Performance Evaluation on Imbalanced UCI Bank Marketing Dataset

Fahim Nasir created a dataset evaluating machine learning models on imbalanced class distributions for bank marketing customer conversion forecasting. The dataset, last updated on 2026-05-15, contains results from a two-stage evaluation framework testing ensemble and deep learning models with five sampling strategies. The framework assesses models across discrimination, calibration, computational cost, and explainability to select suitable model-sampling combinations.

Use Cases

• Comparing the performance of XGBoost with Borderline2SMOTE sampling against other model-sampling combinations.
• Evaluating the trade-offs between discrimination, calibration, and computational cost introduced by synthetic sampling techniques.
• Identifying key drivers of customer responses using SHAP (Shapley Additive exPlanations) for marketing strategy development.
• Benchmarking ensemble learning and deep learning models on imbalanced datasets using k-fold cross-validation and hyper-parameter tuning.

Strengths

• Dataset is licensed under CC-BY-4.0, allowing for open sharing and adaptation.
• Evaluation framework assesses models across four criteria: discrimination, calibration, computational cost, and explainability.
• Includes results from testing five distinct sampling strategies, including Borderline2SMOTE.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Dataset size is 5.5 KB, indicating a very limited scope, likely containing summary results rather than raw data.

Provenance

Source

figshare

Collection Method

Results from a two-stage quadrilateral evaluation framework applied to a UCI dataset.

Freshness

Last updated 2026-05-15 17:43:51; freshness should be verified.