New Perspective for the Prediction of Pollutant Removal Efficiency in Constructed Wetlands: Using a Genetic Algorithm-Assisted Machine Learning Model

Constructed wetlands (CWs) are widely used for wastewater treatment, but their performance is difficult to predict due to varying factors like local weather, hydraulic conditions, vegetation, and wastewater composition. Here, we proposed a model method for predicting CW processing efficiency based on published literature simulations using machine learning methods. Through data mining, we divided the obtained variables into six different categories and proposed different data repair strategies for each category. To improve the model performance, a genetic algorithm-assisted database dimensionality reduction method was introduced in the model destruction. After model selection and hyperparameter optimization, the random forest algorithm was selected as the final algorithm, and the model performances for all four predictions (ammonia nitrogen, total nitrogen, total phosphorus, and chemical oxygen demand removal efficiency) were 0.9405, 0.8277, 0.8136, and 0.8877, respectively. Generally, the magnitude of the influence of the different categories is listed in the following order: meteorology/location > hydraulic condition > substrate property ≈ water quality ≈ reactor size > vegetation. Based on this work, the future design and operation of CWs might find an efficient and environmentally friendly method that could ideally maximize pollution control and economic benefits at the same time.