Analisis Perbandingan Pemodelan Spasial Kerawanan Bencana Tanah Longsor Metode Random Forest Dan Naïve Bayes (Studi Kasus : Kabupaten Malang).

Ummah, Muhammad Hidayatul (2022) Analisis Perbandingan Pemodelan Spasial Kerawanan Bencana Tanah Longsor Metode Random Forest Dan Naïve Bayes (Studi Kasus : Kabupaten Malang). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 03311840000070-UNDERGRADUATE_THESIS.pdf Restricted to Repository staff only Download (9MB)

Abstract

Tanah longsor merupakan salah satu bencana yang menyebabkan kerugian besar baik pada kehidupan manusia maupun infrastruktur. Penilaian kerawanan tanah longsor perlu dilakukan pengembangan terus menurus untuk mendapatkan model yang akurat. Teknologi machine learning dapat meningkatkan akurasi pemodelan kerawanan tanah longsor. Oleh karena itu, dalam penelitian ini bertujuan untuk mengevaluasi perbandingan pemodelan spasial dengan menggunakan algoritma machine learning naïve bayes (NB) dan random forest (RF) di Kabupaten Malang guna mendapatkan model terbaik. Dalam pemodelan, 12 parameter digunakan yaitu elevasi, slope, aspect, jenis tanah, jenis geologi, jarak dari sesar, NDVI (Normalized Different Vegetation Index), jarak dari sungai, kerapatan sungai, TWI (Topographic Wetness Index), curah hujan, dan tutupan lahan. Setiap model dilakukan evaluasi terhadap 9 parameter diantaranya ROC (Receiver Operator Characteristic)-AUC (Area Under Curve), accuracy (acc), sensitivity (sn), specificity (sp), balanced accuracy (ba), g-mean (gm), cohen’s kappa (CK), dan matthew’s correlation coefficient (MCC). Pada penelitian ini terdapat tiga skenario splitting ratio training dan testing dataset untuk algoritma NB yaitu 50:50,60:40, dan 70:30. Sedangkan RF hanya menggunakan 70:30. Berdasarkan hasil pemodelan didapatkan model RF 70:30 merupakan algoritma terbaik dalam studi kasus ini dengan nilai 0,884 untuk ACC, 0,765 untuk SN, 0,962 untuk SP, 0,863 untuk GM, 0,857 untuk BA, 0,749 untuk CK, 0,876 untuk MCC, dan 0,943 untuk AUC. Dari keseluruhan model, parameter elevasi merupakan parameter dengan nilai tingkat kontribusi relatif tertinggi yaitu 100%. Berdasarkan model terbaik, didapatkan wilayah Kabupaten Malang didominasi oleh tingkat kerawanan tinggi dengan luasan sebesar 177.208,83 ha (51% dari keseluruhan wilayah).
===================================================================================================================================
Landslides are one of the disasters that cause huge losses to both human life and infrastructure. Landslide susceptibility assessment needs to be improved continuously to get an accurate model. Machine learning technology can improve the accuracy of landslide susceptibility modelling. Therefore, this research aims to compare spatial modeling using machine learning nave Bayes (NB) and random forest (RF) algorithms in Malang Regency to get the best model. In the modeling, 12 parameters are used: elevation, slope, aspect, soil type, geological type, distance from the fault, NDVI (Normalized Different Vegetation Index), distance from the river, river density, and TWI (Topographic Wetness Index), rainfall, and land cover. Each model was evaluated by eight parameters, including ROC (Receiver Operator Characteristic)-AUC (Area Under Curve), accuracy (acc), sensitivity (sn), specificity (sp), balanced accuracy (ba), g-mean (gm) , Cohen's kappa (CK), and Matthew's correlation coefficient (MCC). In this research, there are three scenarios of splitting ratio training and testing datasets for the NB algorithm, namely 50:50, 60:40, and 70:30. While RF only uses 70:30. Based on the modeling results, the RF 70:30 model is the best algorithm in this case study with a value of 0.884 for ACC, 0.765 for SN, 0.962 for SP, 0.863 for GM, 0.857 for BA, 0.749 for CK, 0.876 for MCC, and 0.943 for AUC. From the whole model, the elevation parameter is the parameter with the highest relative contribution level value, 100%. Based on the best model, it was found that the Malang Regency area was dominated by a high level of vulnerability, with an area of 177,208.83 ha (51% of the total area).

Item Type:	Thesis (Other)
Additional Information:	RSG 551.307 Umm a-1 2022
Uncontrolled Keywords:	Naive bayes, random forest, tanah longsor. Landslide, Naive bayes, random forest.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29202-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	20 May 2026 03:53
Last Modified:	20 May 2026 03:53
URI:	http://repository.its.ac.id/id/eprint/133268

Actions (login required)

View Item