Exploring the effects of different combinations of predictor variables for the treatment of wastewater by microalgae and biomass production

Abstract

Microalgae provide an alternative and cost-efficient method for treating wastewater with simultaneous recovery of resources in the form of biomass. In order to increase the biomass productivity and wastewater treatment capability of microalgae, it is essential to provide the right combination of predictor variables. However, it is practically impossible to determine the best combination of variables through conventional methods from a large dataset of published research. Machine learning algorithms can easily detect the pattern in the large dataset and provide the best combination as desired. In the present analysis, decision tree algorithm was used to determine the effects and the best combination of predictor variables, including microalgae class, pre-cultivation stage deciding factors and operating variables, resulting in high biomass productivity and wastewater treatment capability. Decision tree analysis detected 10 different combinations of predictor variables leading to high nitrogen removal efficiency, 10 combinations for high phosphorus removal efficiency and 8 combinations for increased biomass production. These combinations were tested on recently published experimental findings and nearly 80% accuracy was obtained. The results obtained through machine learning analysis can be used in constructing high throughput experimental designs, which may assist in carrying out the efficient wastewater treatment at large scale. © 2021 Elsevier B.V.