Integrating biosorption and machine learning for efficient remazol red removal by algae-bacteria co-culture and comparative analysis of predicted models

Abstract

This research investigates into the efficacy of algae and algae-bacteria symbiosis (ABS) in efficiently decolorizing Remazol Red 5B, a prevalent dye pollutant. The investigation encompasses an exploration of the biosorption isotherm and kinetics governing the dye removal process. Additionally, various machine learning models are employed to predict the efficiency of dye removal within a co-culture system. The results demonstrate that both Desmodesmus abundans and a composite of Desmodesmus abundans and Rhodococcus pyridinivorans exhibit significant dye removal percentages of 75 ± 1% and 78 ± 1%, respectively, after 40 min. The biosorption isotherm analysis reveals a significant interaction between the adsorbate and the biosorbent, and it indicates that the Temkin model best matches the experimental data. Moreover, the Langmuir model indicates a relatively high biosorption capacity, further highlighting the potential of the algae-bacteria composite as an efficient adsorbent. Decision Trees, Random Forest, Support Vector Regression, and Artificial Neural Networks are evaluated for predicting dye removal efficiency. The Random Forest model emerges as the most accurate, exhibiting an R² value of 0.98, while Support Vector Regression and Artificial Neural Networks also demonstrate robust predictive capabilities. This study contributes to the advancement of sustainable dye removal strategies and encourages future exploration of hybrid approaches to further enhance predictive accuracy and efficiency in wastewater treatment processes.

Keywords: Algae-bacteria dye removal; Artificial neural network; Biosorption; Levenberg marquardt; Machine learning; Support vector regression.