Artificial intelligence-driven prediction models for the cultivation of Chlorella vulgaris FSP-E in food waste culture medium: A comparative analysis and validation of models

Abstract

Recent advancements in biotechnological processes have increasingly relied on machine learning (ML) to enhance efficiency particularly in optimizing microalgae cultivation using food waste as an alternative culture medium. This research work explores the influence of input features derived from food waste culture medium on the growth of Chlorella vulgaris FSP-E by utilizing machine learning models to predict the dry cell weight (DCW) with high precision. Significant correlations between DCW and variables such as concentration, total phosphorus (TP), total nitrogen (TN), and chemical oxygen demand (COD) were identified using scatter plot and heatmap analysis. The heatmap analysis revealed substantial positive correlations among concentration, TP, TN, and COD with DCW, underscoring the importance of these nutrients in fostering microalgae growth. Conversely, the proteins content exhibited a negative correlation with DCW, indicating their limited role in promoting biomass production. Among the machine learning models tested, the Artificial Neural Network (ANN) outperformed among other model in which the results exhibited an R² of 0.9897, MAE of 0.0632, MSE of 0.0098, and RMSE of 0.0992, respectively. The k-Nearest Neighbors (k-NN) and Support Vector Regression (SVR) models also demonstrated strong predictive capabilities, with k-NN achieving an R² of 0.9894, MAE of 0.0446, MSE of 0.0102, and RMSE of 0.1011, while SVR achieved an R² of 0.9844, MAE of 0.0921, MSE of 0.0150, and RMSE of 0.1225. Optimal hyperparameters for the ANN included the activation function, hidden layer sizes, solver and learning rate of ReLu, (32, 64), lbfgs, and constant, respectively. Validation confirmed ANN superior performance, demonstrating its potential for optimizing microalgae cultivation for sustainable biofuel production. This study underscores the potential of using machine learning models to optimize biomass production in microalgae cultivation, providing valuable insights for enhancing sustainable biofuel production.