<![CDATA[Landslides are one of the most frequent natural disasters in Indonesia, causing significant environmental and socio-economic impacts. However, existing early warning systems remain limited in accuracy and are often unable to capture the complex spatial and temporal dynamics of landslide- triggering factors. Therefore, this study aims to develop an artificial intelligence model based on a spatiotemporal transformer to improve landslide prediction and support early warning systems. This research employs a quantitative experimental approach by integrating spatial data such as slope, soil type, and land cover with temporal data including rainfall, temperature, and humidity, as well as historical landslide records. The data are processed through preprocessing stages including normalization and sequence construction, followed by model training using a spatiotemporal transformer architecture. The model performance is evaluated using accuracy, precision, recall, F1 score, mean absolute error, and root mean square error. The results indicate that rainfall and humidity exhibit high variability and play a major role as dynamic triggering factors, while slope represents a dominant static factor contributing to landslide susceptibility. The developed model demonstrates high predictive performance with an accuracy of 0.97, indicating its effectiveness in classifying landslide risk. However, the findings also suggest the need for further validation using more diverse datasets to ensure model generalization. In conclusion, the integration of artificial intelligence and spatiotemporal data provides a promising approach for enhancing landslide prediction and developing more adaptive and reliable early warning systems.]]>
Copyrights © 2026
