Metody strojového učení pro klasifikaci KTG záznamů

Abstract

The aim of this thesis is to find, examine and compare the various scientific works that also deal with machine learning on cardiotocographic records. These records are compared both among themselves and between this work and the results obtained. Overall, the thesis is organized into several chapters, where first the theory regarding the cardiotocograph itself is written. Its principles and functionality are described in Chapter 1. In Chapter 2, the most commonly used methods for machine learning, such as ANN, k-NN, RF, or SVM methods, are summarized. Chapter 3 describes the two main databases, then the research works that deal with this issue are described and summarized, and their results are compared with each other. The fourth chapter contains descriptions of the dataset used and describes the training procedure of the chosen methods – ANN and ANN optimized. Subsequently, the fifth chapter describes the implementation of the method and lastly, a comparison is made between the previously mentioned works and this one and a discussion with the obtained results.