Expertise
Four research directions
Each direction addresses a family of real urban and infrastructure problems and has its own responsible members and publication record.
Direction 1
ML/DL for urban geotechnical hazard and risk prediction
Machine and deep learning for susceptibility mapping and prediction of land subsidence, settlement of structures on soft soils, and slope and riverbank instability.- Random Forest
- XGBoost
- SVM
- CNN
- LSTM
- Transformer
- InSAR
- GIS
Direction 2
Deep learning for urban infrastructure asset management
Computer vision and deep learning for automatic detection and classification of infrastructure defects from imagery and UAV; degradation forecasting and maintenance optimisation on WebGIS and BIM-GIS platforms.- YOLO
- U-Net
- Object detection
- Semantic segmentation
- UAV
- WebGIS
- BIM-GIS
- Predictive maintenance
Direction 3
Machine learning for construction project and investment risk management
Prediction and control of schedule and cost overrun risk, quality and safety assessment for infrastructure projects, combined with NLP analysis of project documents and multi-criteria decision making.- Risk prediction
- Cost overrun
- Schedule control
- NLP
- AHP
- MCDM
- BIM
Direction 4
Spatial data platform, MLOps and decision support
Shared training datasets, model training and deployment pipelines (MLOps), and decision support systems built on explainable AI (XAI).- Dataset
- MLOps
- XAI
- Decision support system
- pgvector
- Spatial database
- Risk map