KDD Cup 2009 Appetency: Customer Churn Prediction Dataset

230 features from the KDD Cup 2009 small training set, curated for evaluating predictive models on independent and identically distributed tabular data. The dataset uses the 'appetency' task as a binary classification label and originates from a 2008 Orange customer database competition. It was prepared by the TabArena team for their Tabular ML IID Study.

Use Cases

• Benchmarking binary classification algorithms based on the appetency churn prediction task.
• Evaluating model robustness to missing values, which are described as abundant.
• Studying feature engineering techniques on data where feature names and categorical values lack semantic meaning.

Strengths

• Derived from a well-known academic competition (KDD Cup 2009), providing a standard benchmark.
• Contains 230 features after curation, offering a moderate-dimensional tabular problem.
• Explicitly curated for IID tabular ML evaluation, aligning with a specific research study.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.
• The description notes many missing values and anonymized features, which complicates interpretation.

Provenance

Source

https://kdd.org/kdd-cup/view/kdd-cup-2009/Intro

Collection Method

Curated from the original KDD Cup 2009 small training data by the TabArena team, with empty columns removed.

Time Range

2008

Freshness

Dataset Year is 2008; last update date is unknown.

Geography

null