Phishing Websites Subset: 2000 Rows and 100 Features for Classification

A 2000-row, 100-feature subset of the PhishingWebsites dataset, created via a reproducible subsampling script with a random seed of 1. The data was originally authored by Eddie Bergman and is shared under a US public domain license. The subsampling process selected up to 10 classes and used stratified sampling.

Use Cases

• Training binary or multi-class classifiers to identify phishing websites based on the 100 available features.
• Benchmarking feature selection algorithms on a reduced-dimensionality cybersecurity dataset.
• Studying the impact of stratified sampling on model performance for imbalanced security datasets.

Strengths

• Subsampling parameters are explicitly documented: seed=1, nrows=2000, ncols=100, nclasses=10.
• The creation method is reproducible, with the full source code for subsampling provided in the description.
• Stratified sampling was used, which likely helps preserve class distribution in the subset.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the original full dataset is unknown, which may limit suitability assessment of the subset.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML, derived from the 'PhishingWebsites' dataset (ID 4534).

Collection Method

Algorithmic subsampling from a larger parent dataset using a specified random seed and parameters.