Machine Learning\nAssisted Image Analysis for Microalgae\nPrediction

Microalgae-based wastewater treatment\nhas resulted in\na paradigm\nshift toward nutrient removal and simultaneous resource recovery.\nHowever, traditionally used microalgal biomass quantification methods\nare time-consuming and costly, limiting their large-scale use. The\naim of this study is to develop a simple and cost-effective image-based\nmethod for microalgae quantification, replacing cumbersome traditional\ntechniques. In this study, preprocessed microalgae images and associated\noptical density data were utilized as inputs. Three feature extraction\nmethods were compared alongside eight machine learning (ML) models,\nincluding linear regression (LR), random forest (RF), AdaBoost, gradient\nboosting (GB), and various neural networks. Among these algorithms,\nLR with principal component analysis achieved an <i>R</i><sup>2</sup> value of 0.97 with the lowest error of 0.039. Combining\nimage analysis and ML removes the need for expensive equipment in\nmicroalgae quantification. Sensitivity analysis was performed by varying\nthe train–test splitting ratio. Training time was included\nin the evaluation, and accounting for energy consumption in the study\nleads to the achievement of high model performance and energy-efficient\nML model utilization.