Hyperparameter Optimisation Results for Reservoir Flood Control Scenario Reduction

5.5 KB of experimental results evaluating four distance measures for scenario reduction in reservoir flood control. The dataset, authored by Ja-Ho Koo and shared under CC-BY-4.0, contains results from a study using a Bayesian Neural Network for inflow scenario generation. It was last updated on 2026-05-27.

Use Cases

• Comparing the performance of Manhattan, Euclidean, Wasserstein, and energy distances for scenario reduction based on the study's evaluation.
• Evaluating the trade-off between preserving extreme inflow scenarios and computational cost based on the described custom index.
• Benchmarking scenario generation models for reservoir inflow dynamics based on the described Bayesian Neural Network approach.

Strengths

• The dataset is small (5.5 KB), facilitating quick download and inspection.
• The accompanying description provides specific performance metrics (NSE, RMSE) and a clear methodological context.
• It includes a direct comparison of four distinct distance measures with noted computational complexities.

Limitations

• The dataset's scale is very limited (5.5 KB), suggesting a narrow scope of results.
• Row count and column-level documentation are absent, requiring manual inspection after download.
• The description notes the underlying model has limitations in predicting peak inflows due to data scarcity.

Provenance

Source

Ja-Ho Koo via figshare.

Collection Method

Results from a study applying a Bayesian Neural Network for scenario generation and evaluating distance measures for reduction.

Freshness

Last updated 2026-05-27 17:41:37; freshness should be verified.