Pemanfaatan XAI (Explainable Artificial Intelligence) dalam Pemetaan Kerentanan Bencana Tanah Longsor

Syah, Nabil Julian and Ezra, Moch Septian (2025) Pemanfaatan XAI (Explainable Artificial Intelligence) dalam Pemetaan Kerentanan Bencana Tanah Longsor. Project Report. [s.n.], [s.l.]. (Unpublished)

Text 5025231023_5025231120-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Praktik ini mengembangkan sistem klasifikasi kerentanan tanah longsor berbasis Deep Learning menggunakan arsitektur Convolutional Neural Network (CNN) yang dilengkapi dengan Squeeze-and-Excitation (SE) dan Spatial Pyramid Pooling (SPP). Sebagai pembanding, menggunakan model Machine Learning tree base seperti Random Forest (RF) dan Gradient Boosting (GB) yang dioptimalkan melalui Parameter tuning. Sistem menerima masukan berupa data raster (.tif) multi-kanal. Proses pemrosesan data dilakukan dengan menggunakan pustaka GDAL untuk membaca dan mengelola data citra spasial. Hasil evaluasi model menunjukkan bahwa pendekatan Tree-based memiliki performa yang lebih stabil dan kemampuan generalisasi yang lebih baik. Sebaliknya, model CNN terindikasi mengalami bias terhadap kelas mayoritas yang disebabkan oleh ketidakseimbangan data. Untuk meningkatkan transparansi keputusan model, pendekatan Explainable AI (XAI) diterapkan menggunakan metode SHAP (SHapley Additive exPlanations). Analisis melalui summary plot dan force plot berhasil mengidentifikasi fitur-fitur geologis dominan serta memvisualisasikan kontribusi spesifik setiap fitur terhadap prediksi longsor.
=================================================================================================================================
This study develops a landslide vulnerability classification system based on Deep Learning using a Convolutional Neural Network (CNN) architecture enhanced with Squeeze-and-Excitation (SE) and Spatial Pyramid Pooling (SPP). As a comparison, tree-based Machine Learning models such as Random Forest (RF) and Gradient Boosting (GB) were employed, optimized through parameter tuning. The system processes multi-channel raster (.tif) data, with spatial image management handled using the GDAL library. Evaluation results indicate that tree-based approaches demonstrate more stable performance and stronger generalization capabilities. In contrast, the CNN model shows bias toward the majority class, largely due to data imbalance. To improve model decision transparency, Explainable AI (XAI) techniques were applied using SHAP (SHapley Additive exPlanations). Analysis through summary plots and force plots successfully identified dominant geological features and visualized the specific contribution of each feature to landslide predictions.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	CNN, Explainable AI (SHAP), Gradient Boosting, Random Forest
Subjects:	G Geography. Anthropology. Recreation > G Geography (General) > G70.217 Geospatial data T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Industrial Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Nabil Julian Syah
Date Deposited:	13 Apr 2026 06:34
Last Modified:	13 Apr 2026 06:34
URI:	http://repository.its.ac.id/id/eprint/132779

Actions (login required)

View Item