Pemetaan Kerentanan Tanah Longsor Berbasis Convolutional Neural Network (CNN) Dan Analisis Interpretasi SHAP Di Kecamatan Pacet, Kabupaten Mojokerto

Bani, Mohammad Alvi Aldi (2025) Pemetaan Kerentanan Tanah Longsor Berbasis Convolutional Neural Network (CNN) Dan Analisis Interpretasi SHAP Di Kecamatan Pacet, Kabupaten Mojokerto. Masters thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 6015241021-Master_Thesis.pdf] Text
6015241021-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (9MB) | Request a copy

Abstract

Kecamatan Pacet, Kabupaten Mojokerto, merupakan salah satu wilayah di Provinsi Jawa Timur yang sering mengalami kejadian tanah longsor. Sebagai upaya mitigasi, penelitian ini bertujuan untuk memetakan distribusi kelas kerentanan tanah longsor di daerah penelitian. Penelitian ini mengaplikasikan teknik analisis machine learning yang menggabungkan model Convolutional Neural Network (CNN) dan SHapley Additive exPlanations (SHAP) untuk menilai distribusi kerentanan tanah longsor. CNN digunakan untuk mengolah citra satelit dan data spasial lainnya, seperti kemiringan lereng, jenis tanah, penggunaan lahan, dan curah hujan, untuk mendeteksi pola-pola kerentanan tanah longsor. Metode SHAP diterapkan untuk mengevaluasi dan menjelaskan kontribusi setiap faktor terhadap prediksi kerentanan yang dihasilkan oleh model CNN, memberikan wawasan yang lebih jelas mengenai faktor-faktor pemicu tanah longsor yang signifikan. Sebanyak 14 faktor pemicu tanah longsor digunakan dalam penelitian ini, yaitu: TWI (Topographic Wetness Index), SPI (Stream Power Index), STI (Sediment Transport Index), kemiringan lereng, aspect lereng, elevasi, curvature profile, curvature plan, jarak dari sungai, litologi, curah hujan, penggunaan lahan, jarak dari jalan dan NDVI. Pembagian data untuk pelatihan dan pengujian model adalah 70% dan 30%, yang dihasilkan dari inventarisasi data history kejadian longsor. Model CNN yang dihasilkan menunjukkan performa yang sangat baik dengan tingkat akurasi yang tinggi, mencapai nilai AUC sebesar 97.94% pada data pelatihan dan 97.31% pada data validasi. Desa Cembor dan Claket menjadi desa dengan luas wilayah dengan risiko paling tinggi dibanding dengan desa-desa lainnya. Curvature profile (kelengkungan profil lereng) teridentifikasi sebagai faktor dengan pengaruh prediktif tertinggi, diikuti oleh kemiringan lereng, penggunaan lahan, jarak dari jalan, lithologi dan curah hujan.
===========================================================================================================================================
Pacet Sub-district, Mojokerto District, is one of the areas in East Java Province that frequently experiences landslides. As a mitigation effort, this study aims to map the distribution of landslide susceptibility classes in the study area. This research applies machine learning analysis technique that combines Convolutional Neural Network (CNN) and SHapley Additive exPlanations (SHAP) models to assess the distribution of landslide susceptibility. CNN is used to process satellite images and other spatial data, such as slope, soil type, land use, and rainfall, to detect patterns of landslide susceptibility. The SHAP method was applied to evaluate and explain the contribution of each factor to the susceptibility predictions generated by the CNN model, providing clearer insights into the significant landslide-inducing factors. A total of 14 landslide triggering factors were used in this study, namely: TWI (Topographic Wetness Index), SPI (Stream Power Index), STI (Sediment Transport Index), slope slope, slope aspect, elevation, curvature profile, curvature plan, distance from river, lithology, rainfall, land use, distance from road and NDVI. The data split for model training and testing is 70% and 30%, which is generated from the inventory of historical landslide data. The resulting CNN model showed excellent performance with a high level of accuracy, achieving an AUC value of 97.94% on the training data and 97.31% on the validation data. Cembor and Claket villages have the highest risk area compared to other villages. Curvature profile was identified as the factor with the highest predictive influence, followed by slope, land use, distance from roads, lithology and rainfall.

Item Type: Thesis (Masters)
Uncontrolled Keywords: kerentanan bencana, tanah longsor, (Artificial Intelligence) Ai, CNN, SHAP, Kec, Pacet, Kab. Mojokerto disaster susceptibility, landslide, Ai, CNN, SHAP, Kec. Pacet, Kab. Mojokerto
Subjects: G Geography. Anthropology. Recreation > G Geography (General) > G70.212 ArcGIS. Geographic information systems.
G Geography. Anthropology. Recreation > G Geography (General) > G70.217 Geospatial data
G Geography. Anthropology. Recreation > G Geography (General) > G70.5.I4 Remote sensing
T Technology > T Technology (General) > T174.5 Technology--Risk assessment.
T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
Divisions: Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Regional & Urban Planning > 35101-(S2) Master Thesis
Depositing User: Mohammad Alvi Aldi Bani
Date Deposited: 05 Aug 2025 04:02
Last Modified: 18 Sep 2025 02:59
URI: http://repository.its.ac.id/id/eprint/125407

Actions (login required)

View Item View Item