Jackysyj/ml-water-treatment-deployment-gap: V2.0

This repository contains the data and analysis code supporting the npj Clean Water Comment "Operational Evidence Standards for Machine Learning in Wastewater Treatment" (Jiang, Yang, Zhang, Mao & Cheng). Contents dataset_423_analysis_corpus.csv — the 423-study full-text analysis corpus with LLM-extracted structured fields (deployment, real-time testing, validation timing, uncertainty quantification, reported R², algorithms, sub-field, data source and scale). source_extraction_pool_500.csv — the 500-record source extraction pool from which the 423-study full-text corpus is derived. derived_outputs/ — figure source data and anonymised six-plant derived outputs (per-plant and per-target performance, walk-forward and random-CV summaries, effluent variability), sufficient to reproduce all reported figures and summary statistics. benchmark/ — LLM extraction benchmark summaries and the 50-paper gold-standard evaluation. prompts/ — the extraction and screening prompts. manual_review/ — manual full-text verification records for the load-bearing deployment and validation fields. scripts/ — corpus cleaning, extraction-reliability checks, six-plant analysis, figure generation and manuscript validation code. Reproducibility. The released anonymised derived outputs reproduce the manuscript's evidence map (423 studies; plant deployment 2.8%, real-time testing 5.2%, future-facing validation 18.2%, uncertainty quantification 8.5%; median reported R² 0.97 among 202 studies) and the bounded six-plant operational illustration (6,006 timestamped records; median feedforward R² 0.063 under random cross-validation and 0.029 under walk-forward evaluation across 29 plant-target tasks). Data confidentiality. Raw plant records are not publicly available because they are covered by utility confidentiality agreements and could identify operating facilities. The anonymised derived outputs (relative time, no facility names) are released here and are sufficient to reproduce the reported plant-level results. License: CC-BY-4.0. Related code: https://github.com/Jackysyj/ml-water-treatment-deployment-gap (V2.0).