Identifikasi Daerah Rawan Banjir Menggunakan Algoritma Random Forest dan Support Vector Machine (SVM) (Studi Kasus : Kabupaten Pasuruan)

Purba, Rick Owen Handel (2024) Identifikasi Daerah Rawan Banjir Menggunakan Algoritma Random Forest dan Support Vector Machine (SVM) (Studi Kasus : Kabupaten Pasuruan). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5016201092-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (9MB) | Request a copy

Abstract

Peristiwa bencana banjir berpotensi untuk merusak dan menyebabkan kerugian pada daerah yang terdampak beserta dengan komponen ekosistem di dalamnya. Kabupaten Pasuruan merupakan salah satu daerah rawan banjir yang berpotensi menyebabkan kerugian besar dan terhambatnya kegiatan masyarakat setempat. Kerugian yang disebabkan oleh banjir tersebut menjadi salah satu landasan terhadap urgensi identifikasi daerah rawan banjir yang akurat. Implementasi identifikasi daerah rawan banjir tersebut juga merupakan usaha dalam mewujudkan tujuan dari SDGs, terutama pada tujuan nomor 11 dan 15, terkait mewujudkan pemukiman manusia yang berkelanjutan dan melindungi ekosistem daratan. Oleh karena itu, penelitian ini bertujuan untuk mengidentifikasi daerah rawan banjir menggunakan metode Machine Learning dengan algoritma Random Forest (RF) dan Support Vector Machine (SVM) di Kabupaten Pasuruan. Untuk melakukan identifikasi, digunakan 11 parameter yang terpilih berdasarkan tahap seleksi parameter, yaitu kerapatan saluran sungai, jarak terhadap sungai, indeks kerapatan bangunan (NDBI), jenis tanah, tutupan lahan, curah hujan, Manning’s Roughness Coefficients, kelerengan, Band 5, TWI, dan ketinggian. Pemodelan dengan Machine Learning menggunakan dataset (banjir dan non-banjir) dan seluruh raster parameter yang terseleksi. Adapun tahap pre-processing mencakup scalling dan konversi data kategorik menjadi numerik. Pada algoritma Random Forest (RF) dilakukan dengan rasio perbandingan 70:30 dan dengan algoritma Support Vector Machine (SVM) dilakukan dengan rasio perbandingan 50:50, 60:40, dan 70:30. Diketahui model terbaik yang dihasilkan adalah SVM ratio 70:30 dengan Accuracy senilai 0.9828, Sensitivity (Sen) senilai 0.9655, Specificity senilai 1.0000, Balanced Accuracy senilai 0.9828, dan AUC senilai 1.0000. Pada model ini, parameter yang paling signifikan berdasarkan persentase Importance Degree Index (IDI) adalah elevasi dengan persentase senilai 100%. Berdasarkan validasi model SVM 70:30 diketahui bahwa daerah yang sangat rawan terhadap bencana banjir, yaitu Kecamatan Bangil, Beji, Gempol, Gondang Wetan, Grati, Kejayan, Kraton, Lekok, Nguling, Pandaan, Pohjentrek, Rejoso, Rembang, Sukorejo, Winongan, dan Wonorejo.
=================================================================================================================================
Flood events have the potential to cause damage and loss in affected areas along with their ecosystem components. Pasuruan Regency is one of the flood-prone areas that has the potential to cause significant losses and disrupt local community activities. The losses caused by these floods underline the urgency of accurately identifying flood-prone areas. The implementation of flood-prone area identification also aligns with the Sustainable Development Goals (SDGs), particularly goals 11 and 15, which aim to create sustainable human settlements and protect terrestrial ecosystems. Therefore, this study aims to identify flood-prone areas using Machine Learning methods with the Random Forest (RF) and Support Vector Machine (SVM) algorithms in Pasuruan Regency. To perform the identification, 11 parameters were selected based on a parameter selection process: river channel density, distance to river, Normalized Difference Built-up Index (NDBI), soil type, land cover, rainfall, Manning’s Roughness Coefficients, slope, Band 5, Topographic Wetness Index (TWI), and elevation. Machine Learning modeling utilized datasets (flood and non-flood) and all selected raster parameters. The pre-processing stage included scaling and converting categorical data into numerical format. The Random Forest (RF) algorithm was applied with a 70:30 split ratio, and the Support Vector Machine (SVM) algorithm was applied with split ratios of 50:50, 60:40, and 70:30. The best model obtained was the SVM with a 70:30 ratio, which achieved an accuracy of 0.9828, sensitivity of 0.9655, specificity of 1.0000, balanced accuracy of 0.9828, and an AUC of 1.0000. In this model, the most significant parameter, based on the Importance Degree Index (IDI) percentage, was elevation, with a percentage value of 100%. Based on the validation of the 70:30 SVM model, the areas identified as highly vulnerable to flood disasters include the districts of Bangil, Beji, Gempol, Gondang Wetan, Grati, Kejayan, Kraton, Lekok, Nguling, Pandaan, Pohjentrek, Rejoso, Rembang, Sukorejo, Winongan, and Wonorejo.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Flood, Support Vector Machine (SVM), Random Forest, SDGs, Banjir
Subjects:	G Geography. Anthropology. Recreation > G Geography (General) > G70.212 ArcGIS. Geographic information systems.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29202-(S1) Undergraduate Thesis
Depositing User:	Rick Owen Handel Purba
Date Deposited:	09 Jul 2024 08:53
Last Modified:	09 Jul 2024 08:53
URI:	http://repository.its.ac.id/id/eprint/108209

Actions (login required)

View Item