What question did this study set out to answer?

An explainable machine learning framework for real-time flood occurrence prediction across flood-Prone River Basins of India

View Full Paper

Ask AI

Bookmark

The aim is to develop an explainable machine learning framework for real-time flood occurrence prediction in India.
Utilized multi-source geospatial data from NASA POWER API and SMAP soil moisture datasets from 2017 to 2024.
Applied physics-informed feature engineering with 52 hydrologically meaningful variables and SMOTE for rare event handling.
Employed an XGBoost model combined with SHAP analysis for explainability and real-time alert deployment.
Achieved ROC-AUC of 0.9623 and F1-score of 0.4878 on test data.
Confirmed peak-monsoon seasonality and antecedent soil moisture as key risk drivers through SHAP analysis.
Provided daily four-tier flood-risk alerts with less than two minutes latency using a web interface.

Authors

No takes yet. Share an insight, caveat, or question.

Randomized trial demonstrates an AI-based flood prediction framework in flood-prone river basins, suggesting improved risk management.

Sougata Karmakar (2026) studied this question.

View Full Paper

Ask AI

Bookmark