Sulistyo, Rano Noumi (2025) Pemanfaatan XAI (Xplainable Artificial Intelligence) dalam Pemetaan Kerentanan Bencana Tanah Longsor di Kecamatan Pacet Kab. Mojokerto. Project Report. [s.n.], [s.l.]. (Unpublished)
![]() |
Text
5025211185_Project_Report.pdf - Accepted Version Download (1MB) |
Abstract
Praktik ini mengembangkan sistem klasifikasi kerentanan tanah longsor berbasis Convolutional Neural Network (CNN) yang menerima masukan berupa data raster (.tif) multi-kanal. Proses pemrosesan data dilakukan dengan menggunakan pustaka GDAL untuk membaca dan mengelola data citra spasial. Dataset hasil ekstraksi kemudian dilatih dalam model CNN yang dilengkapi dengan Squeeze-and-Excitation (SE) dan Spatial Pyramid Pooling (SPP). Hasil pelatihan menunjukkan akurasi tinggi, dengan nilai Area Under Curve (AUC) sebesar 97,94% pada data pelatihan dan 97,31% pada data validasi, serta kestabilan loss selama 500 epoch. Untuk meningkatkan interpretabilitas model, digunakan pendekatan Explainable AI (XAI) dengan metode SHAP (SHapley Additive exPlanations). Analisis SHAP menunjukkan fitur-fitur dominan seperti curvature profile, kemiringan lereng, penggunaan lahan, dan jarak dari jalan sebagai kontributor utama dalam prediksi. Visualisasi summary plot dan force plot berhasil mengungkap kontribusi positif dan negatif setiap fitur pada setiap prediksi, serta menunjukkan bahwa model membentuk keputusan berdasarkan kombinasi fitur yang kompleks. Hasil ini membuktikan bahwa integrasi CNN dan SHAP efektif dalam mendukung sistem pemetaan kerentanan bencana berbasis data geospasial.
=================================================================================================================================
This project develops a landslide susceptibility classification system based on a Convolutional Neural Network (CNN) that accepts multi-channel raster (.tif) data as input. The data preprocessing stage utilizes the GDAL library to read and manage spatial imagery data. The extracted dataset is then trained using a CNN model equipped with Squeeze-and-Excitation (SE) and Spatial Pyramid Pooling (SPP). The training results demonstrate high accuracy, with an Area Under Curve (AUC) of 97.94% on the training set and 97.31% on the validation set, along with stable loss values throughout 500 epochs. To enhance model interpretability, an Explainable AI (XAI) approach using SHAP (SHapley Additive exPlanations) is employed. SHAP analysis identifies dominant features such as curvature profile, slope, land use, and distance to roads as the primary contributors to predictions. The summary plot and force plot visualizations successfully reveal both positive and negative feature contributions to individual predictions and show that the model's decisions are based on a complex combination of features. These findings confirm that the integration of CNN and SHAP is effective for supporting geospatial data-driven disaster risk mapping systems.
Item Type: | Monograph (Project Report) |
---|---|
Uncontrolled Keywords: | CNN, SHAP, GDAL, Longsor, Geospasial |
Subjects: | T Technology > T Technology (General) > T11 Technical writing. Scientific Writing T Technology > T Technology (General) > T174.5 Technology--Risk assessment. T Technology > T Technology (General) > T385 Visualization--Technique T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.62 Decision support systems |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis |
Depositing User: | Rano Noumi Sulistyo |
Date Deposited: | 04 Jul 2025 02:12 |
Last Modified: | 04 Jul 2025 02:12 |
URI: | http://repository.its.ac.id/id/eprint/119344 |
Actions (login required)
![]() |
View Item |