Hybrid modelling of Bioprocess Dynamics

This repository contains MATLAB scripts for modeling, predicting, and analyzing microbial growth and polymer production. In the dynamic modeling step (Folder: Dynamic_modelling), experimental datasets are loaded, optimized kinetic parameters are estimated, and predictions are generated, with model performance evaluated via cross-validation, bootstrap, and error metrics. The hybrid modeling step (Folder: Hybrid_modelling) integrates neural networks with optimized parameters to simulate system behavior, assess parameter sensitivity, and generate feature explanations using LIME, Shapley, and partial dependence plots. The cross-validation of hybrid models (Subfolder: Cross_validation_hybrid_modelling) involves encoding and decoding network parameters, simulating biomass and polymer growth, calculating prediction errors, and evaluating model performance. Finally, the statistics calculation (Folder: Statistics_calculation) step consolidates results from both dynamic and hybrid models, computing MAE, RMSE, MAPE, and R² to quantify model accuracy and reliability for validation datasets.