A Stacking Ensemble Model for K_OC Prediction and Environmental Mobility Assessm

Global mobility assessments for 129,875 neutral organic compounds (NOCs) from 12 national and regional inventories worldwide. The dataset contains predictions from a descriptor-light stacking model (DL-SM) trained on 1987 NOCs, achieving a test R² of 0.825. The model was developed by Fu Liu and published on figshare in March 2026.

Use Cases

• Predicting organic carbon sorption coefficient (K_OC) for neutral organic compounds based on 13 molecular descriptors.
• Assessing the intrinsic mobility potential of chemicals across global inventories based on predicted mobility scores.
• Analyzing the relationship between molecular size (e.g., nC ≤ 15) and chemical mobility.
• Identifying neutral persistent, mobile, and toxic (PMT) candidates for regulatory management.

Strengths

• Model trained on 1987 neutral organic compounds (NOCs) and validated on an independent test set.
• Predictions generated for a large set of 129,875 NOCs from 12 global inventories.
• Model performance reported with a specific test R² of 0.825.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the primary prediction dataset is unknown, which may limit suitability assessment.

Provenance

Source

Fu Liu via figshare.

Collection Method

Predictions generated by a machine learning stacking ensemble model applied to chemical inventories.

Time Range

Model development and prediction date is circa 2026.

Freshness

Last updated 2026-03 25 21:04:05.

Geography

Global, based on 12 national and regional chemical inventories.