Machine learning optimization of environmental factors influencing biomass and nutritional composition in local algal species

Abstract

Algae are recognized for their potential in biofuel production due to their high biomass yield, protein and lipid. This study investigates the influence of pH, temperature, light intensity, light colour and CO2 CO2. Machine learning on biomass and biochemical composition in five algal genera (Chlorella, Botryococcus, Chlamydomonas, Tetraselmis and Closterium). Algal samples were isolated from aquatic environments in KPK-Pakistan and cultured under controlled conditions. Environmental variables were systematically varied: pH, temperature, light intensity, light colour and CO2 concentration. Biochemical analyses revealed biomass ranging from 0.2 to 2.1 gl-1, lipids 7.2-24.5% and proteins 8-49.5%, with optimal conditions of pH 7, 30°C, red light, 3000 lux and 9% CO2. Machine learning was applied to optimize environmental conditions, with random forest (RF) identified as the most effective model. A novel metric, W_new, combining performance and error metrics, facilitated robust model evaluation and hyperparameter tuning. The model's feature importance analysis ranked CO₂ concentration and pH as the most influential factors. RF achieved R2 scores of 0.686 (training) and 0.534 (validation), demonstrating strong predictive performance. This study integrates experimental and computational approaches, providing a detailed framework for optimizing algal cultivation. We highlighted the utility of machine learning in enhancing biomass and lipid productivity, advancing the sustainable production of biofuel.