Exploring spatial and temporal importance of input features and the explainability of machine learning-based modelling of water distribution systems
Bibliographic record
Abstract
Ensuring safe drinking water necessitates advanced management and monitoring techniques for water quality in distribution systems. This study leverages machine learning (ML) to model chlorine decay in a water distribution system (WDS) in British Columbia, Canada. A four-layer long short term memory (LSTM) network was trained to predict chlorine concentrations at a reservoir >24,000 m from the treatment plant. Explainable AI (XAI) techniques were applied to the trained network to address critical issues, such as enhancing the transparency and reliability of ML models. Several XAI methods were used to investigate the importance of sensor placement, identify the most significant features, understand feature ranges that result in poor performance, and validate model logic. Results demonstrated that for ML-based WDS control, sensor location is not critical, with high prediction accuracy achieved (mean absolute error <0.025 mg/L) even when exclusively using data from nodes spatially distant from the prediction site. XAI techniques showed the capability of identifying essential features and demonstrated that the behaviour of the ML model conformed with the expectations of chlorine behaviour. Superfluous variables were ranked low in importance, and the model learned fundamental aspects of chemical kinetics, such as temperature dependence and decay rate. Most importantly, the XAI methods applied showed the capability to communicate the reasoning for specific predictions, even at a local or sample-specific level. This study underscores the importance of transparency and trust in ML models, especially as the field transitions towards digital twin and Internet of Things (IoT) technologies, to enhance the effective management of water quality systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".