Assessing industrial wastewater effluent toxicity using boosting algorithms in machine learning: A case study on...

Area of Science:

• Environmental Science
• Artificial Intelligence
• Ecotoxicology

Background:

• Industrial wastewater effluent often contains persistent trace heavy metals, posing risks to aquatic ecosystems.
• Traditional ecotoxicity assessment methods, including animal testing and bioassays, are ethically problematic, costly, and time-consuming.
• There is a need for rapid, cost-effective, and reliable methods for assessing industrial wastewater ecotoxicity.

Purpose of the Study:

• To propose and validate an artificial intelligence-powered water quality assessment (AiWA) approach for predicting industrial effluent ecotoxicity.
• To enhance the speed and cost-effectiveness of ecotoxicity assessment processes.
• To identify key parameters influencing ecotoxicity for improved wastewater management.

Main Methods:

• Collected 99 wastewater samples from 21 diverse industries in the Republic of Korea.
• Measured 14 physicochemical and ecotoxicological parameters.
• Developed and compared boosting algorithms, specifically extreme gradient boosting (XGBoost) and adaptive boosting (AdaBoost), for predictive modeling.

Main Results:

• The XGBoost-based AiWA model achieved high prediction performance (R² > 0.94, RMSE = 3.5 toxicity units) for integrated toxicity units (ITU).
• Feature selection identified conductivity, copper, lead, selenium, pH, and zinc as critical predictors of ecotoxicity.
• Optimal conditions for non-toxic discharge (TU ≤ 1) were identified as pH 6.8–8.4 and conductivity < 1651 μS/cm.

Conclusions:

• The XGBoost-based AiWA model provides a significantly improved and efficient approach for predicting industrial effluent ecotoxicity.
• pH and conductivity are identified as crucial indicators for assessing ecotoxicity levels.
• The findings support rapid, cost-effective detection of heavy metal ecotoxicity, aiding wastewater management decisions.