TRAP@NCI

Application of Machine Learning Techniques to classify Fetal Hypoxia

Mishra, Krishna Mohan (2016) Application of Machine Learning Techniques to classify Fetal Hypoxia. Masters thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Master of Science)
Download (1MB) | Preview

Abstract

Aim of this research is to classify fetal hypoxia using machine learning approach based on Cardiotocography (CTG) data and patient's previous complications records. Classification method is very popular in analysing new born baby's health in critical cases. CTG data had been used by obstetricians for analysing fetal well-being during pregnancy complications which provides fetus information in detail. CTG data consist of fetal heart rate (FHR) signals from mothers abdomen in the form of continuous electronic time varying pattern. CTG data provides information about FHR which is used for prior prediction and diagnosis of long term embryo impairment. In this study high dimensional CTG data is used to classify fetal hypoxia having attributes like FHR and uterine contraction (UC) in normal and pathological terms. CTG data has been taken from UCI machine learning repository consist of 1832 instances and 21 attributes used in this study out of which 8 are continuous and 13 are discrete. Research follows stacked generalised ensemble approach consist of 6 machine learning classifiers (Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), C4.5 Decision Tree Classier (J48), Simple Logistics (SL), Adaptive Boosting (AdaBoost)) for CTG data classification. Robustness of the model has been evaluated by 10-fold cross validation technique with accuracy, sensitivity and specificity as evaluation features for model. F-measure, ROC (receiver operation characteristics) and Kappa Statistics also used as tools to study test performance. We also compared Model Testing Time for each individual model with accuracy to find out which model performs best. Highest accuracy of 98.79 % provided by Stacking 2 (SL+C4.5(J48)+RF) model. If experienced obstetricians are not available this research will help physicians in prior detection and diagnosis of fetal hypoxia.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: CAOIMHE NI MHAICIN
Date Deposited: 27 Jan 2017 15:55
Last Modified: 27 Jan 2017 15:58
URI: http://trap.ncirl.ie/id/eprint/2522

Actions (login required)

View Item View Item