The comparative study of microbial classification using machine learning and neural network models based on PCA for feature generation

The classification of microorganisms is an important branch of microbiology and plays a crucial role in understanding microbial diversity. Microorganisms include bacteria, viruses, fungi, protozoa, algae, etc, and they exhibit significant differences in morphology, physiology, and ecological niches. By classifying microorganisms with similar characteristics into groups, it aids in studying both their common properties and differences. In this study, PCA, GBDT, and XGBoost were used to generate new features, and machine learning methods (logistic regression, decision trees, and random forests) were compared with neural networks to determine the best predictive model and feature engineering approach. According to the experimental results, in machine learning models, the random forest model with three new features generated by PCA performed the best. In neural network models, the BP neural network with two new features generated by PCA performed the best.