Component-Wise Ablation Analysis: Bank Marketing Data for Imbalanced Classification

A 5.5 KB Excel dataset supports a study proposing a two-stage quadrilateral evaluation framework for forecasting customer conversion in bank marketing. Authored by Fahim Nasir and last updated in May 2026, the data was used to test ensemble and deep learning models under five sampling strategies to address class imbalance. Results indicate the XGBoost model with Borderline2SMOTE sampling showed improved adaptability with reduced synthetic distortion.

Use Cases

• Evaluating sampling techniques like Borderline2SMOTE for imbalanced classification based on the described framework.
• Benchmarking model performance across discrimination, calibration, cost, and explainability metrics as outlined in the study.
• Applying SHAP explanations to identify key drivers of customer responses for predictive inference.
• Selecting context-aware model-sampling combinations for responsible predictive analytics in banking.

Strengths

• Dataset is openly licensed under CC-BY-4.0.
• The accompanying research describes a specific, reproducible evaluation framework for imbalanced data.

Limitations

• The dataset is very small at 5.5 KB, suggesting limited scope or a summary of results rather than raw data.
• Column-level documentation is absent; field semantics must be inferred from the research description.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Likely generated or compiled as part of the described research methodology.

Freshness

Last updated 2026-05-15 17:43:58; freshness should be verified.