Air-Pollution Evidence-Synthesis Analysis Dataset and Reproduction Code

This repository contains the derived study-level dataset and Python code used to reproduce the quantitative deployment-evidence analysis reported in the manuscript, “A Computational Evidence-Synthesis Framework for Engineering Performance and Deployment Orientation in Air Pollution Control Technologies.” The dataset contains 202 studies from the evidence-synthesis workflow. Deployment evidence was classified using the Deployment Evidence Index (DEI). Three studies had unclear deployment evidence and were excluded from the primary predictive analysis, leaving 199 classifiable studies: 187 studies with no demonstrated deployment (DEI 0–1) and 12 studies with demonstrated deployment (DEI 2–3). The Random Forest analysis uses technology class, removal efficiency, and pressure drop as predictors. The repository includes the Python reproduction script and the derived dataset required to reproduce the reported independent 80/20 holdout analysis and repeated stratified 5-fold cross-validation with 10 repeats. The repository does not include the demonstrative Physics-Guided Active Learning–Computational Fluid Dynamics (PGAL–CFD) implementation, CFD cases, meshes, solver files, UDFs, or ANSYS Fluent files.