Graph-learning-based machine learning improves prediction and cultivation of commercial-grade marine microalgae Porphyridium.

[Display omitted] • A graph-learning-based BANE model improves the prediction performances. • The BANE-XGBoost has the higher performance improvement and best prediction. • Illumination intensity and KH 2 PO 4 (TOP 2) dominate the growth of Porphyridium. • SHAP-based heatmap and group reach the combined roles of cultivation parameters. • The one-way and two-way PDP models find the optimal factors for the two targets. A graph learning [Binarized Attributed Network Embedding (BANE)] model enhances the single-target and multi-target prediction performances of random forest and eXtreme Gradient Boosting (XGBoost) by learning complex interrelationships between cultivation parameters of Porphyridium. The BANE-XGBoost has the best prediction performance (train R2 > 0.96 and test R2 > 0.87). Based on Shapley Additive Explanation (SHAP) model, illumination intensity, culture time, and KH 2 PO 4 are the most critical factors for Porphyridium growth. The combined facilitating roles of cultivation parameters are found using the SHAP value-based heat map and group. To reach high biomass and daily production rate concurrently, one-way and two-way partial dependent plots models find the optimal conditions. The top 2 critical parameters (illumination intensity and KH 2 PO 4) were selected to verify using the graphical user interface website based on the optimized model and lab experiments, respectively. This study shows the graph-learning-based model can improve prediction performance and optimize intricate low-carbon microalgal cultivation. [ABSTRACT FROM AUTHOR]