan pada data geomagnetik komponen H, D, dan Z dari stasiun di Cilacap (CLP), Sukabumi (SKB), dan Trestes (TRT) yang ada di Pulau Jawa. Data dinormalisasi menggunakan metode z-score sebelum dilakukan pembangkitan data sintetis. Validasi dilakukan menggunakan koefisien korelasi Pearson, nilai root mean square error (RMSE), serta uji Kolmogorov-Smirnov untuk mengukur kemiripan data sintesis terhadap data asli. Hasil penelitian menunjukkan bahwa seluruh metode belum mampu menghasilkan data sintetis dengan karakteristik mendekati data asli, yang mana ditunjukan pada skor indeks komposit yang dihasilkan belum mencapai ambang batas validasi ≥0,6. Performa metode juga bervariasi tergantung stasiun: MBB menunjukkan kinerja terbaik di stasiun CLP (skor 0.389) dan TRT (skor 0.3925), sedangkan SKB (skor 0.3882) paling sesuai untuk stasiun SKB. Meskipun demikian, penelitian ini berhasil merancang generator data sintetis berbasis block bootstrap yang berpotensi dikembangkan lebih lanjut untuk augmentasi data prekursor gempa bumi, dengan catatan perlu optimasi parameter dan penambahan data yang lebih representatif.
=================================================================================================================================
Limited earthquake precursor data with interdependent time series characteristics pose a challenge for earthquake prediction model development. This study aims to design a synthetic data generator based on the block bootstrap method as a data augmentation technique for earthquake precursor data to overcome the limitations of extremely minimal data. The methods used include Moving Block Bootstrap (MBB), Circular Block Bootstrap (CBB), and Stationary Block Bootstrap (SBB), which are applied to geomagnetic data components H, D, and Z from stations in Cilacap (CLP), Sukabumi (SKB), and Trestes (TRT) on the island of Java. The data was normalized using the z-score method before synthetic data generation. Validation was performed using Pearson's correlation coefficient, root mean square error (RMSE) values, and the Kolmogorov-Smirnov test to measure the similarity of the synthetic data to the original data. The results showed that none of the methods were able to produce synthetic data with characteristics close to the original data, as indicated by the composite index scores that did not reach the validation threshold of ≥0.6. The performance of the methods also varied depending on the station: MBB showed the best performance at the CLP station (score of 0.389) and TRT station (score of 0.3925), while SKB (score of 0.3882) was most suitable for the SKB station. Nevertheless, this study successfully designed a block bootstrap-based synthetic data generator that has the potential to be further developed for earthquake precursor data augmentation, with the caveat that parameter optimization and the addition of more representative data are needed.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Block Bootstrap, Data Sintetis, Gempa Bumi, Prekursor Gempa,Bootstrap Block, Synthetic Data, Earthquake, Earthquake Precursor
Subjects:	Q Science Q Science > QC Physics Q Science > QE Geology Q Science > QE Geology > QE1.F557 Magnetic anomalies--Measurement.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Edo Dwi Yusriza Ramadhan
Date Deposited:	04 Feb 2026 06:55
Last Modified:	04 Feb 2026 06:55
URI:	http://repository.its.ac.id/id/eprint/132047

Actions (login required)

View Item