Phishing Websites Subset: 2,000 Rows and 100 Features for Classification

A 2,000-row, 100-feature subset of the PhishingWebsites dataset, generated via a controlled subsampling process with a random seed of 3. The data, originally from OpenML, was created by Eddie Bergman and is released under a US public domain license. It contains 10 target classes, sampled using a stratified method.

Use Cases

• Training phishing website classifiers based on the 100 available features.
• Benchmarking model performance on a stratified subset with 10 target classes.
• Feature selection analysis on a controlled sample of web security attributes.

Strengths

• Contains 2,000 rows, providing a substantial sample for model training.
• Includes 100 features, offering a range of potential predictors for classification.
• Subsampling was stratified, which likely helps maintain class distribution.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count is known, but the original dataset's full context and potential biases are not detailed.

Provenance

Source

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

Collection Method

Algorithmic subsampling with specified parameters for rows, columns, and classes.