Srovnání metod důležitosti rysů pro mikrobiologická data

This bachelor thesis focuses on the comparison of feature importance methods on microbiological data. The aim of the work is to compare the accuracy of the prediction model on a given dataset when manipulating the data according to the selected feature importance methods. Specifically, the SHAP and ARA methods are considered. The theoretical part of the thesis contains definitions of terms and description of the two selected feature importance methods. The methodological part describes how the research of the given issue was carried out, contains a general description of the methods used and their practical application in the practical part of the work. The practical part is implemented in the form of data processing using the Python language in Jupyter Notebook. These tools are used to measure the accuracy of the Catboost prediction model in adjusting the dataset according to importance using the SHAP and ARA methods, as well as factor analysis or Principal Component Analysis.