Using Machine Learning to Predict Environmental Parameters and Geographical Locations of Wastewater Treatment Plants Based on Activated Sludge Microbial Communities

Activated sludge is a key component of wastewater treatment systems. Microbial community composition of activated sludge is closely related to environmental parameters and the geographical location of wastewater treatment plants (WWTPs). In this study, we used a machine learning approach to predict environmental parameters and geographical locations of WWTPs based on microbial communities, aiming to improve the efficiency and quality of wastewater treatment. The results showed that eXtreme Gradient Boosting (XGBoost) was the most suitable method for the prediction of environmental parameters. The correlation between predicted values and actual values ranged from weak (R2 = 0.47) to strong (R2 = 0.76). Key operational taxonomic units (OTUs) and key phyla with an important influence on each environmental parameter were identified. Furthermore, geographical location could be well predicted using Random forest (RF) with an accuracy of 98.4%. Key OTUs were identified for each continent. Overall, our results demonstrated that environmental parameters and geographical locations of WWTPs are predictable. This study provided valuable information for optimizing wastewater treatment systems and selecting effective wastewater treatment methods.