QSAR-Biodeg: Subsampled Molecular Properties for Biodegradability Prediction

A 2000-row, 100-column subsample of the QSAR-biodeg dataset, generated with a random seed of 0 and stratified sampling. The dataset was created by Eddie Bergman and is licensed as public domain in the United States. It likely contains molecular descriptors for modeling chemical biodegradability.

Use Cases

• Training classification models to predict chemical biodegradability based on molecular descriptors.
• Benchmarking feature selection algorithms on a dataset with 100 sampled molecular properties.
• Developing stratified sampling techniques for imbalanced chemical datasets.
• Analyzing the relationship between specific molecular features and environmental persistence.

Strengths

• Subsampling parameters are explicitly defined: seed=0, nrows=2000, ncols=100, nclasses=10.
• Stratified sampling was used, which likely preserves class distribution from the original dataset.
• Derived from a known QSAR dataset focused on a specific, applied domain (biodegradation).

Limitations

• The specific 100 columns and their semantics are unknown, requiring inspection after download.
• Row count is defined by the subsampling script but the original dataset's total size is not provided.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Subsampling of the 'qsar-biodeg' dataset (OpenML ID 1494).

Collection Method

Algorithmic subsampling using a Python script with random column and row selection, and stratified sampling.