Development of a Prediction Algorithm using Boosted Decision Trees for Earlier Diagnoses on Obstructive Sleep Apnea

Sim, Doreen Ying Ying (2018) Development of a Prediction Algorithm using Boosted Decision Trees for Earlier Diagnoses on Obstructive Sleep Apnea. PhD thesis, Universiti Malaysia Sarawak (UNIMAS).

[img] PDF (Please get the password from ACADEMIC REPOSITORY UNIT, ext: 082-583932/ 082-583914)
Development of a Prediction Algorithm ft.pdf
Restricted to Registered users only

Download (6MB)

Abstract

This research develops a knowledge-based system by using computational intelligent approaches based on Boosting algorithms on decision trees augmented by pruning techniques and Association Rule Mining. This system can provide better prediction accuracies and speedier medical analyses in order to help medical doctors in the earlier clinical diagnoses of Obstructive Sleep Apnea, i.e. OSA. The prediction algorithms developed are based on the OSA datasets collected mainly from the public hospitals in Selangor, Malaysia. The proposed OSA questionnaires have been newly designed after the data collection has been completed. The newly proposed and designed OSA questionnaires are customizable to best fit for Malaysians and have significant differences with the internationally standardized OSA questionnaires since these questionnaires are tailor-made based on the raw data collected within populations in Malaysia only. The parameters involved in the prediction algorithms developed are based on common OSA risk factors and visual-inspected variables found in the patients’ records in the OSA datasets collected. As benchmarked comparisons and contrasts, the performance of the computational intelligent system developed in this research is testified with quite a number of standard online databases downloaded mainly from the University of California Irvine data repositories. This is to ensure the generalized performance, flexible strengths and robustness of the OSA prediction system developed. Research outputs showed that there is a significant prediction improvement of the algorithms and system developed based on all types of datasets being accessed against the classical boosting approaches on the same datasets. There is a stepwise prediction improvement from the classical approach of Boosted Decision Trees to the developed Boosted Pruned-Decision Trees and then to Boosted Pruned-Association-Rule-Minded-Decision Trees. The developed prediction iv algorithms have been proven to help medical doctors with earlier clinical diagnoses on OSA cases, especially in Malaysia.

Item Type: Thesis (PhD)
Additional Information: Thesis (Ph.D) -- Universiti Malaysia Sarawak, 2018.
Uncontrolled Keywords: Boosting algorithms; pruning techniques; Association Rule Mining; prediction algorithms; OSA questionnaires; OSA risk factors; visual-inspected variables, unimas, university, universiti, Borneo, Malaysia, Sarawak, Kuching, Samarahan, ipta, education, Postgraduate, research, Universiti Malaysia Sarawak.
Subjects: B Philosophy. Psychology. Religion > BF Psychology
Divisions: Academic Faculties, Institutes and Centres > Faculty of Cognitive Sciences and Human Development
Depositing User: Gani
Date Deposited: 27 Aug 2019 08:28
Last Modified: 09 Jun 2020 03:17
URI: http://ir.unimas.my/id/eprint/26594

Actions (For repository members only: login required)

View Item View Item