Urban Road Collapse Points Across China from 2018 to 2024

From 2018 to 2024, this dataset comprises 1,544 georeferenced urban road collapse points across China. It was compiled and georeferenced by Minghui Guo based on publicly available online information and is formatted for ArcGIS. The associated analytical code includes predictive models using XGBoost, xLSTM, and PINN algorithms.

Use Cases

• Predicting road collapse susceptibility based on geospatial and temporal features mentioned in the description
• Interpreting model predictions for road failure using SHAP summary plots as referenced
• Training and evaluating machine learning models like XGBoost and xLSTM on georeferenced infrastructure data
• Generating ROC curves and confusion matrices to assess model performance for hazard prediction

Strengths

• Contains 1,544 specific data points for model training and validation
• Covers a 6-year time range from 2018 to 2024 for temporal analysis
• Includes pre-developed analytical code for XGBoost, xLSTM, and PINN algorithms

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count for the underlying data is unknown, which may limit suitability assessment
• Data may reflect geographic and source bias inherent to the compilation from public online information

Provenance

Source

Minghui Guo

Collection Method

Compiled and georeferenced by the author based on publicly available online information

Time Range

2018 to 2024

Freshness

Last updated 2026-05-20 14:11:24; freshness should be verified

Geography

Urban areas across China