IMPLEMENTATION OF MAPPING-BASED MACHINE LEARNING ALGORITHM AS NON-STRUCTURAL DISASTER MITIGATION TO DETECT LANDSLIDE SUSCEPTIBILITY IN TAKARI DISTRICT

Aug 5, 2026

Upload by Ezra

Research presents a cartographic-based machine learning framework to identify landslide-susceptible areas in Takari District using non-structural disaster mitigation principles. Machine learning methods—Support Vector Machine, Naive Bayes Classifier, Ordinal Logistic Regression, Random Forest, and Decision Tree—are applied to rainfall-related conditions and evaluated with accuracy and Kappa metrics. Ordinal Logistic Regression and Random Forest reach 74.36% accuracy, while Random Forest achieves the highest Kappa (0.5397). Prediction outputs are interpreted together with rainfall factors and environmental variables, and susceptibility mapping is produced via overlay of slope, geology, and soil maps, with suggestions for spatial validation using additional datasets or remote sensing.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Abstract

Introduction

Landslides as a major hazard in Indonesia

Causes and contributing factors

Study area: Takari District

IMPLEMENTATION OF MAPPING-BASED MACHINE LEARNING ALGORITHM AS NON-STRUCTURAL DISASTER MITIGATION TO DETECT LANDSLIDE SUSCEPTIBILITY IN TAKARI DISTRICT Sefri Imanuel Fallo 1*, Lidia Paskalia Nipu2 1Mathematics Study Program, Faculty of Mathematics and Natural Sciences, Universitas San Pedro 2Environmental Engineering Study Program, Faculty of Engineering and Planning, Universitas San Pedro Jln. Ir. Soekarno, Kupang, 85112, Indonesia Corresponding author’s e-mail: *[fallosefriimanuel@gmail.com](fallosefriimanuel@gmail.com) ABSTRACT Article History: Received: 21st November 2023 Revised: 7th January 2024 Accepted: 10th March 2024 Published: 1st June 2024 Keywords: Machine Learning; Landslide Susceptibility; Non-Structural Mitigation; Mapping. This research is primarily dedicated to providing a comprehensive exposition of the methodology applied in the deployment of a cartographic-based machine learning algorithm designed for the precise identification of areas susceptible to landslides within the geographical confines of the Takari District. This research delves into the application of mapping-based machine learning algorithms in the domain of non-structural disaster mitigation, with a specific emphasis on the detection of landslide susceptibility within the Takari District. A range of machine learning algorithms, including Support Vector Machine, Naive Bayes Classifier, Ordinal Logistic Regression, Random Forest, and Decision Tree, were harnessed to evaluate rainfall data within the context of landslide susceptibility. An evaluation of model performance, anchored in accuracy and Kappa metrics, unveiled that both the Ordinal Logistic Regression and Random Forest models exhibited noteworthy precision, reaching a commendable 74.36%. Nevertheless, a meticulous examination of Kappa values disclosed the ascendancy of the Random Forest model, which achieved a superior Kappa value of 0.5397. As portrayed in the visual representation provided, it becomes manifest that the Random Forest algorithm's prognostications yield 66 instances of cloudy atmospheric conditions, 48 occurrences of light precipitation, and 3 episodes of moderate rainfall. These predictions are influenced by several factors, including average temperature, humidity levels, wind speed, duration of sunlight, and wind direction at maximum speed. Consequently, this comprehensive analysis underscores the Random Forest algorithm as the most efficacious model for landslide susceptibility prediction. Furthermore, the study seamlessly integrated overlay maps, encompassing the Slope Inclination Map of the Takari District, Geological Map of the Takari District, and Soil Type Map of the Takari District, to contribute to the formulation of a definitive map delineating the susceptibility to landslides in the Takari District. Furthermore, further research could conduct spatial validation of the model predictions using additional datasets or remote sensing data to validate the accuracy of the landslide usceptibility map and ensure its applicability across different geographica regions. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-ShareAlike 4.0 International License. How to cite this article: S. I Fallo and L. P. Nipu.,“IMPLEMENTATION OF MAPPING-BASED MACHINE LEARNING ALGORITHM AS NON-STRUCTURAL DISASTER MITIGATION TO DETECT LANDSLIDE SUSCEPTIBILITY IN TAKARI DISTRICT,”BAREKENG: J. Math. & App., vol. 18, iss. 2, pp. 0877-0892, June, 2024. Copyright © 2024 Author(s) Journal homepage: [https://ojs3.unpatti.ac.id/index.php/barekeng/](https://ojs3.unpatti.ac.id/index.php/barekeng/) Journal e-mail: [barekeng.math@yahoo.com](barekeng.math@yahoo.com); [barekeng.journal@mail.unpatti.ac.id](barekeng.journal@mail.unpatti.ac.id) Research Article ∙ Open Access 1. INTRODUCTION Natural disasters are unpredictable events that can occur at any time and in any location, resulting in

FAQs

What is the main goal of this research in Takari District?

To deploy a mapping-based machine learning approach that detects and maps landslide susceptibility as part of non-structural disaster mitigation.

Which machine learning algorithms are compared for landslide susceptibility detection?

Support Vector Machine, Naive Bayes Classifier, Ordinal Logistic Regression, Random Forest, and Decision Tree are used to evaluate landslide susceptibility.

How is model performance measured, and which model performs best?

Performance is assessed using accuracy and Kappa metrics; Random Forest achieves the highest Kappa value (0.5397) and is concluded as the most effective model.

Aug 5, 2026

Upload by Ezra