Volatile Compound Sensor Data for Machine Learning-Based Property Prediction

A single-sensor approach identifies volatile organic compounds by analyzing non-equilibrium mass-transport dynamics. Machine learning models trained on the sensor's time-dependent reflectance spectra can predict chemical properties like boiling point and vapor pressure. The data originates from research at Harvard University, implementing both passive and active sniffing modalities.

Use Cases

• Classifying polar and nonpolar volatile organic compounds based on time-dependent sensor response patterns.
• Determining mixing ratios of binary chemical mixtures using non-equilibrium mass-transport dynamics.
• Predicting physicochemical properties like boiling point and viscosity from sensor output for known and unknown compounds.
• Reducing data collection time by implementing a bioinspired active sniffing approach that controls analyte delivery.

Strengths

• Data enables property prediction for compounds unknown to the model, suggesting generalizable features.
• Implements an active sniffing modality that reduces data collection and analysis duration to seconds.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

Harvard University

Collection Method

Sensor readings from a mesoporous one-dimensional photonic crystal, with data processing functions available on GitHub.