Skip to content

Latest commit

 

History

History
42 lines (33 loc) · 3.15 KB

README.md

File metadata and controls

42 lines (33 loc) · 3.15 KB
Raw

XAI - Lateral Spreading

This project investigates the application of eXplainable AI (XAI) techniques on predictive machine learning models for lateral spreading phenomena. We have developed multiple XGBoost models using a dataset sourced from Durante and Rathje (2022). The repository provides resources for data preprocessing, model training, and interpretation using SHAP (SHapley Additive exPlanations) explainers.

Folder Structure

data Folder: Contains both the original and processed datasets. The original dataset, derived from Durante and Rathje (2021), comprises 6,500 datapoints from Christchurch, New Zealand, pertaining to the 2011 Christchurch Earthquake. It includes various features such as geometry features, event-specific features like groundwater depth (GWD) and peak ground acceleration (PGA), CPT (cone penetration test) related features, and binary indicators for lateral spreading. Refer to Table 1 for a breakdown of features used in each model.

Table 1. Summary of features used in each XGBoost model.

Model L
(km)		 GWD
(m)		 PGA
(g)		 Elevation
(m)		 Slope
(%)		 Ic
(med)		 Ic
(std)		 qc1Ncs
(med)		 qc1Ncs
(std)
A ✓ ✓ ✓ ✓ ✓ O O O O
B ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
C ✓ ✓ ✓ ✓ O ✓ ✓ O O

model_development Folder: Includes Jupyter notebooks for data preprocessing (data_preprocessing.ipynb) and XGBoost model training (xgb_training.ipynb). The data preprocessing notebook loads the dataset, performs data splitting, and feature selection according to Table 1, saving the processed data as pickle files (data_x.pkl) in data folder. The XGBoost training notebook demonstrates the model training process and saves the trained models as pickle files (opt_XGB_X.pkl) in the xgb_models folder.

model_usage Folder: Contains Jupyter notebooks (shap_explainer_X.ipynb) for generating SHAP explanations for each XGBoost model. These notebooks load the trained models and corresponding data to create SHAP visualizations.

xgb_models Folder: Stores the trained XGBoost models developed from different datasets in the data folder.

References

Durante, M. G. and Rathje, E. (2022). Machine learning models for the evaluation of the lateral spreading hazard in the Avon river area following the 2011 Christchurch earthquake. doi:10.17603/DS2-3ZDJ-4937

Durante, M. G. and Rathje, E. M. (2021). An exploration of the use of machine learning to predict lateral spreading. Earthquake Spectra 37, 2288–2314. doi:10.1177/87552930211004613

Citation

@software{krishna_kumar_2024_11003110,
  author       = {Krishna Kumar and
                  Cheng-Hsi Hsiao},
  title        = {geoelements/xai-lateral-spreading: v1.0.1},
  month        = apr,
  year         = 2024,
  publisher    = {Zenodo},
  version      = {v1.0.1},
  doi          = {10.5281/zenodo.11003110},
  url          = {https://doi.org/10.5281/zenodo.11003110}
}