Machine learning optimization for microbial surfactant production
Description
This dataset contains the raw experimental data, source code, and generated outputs used for the development and evaluation of multiple linear regression (MLR), Random Forest (RF), and artificial neural network (ANN) models. It also includes the scripts used for data augmentation, multi-objective optimization, TOPSIS based ranking, and the figures presented in the associated study.
Files
Steps to reproduce
The script "d5_Carbon_Screening.m" uses the raw dataset "d5_carbon_source.mat" to generate the graph for comparing different carbon sources and identifying the optimal carbon source. The script "a2_Fit_lm_model.m" loads the Central Composite Design (CCD) raw dataset and develops the MLR model. Subsequently, "b0_Gendata_transformed.m" uses the same CCD dataset to generate an augmented dataset through data transformation and augmentation. This augmented dataset is then used by "b2_Bayesian_tree.m" and "b4_Bayesian_fitrnet.m" to develop the RF and ANN models, respectively. The script "b6_model_indic.m" calculates the performance metrics and evaluation indices for all developed models. Multi-objective optimization is performed by sequentially executing "c1_GA_Ind_min_max.m", "c3_ANN_multiobj.m", and "c4_Topsis.m", which normalize the objectives, optimize the process using a Genetic Algorithm, and rank the Pareto-optimal solutions using the TOPSIS. The script "e2_Biomass_calibration.m" uses the biomass calibration dataset to generate the regression plot showing the relationship between optical absorbance and cell dry weight. The remaining MATLAB scripts are used to generate the model performance plots, optimization results, explainable AI visualizations, visualize augmented data quality; and show all other relations between the inputs and outputs of CCD datasets.
Institutions
- Karunya UniversityTamil Nadu, Coimbatore