Klasifikace aktuálního stavu plodu z kardiotokografických záznamů

Abstract

The Cardiotocography is a diagnostic method, which measures fetal heart rate variability, fetal movement and uterine contractions. The main aim of this work is to improve recent results, which were achieved in the task of actual fetal state determination. Each fetal state can be classified to the one of three basic states. The states are: physiological, suspicios and pathological. These classes were determined by the international federation of obstetricians and gynecologists. Recent methods achieved unsatisfactory results. The Random Forest algorithm belongs to the group models and its options can improve these results. The second main aim is to reduce original feature space by the Principal Component Analysis or by the Correlation Feature Subset method. This reduction can bring improvement of recent results as well as it can simplify the complexicity of the model.