Comparative Feature Selection Methods for Flood Susceptibility Mapping in Khuzestan, Iran

Mohammad Kazemi's dataset provides a comparative table of nine feature selection methods applied to flood susceptibility mapping. The data, last updated in April 2026, originates from a study using 19 environmental factors and 1,000 sample points from flood-prone Khuzestan Province, Iran. It details the consensus selection of key predictors like NDVI and daily minimum temperature, and the performance of metaheuristic-optimized LSTM models.

Use Cases

• Benchmarking feature selection algorithm performance based on the nine listed methods (e.g., Boruta, RFE, Mutual Information).
• Training or validating flood prediction models using the identified critical predictors like NDVI and Daily Minimum Temperature.
• Studying consensus-based variable selection strategies for geospatial datasets.
• Comparing the efficacy of metaheuristic optimization algorithms (WOA, GWO, OOA, CSA, HOA) for tuning deep learning model hyperparameters.

Strengths

• Compares nine distinct feature selection methods, providing a direct performance benchmark.
• Based on a study using 1,000 specifically sampled data points (500 flood, 500 non-flood).
• Identifies concrete influential variables (NDVI, Daily Minimum Temperature) through a formalized consensus rule.
• Includes validation results with specific performance metrics (F1-Score of 0.88, Cohen’s Kappa of 0.75).

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (5.5 KB), indicating it is likely a summary table rather than the underlying raw spatial data.

Provenance

Source

Mohammad Kazemi via figshare.

Collection Method

Features were sourced from Google Earth Engine, with model development using sampled points from Khuzestan Province, Iran.

Freshness

Last updated 2026-04-29 17:42:19; freshness should be verified.

Geography

Khuzestan Province, Iran.