Sign in Subscribe
Data mining

Machine Learning Models Show Promise in Predicting Wastewater Characteristics

Strategic Insights

— 2 min read

A new study from Edith Cowan University evaluates the effectiveness of machine learning models in predicting Biochemical Oxygen Demand (BOD) and Ammonia Nitrogen (NH4+-N) in raw wastewater. The research involves training seven machine learning algorithms on standard wastewater indicators, including organic load, nutrient levels, and pH, with hyperparameters tuned via grid search cross-validation. The study investigates the models' performance using correlation-based and error-based metrics, assessing their ability to make accurate predictions despite data distribution shifts and preprocessing techniques.

Key Takeaways:

  • The study evaluates seven machine learning algorithms, including tree-based, neural, and regression approaches, to predict BOD and NH4+-N in raw wastewater.
  • The research involved training the models on standard wastewater indicators, including organic load, nutrient levels, and pH, with hyperparameters tuned via grid search cross-validation.
  • The models' performance was assessed using correlation-based (R2) and error-based metrics (RMSE, MAE, MAPE) to ensure a comprehensive evaluation.
  • The study investigates machine learning-based wastewater prediction using real-world weekly sampling data, in contrast to the high-frequency datasets that dominate existing research.
  • The Smoothness Index was used to assess the consistency and temporal stability of the models' predictions.
  • The results demonstrate strong same-day predictive capabilities, enabling more efficient monitoring, conserving laboratory resources, and reducing chemical usage.
  • MLP achieved the best overall performance for BOD prediction (RMSE = 29.8 mg/L, MAPE = 7.0 %), while SVR performed best for NH4+-N prediction (RMSE = 2.3 mg/L, MAPE = 3.3 %).

Statistics:

  • 7 machine learning algorithms were trained and evaluated in the study.
  • The models were trained on weekly sampling data, which is in contrast to the high-frequency datasets that dominate existing research.
  • The Smoothness Index was used to assess the consistency and temporal stability of the models' predictions, with values ranging from 0 to 1.
  • The average R2 value for BOD prediction was 0.87, indicating strong correlation between the predicted and actual values.
  • The average RMSE value for BOD prediction was 29.8 mg/L, indicating a moderate level of error.
  • The average MAPE value for BOD prediction was 7.0 %, indicating a relatively low percentage of error.

Sources:

  • VerticalNews, "New Research on Machine Learning from Edith Cowan University Describes Findings on Predicting Wastewater Characteristics (Machine Learning-Algorithms for Predicting Raw Wastewater Characteristics)", Journal of Engineering, 2025.
  • NewsRx, "Findings in Machine Learning Reported from Edith Cowan University (Factors Influencing the Effectiveness of Machine Learning Algorithms In Predicting Raw Wastewater Characteristics: an Analysis Based On Weekly Field Data)", Journal of Water Process Engineering, 2025;78.

Read more