Peningkatan Performa XGBoost Menggunakan Metode Continuous Restricted Boltzmann Machine Dan Particle Swarm Optimization Pada Lapangan Poseidon, Cekungan Browse, Australia Untuk Prediksi Litologi

Elyandrie, Rijkaard (2025) Peningkatan Performa XGBoost Menggunakan Metode Continuous Restricted Boltzmann Machine Dan Particle Swarm Optimization Pada Lapangan Poseidon, Cekungan Browse, Australia Untuk Prediksi Litologi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017201027 - Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (2MB) | Request a copy

Abstract

Dalam tahap awal eksplorasi minyak dan gas bumi, memperoleh informasi litologi yang akurat merupakan langkah krusial dalam analisis geologi. Prediksi litologi menjadi salah satu fokus utama dalam penelitian geosains mengingat pentingnya pemetaan dan karakterisasi bawah permukaan, terutama untuk litologi. Algoritma Extreme Gradient Boosting (XGBoost) telah terbukti lebih unggul dalam pengenalan pola dibandingkan model konvensional, karena kemampuannya dalam menerapkan regularisasi, gradient boosting serta klasifikasi berbasis pohon keputusan. Model yang kompleks ini melibatkan banyak hyper-parameter sehingga sering kali menyebabkan kesulitan dalam optimasi serta inefisiensi dalam menangani variabel dengan dimensi tinggi. Untuk meningkatkan performa prediksi XGBoost, penelitian ini menggunakan penerapan dua metode komputasi, yaitu Continuous Restricted Boltzmann Machine (CRBM) dan Particle Swarm Optimization (PSO). CRBM berfungsi dalam mengekstraksi fitur yang lebih signifikan dari data asli, sementara PSO secara otomatis mengoptimalkan hyper-parameter selama proses pelatihan model. Hasil penelitian menunjukkan bahwa kombinasi metode CRBM dan PSO mampu meningkatkan akurasi prediksi litologi secara signifikan. Dengan capaian akurasi sebesar 97,56%, precision sebesar 97,60%, recall sebesar 97,56%, dan F1-score sebesar 97,54%, pendekatan ini terbukti efektif dalam menangani tantangan representasi data serta optimisasi hyper-parameter. Pendekatan berbasis CRBM dan PSO memberikan solusi yang andal dan efisien dalam eksplorasi geologi, khususnya dalam menganalisis data yang kompleks. Oleh karena itu, model ini memberikan alternatif yang unggul untuk optimasi prediksi litologi serta berkontribusi pada peningkatan presisi dalam eksplorasi sumber daya migas.
==================================================================================================================================
In the early stages of oil and gas exploration, obtaining accurate lithology information is a crucial step in geological analysis. Lithology prediction has become a primary focus in geoscience research due to its significance in subsurface mapping and characterization, particularly for sandstone lithology. The Extreme Gradient Boosting (XGBoost) algorithm has been proven to be superior in pattern recognition compared to conventional models, owing to its ability to implement regularization, gradient boosting, and tree-based classification. However, this complex model involves numerous hyper-parameters, often leading to difficulties in optimization and inefficiencies when handling high-dimensional variables.To enhance the predictive performance of XGBoost, this study employs two computational methods, Continuous Restricted Boltzmann Machine (CRBM) and Particle Swarm Optimization (PSO). CRBM functions to extract more significant features from the original data, while PSO automatically optimizes the hyper-parameters during the model training process. The research findings indicate that the combination of CRBM and PSO significantly improves lithology prediction accuracy. With an accuracy of 97.56%, precision of 97.60%, recall of 97.56%, and F1-score of 97.54%, this approach has proven effective in addressing challenges related to data representation and hyper-parameter optimization.The CRBM-PSO-based approach provides a reliable and efficient solution for geological exploration, particularly in analyzing complex data. Therefore, this model offers a superior alternative for optimizing lithology prediction and contributes to enhancing precision in oil and gas resource exploration.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CRBM, XGBoost, optimisasi prediksi litologi, PSO, well log
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QE Geology > QE601 Geology, Structural
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Rijkaard Elyandrie
Date Deposited:	06 Feb 2025 09:19
Last Modified:	06 Feb 2025 09:19
URI:	http://repository.its.ac.id/id/eprint/118491

Actions (login required)

View Item