Plastic Waste-Based Activated Carbon Adsorption Optimization Model Using Machine Learning to Improve the Quality of Domestic Liquid Waste

Environmental pollution from plastic waste and domestic liquid waste requires innovative, sustainable solutions. This study explores the integration of plastic-based activated carbon adsorption technology with a machine learning approach for the optimization of domestic liquid waste treatment through the Systematic Literature Review method. A literature search was conducted on the Scopus, Web of Science, ScienceDirect, and Google Scholar databases for the period 2021-2025, resulting in 10 high-quality articles that were comprehensively analyzed. The synthesis results showed that activated carbon from PET, PVC, and chemical activated plastic waste produced a material with a BET surface area of 800-1500 m²/g and an adsorption capacity of 1086 mg/g comparable to commercial products. Specific surface area parameters and initial concentrations of pollutants were the most dominant factors with a combined contribution of more than 43% to the removal efficiency. Machine learning algorithms such as Extreme Gradient Boosting, Gaussian Process Regression, and multimodal deep learning achieve very high prediction accuracy with R² up to 0.99 and RMSE below 5 mg/g. Optimization based on metaheuristic algorithms increases adsorption capacity by up to 15.40% with material reusability reaching 70-90% after five cycles. Competitive production costs ($13.75/kg) and low carbon footprint (5.92 kg CO₂/kg) make this technology a sustainable solution that supports the circular economy. The framework developed provides a scientific foundation for the implementation of an efficient and environmentally friendly liquid waste treatment system