Journal of Geographical Sciences, Journal Year: 2024, Volume and Issue: 34(11), P. 2111 - 2127
Published: Nov. 1, 2024
Language: Английский
Journal of Geographical Sciences, Journal Year: 2024, Volume and Issue: 34(11), P. 2111 - 2127
Published: Nov. 1, 2024
Language: Английский
Journal of Hydrology Regional Studies, Journal Year: 2025, Volume and Issue: 57, P. 102174 - 102174
Published: Jan. 8, 2025
Language: Английский
Citations
0Published: Jan. 1, 2024
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Language: Английский
Citations
0Published: Jan. 1, 2024
This study endeavors to assess the performance of landslide susceptibility models and differences in factors across various slope unit scales. Initially, we curated a geospatial dataset comprising 3594 historical events 22 initial factors. Subsequently, 30 sets schemes were generated based on different parameter combinations. These datasets trained tested 7:3 ratio using random forest (RF) model, resulting accuracy values ranging from 0.686 0.812 AUC 0.770 0.884. In addition, model optimization was facilitated GridSearchCV (GS), generation mapping (LSM) for three representative scenarios. Lastly, decision-making mechanism at scales elucidated explainable artificial intelligence (XAI). Key findings elucidate: (1) The scale has profound effect efficacy models, with smaller minimum area thresholds improving but cost increased computational time. (2) Overly homogeneous units prove inadequate encapsulating intricacies real-world terrain slopes, thereby precipitating diminished accuracy. (3) optimized scheme may enhance hyperparameter is somewhat more appropriate XAI models. Wise selection optimal cell key training machine learning-based superior performance.
Language: Английский
Citations
0Journal of Geographical Sciences, Journal Year: 2024, Volume and Issue: 34(11), P. 2111 - 2127
Published: Nov. 1, 2024
Language: Английский
Citations
0