HTRU2: Pulsar Candidates from the High Time Resolution Universe Survey

16,259 spurious examples and 1,639 real pulsar examples collected during the HTRU survey. The dataset describes pulsar candidates for binary classification, with all examples checked by human annotators. It was created by Dr Robert Lyon from the University of Manchester using the PulsarFeatureLab tool.

Use Cases

• Train binary classification models to distinguish real pulsars from radio frequency interference based on integrated pulse profile statistics.
• Benchmark anomaly detection algorithms on a dataset with a significant class imbalance between pulsar and non-pulsar candidates.
• Develop feature engineering techniques for astronomical signal processing using the eight continuous variables extracted from candidate files.

Strengths

• Contains 17,898 total examples with verified labels from human annotators.
• Includes 1,639 confirmed real pulsar examples, representing a scientifically interesting minority class.
• Data is presented in both CSV and ARFF formats for compatibility with different tools.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Dr Robert Lyon, University of Manchester, School of Physics and Astronomy.

Collection Method

Candidates collected during the High Time Resolution Universe Survey (South) and features extracted using the PulsarFeatureLab tool.