Estimasi Risiko Kerugian Banjir Berbasis Damage Ratio Menggunakan Quantile Regression Forest Dalam Kerangka Catastrophe Modelling

Prasetia, Faris Rayhanabil (2026) Estimasi Risiko Kerugian Banjir Berbasis Damage Ratio Menggunakan Quantile Regression Forest Dalam Kerangka Catastrophe Modelling. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221093-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Banjir merupakan salah satu bencana hidrometeorologi yang seringkali menimbulkan kerusakan fisik serta kerugian ekonomi yang signifikan. Oleh karena itu, diperlukan pendekatan kuantitatif yang mampu mengestimasi risiko kerugian secara lebih sistematis dan terukur. Penelitian ini bertujuan untuk mengestimasi risiko kerugian akibat banjir pada bangunan rumah tinggal di Provinsi Jawa Timur dengan menggunakan pendekatan Catastrophe Modelling dan metode Quantile Regression Forest (QRF). Dalam penelitian ini, Damage ratio digunakan sebagai proksi tingkat kerusakan relatif bangunan dan bukan sebagai rasio kerugian aktual terhadap nilai aset masing-masing bangunan. Nilai damage ratio disusun berdasarkan kategori kerusakan rumah, yaitu rusak berat, rusak sedang, dan rusak ringan, kemudian digunakan sebagai variabel respons dalam pemodelan QRF. Variabel prediktor yang digunakan meliputi ketinggian banjir, curah hujan, elevasi wilayah, kepadatan penduduk, jumlah penduduk, dan tingkat kemiskinan. Hasil analisis feature importance menunjukkan bahwa ketinggian banjir merupakan faktor yang paling berpengaruh terhadap Damage Ratio dengan kontribusi sebesar 41,70%, diikuti oleh curah hujan sebesar 19,22%, kepadatan penduduk sebesar 15,55%, dan tingkat kemiskinan sebesar 9,10%. Evaluasi model menunjukkan bahwa metode QRF mampu merepresentasikan pola estimasi kuantil, khususnya pada kuantil tinggi, yang ditunjukkan oleh nilai Pinball Loss terendah pada kuantil ke-99 (Q99) sebesar 0,004678 serta Coverage Probability pada kuantil ke-90 (Q90) sebesar 89,47%. Meskipun demikian, kemampuan generalisasi model masih tergolong terbatas, sebagaimana tercermin dari nilai Out-of-Bag R² sebesar 21,68%. Hasil estimasi risiko menunjukkan bahwa distribusi kerugian banjir memiliki karakteristik heavy-tailed, yang mengindikasikan adanya potensi kerugian ekstrem dengan probabilitas yang relatif kecil. Pada tingkat kepercayaan 99%, nilai Value at Risk (VaR) berada pada rentang Rp5,8 miliar hingga Rp8,8 miliar, dengan nilai tertinggi diperoleh pada skenario ELT_Q99 sebesar Rp8.856.694.386. Sementara itu, nilai Tail Value at Risk (TVaR) pada tingkat kepercayaan yang sama menunjukkan besarnya risiko pada bagian ekor distribusi, dengan nilai tertinggi pada skenario ELT_Q99 mencapai Rp122.179.679.040. Temuan ini menunjukkan bahwa pendekatan QRF yang dikombinasikan dengan ukuran risiko berbasis tail risk dapat dimanfaatkan untuk memberikan gambaran yang lebih komprehensif mengenai potensi kerugian banjir ekstrem pada berbagai wilayah.
=======================================================================================================================================
Flooding is one of the hydrometeorological disasters that frequently causes significant physical damage and economic losses. A quantitative approach is therefore needed to estimate loss risk in a more systematic and measurable manner. This study aims to estimate flood loss risk for residential buildings in East Java Province using the Catastrophe Modelling approach and the Quantile Regression Forest (QRF) method. In this study, the Damage Ratio is used as a proxy for the relative level of building damage, rather than as an actual loss ratio to the asset value of each individual building. The Damage Ratio is constructed based on three categories of housing damage, namely severely damaged, moderately damaged, and lightly damaged houses, and is then used as the response variable in the QRF model. The predictor variables used in this study include flood depth, rainfall, regional elevation, population density, total population, and poverty level. The feature importance analysis shows that flood depth is the most influential factor affecting the Damage Ratio, with a contribution of 41.70%, followed by rainfall at 19.22%, population density at 15.55%, and poverty level at 9.10%. The model evaluation indicates that QRF is able to represent quantile estimation patterns, particularly at higher quantiles. This is shown by the lowest Pinball Loss value at the 99th quantile (Q99), amounting to 0.004678, and a Coverage Probability of 89.47% at the 90th quantile (Q90). However, the model’s generalization ability remains limited, as reflected by the Out-of-Bag R² value of 21.68%. The risk estimation results indicate that the flood loss distribution exhibits heavy-tailed characteristics, suggesting the potential for extreme losses with relatively low probability. At the 99% confidence level, the Value at Risk (VaR) ranges from IDR 5.8 billion to IDR 8.8 billion, with the highest value obtained under the ELT_Q99 scenario at IDR 8,856,694,386. Meanwhile, the Tail Value at Risk (TVaR) at the same confidence level reflects the magnitude of risk in the tail of the distribution, with the highest value under the ELT_Q99 scenario reaching IDR 122,179,679,040. These findings show that the QRF approach combined with tail risk-based risk measures can provide a more comprehensive overview of potential extreme flood losses across different regions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Banjir, Flood, Damage Ratio, Quantile Regression Forest, Value at Risk, Tail Value at Risk.
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > T Technology (General) > T58.62 Decision support systems Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Faris Rayhanabil Prasetia
Date Deposited:	17 Jul 2026 03:24
Last Modified:	17 Jul 2026 03:24
URI:	http://repository.its.ac.id/id/eprint/135182

Actions (login required)

View Item