Predicting the Pattern of Under-Five Mortality in Ethiopia Using Data Mining Technology: The Case of Butajira Rural Health Program

Be’emnetu, Tekabe (2012) Predicting the Pattern of Under-Five Mortality in Ethiopia Using Data Mining Technology: The Case of Butajira Rural Health Program. Masters thesis, Addis Ababa University.

PDF (Predicting the Pattern of under-Five Mortality in Ethiopia Using Data Mining Technology: The Case of Butajira Rural Health Program) Be’emnetu, Tekabe.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Introduction: The under-five deaths in Ethiopia represent 48% of all mortality. More than half of the under-five deaths occurred during the first year of life, and 53% of these before 2 months of age. Data mining is a collection of techniques for efficient automated discovery of previously unknown, valid, novel, useful and understandable patterns in large databases. Objective: The main objective of this study is to explore the potential applicability of data mining to predict the determinants, levels and pattern of under-five mortality in Ethiopia, particularly for the Butajira rural health program sites. This can greatly support for policy makers, planners, and healthcare providers working on the control of under-five children mortality in Ethiopia. Methods and Material: The methodology used for this research was a hybrid six-step Cios Knowledge Discovery Process. The required data was collected from Butajira rural health program database covering the period 1987-2008. The researcher used two popular data mining algorithms (C4.5 J48 Decision Trees and Naïve Bayes Classifier) to develop the predictive model using a larger dataset (11,600 cases). The researcher also used a 10-fold cross validation and 90% split test mode for data mining methods of the two predictive models for performance comparison purposes. Results: The results indicated that the decision tree (J48 algorithm) is the best predictor with pruned parameter of the tree of 90% split test mode; it has 97.49% accuracy on the holdout dataset (this predictive accuracy is better than any reported in the literature), Naïve Bayes Classifier came out to be the second with supervised discretization has 96.67% accuracy. Conclusion: The results from this study were very capable and confirmed the belief that applying data mining techniques could indeed support a predictive model building task that predicts the pattern of under-five mortality in Ethiopia; particularly for Butajira rural health program sites are possible. In the future, more classification studies by using a possible large amount of Butajira rural health program demographic and surveillance sites dataset records with epidemiological information and employing other classification algorithms, tools and techniques could yield better results.

Item Type:	Thesis (Masters)
Subjects:	H Social Sciences > HA Statistics R Medicine > RJ Pediatrics R Medicine > RJ Pediatrics > RJ101 Child Health. Child health services T Technology > T Technology (General)
Divisions:	Africana
Depositing User:	Selom Ghislain
Date Deposited:	11 Sep 2018 07:35
Last Modified:	11 Sep 2018 07:35
URI:	http://thesisbank.jhia.ac.ke/id/eprint/5113

Actions (login required)

View Item