Abstract:
Pneumonia is the single leading cause of mortality in under five year children and is a major
cause of child mortality in every region of the world, with most deaths occurring in sub Saharan
Africa and South Asia. It is also known to be one of the predominant causes of mortality for
under-five children in Ethiopia and luck of sufficient pediatricians. Since, conducting this study
is very important to minimize death rate. The main objective of this study is to develop a case based system for the diagnosing and treatment of pneumonia under five-year children. The study
employed a design science research approach to understand the problems in the area and develop
model. The researcher used manual and automated knowledge acquisition techniques, such as
interview, document analysis and data mining. For this study, predictive data mining task mainly
classification technique was performed to generate representative cases from the prepared data.
The required data were acquired from Jimma University Specialized Hospital. WEKA data
mining tool is used for experimentation. Three experiments were conducted by using J48, PART,
and Naïve Bayes classification algorithms to identify the best model and select the best
performing data mining classification algorithm. Based on experimental result, PART
classification algorithm is selected to construct cases for the case based system because it
registered better performance than other classifiers. The developed model was tested with test
instances and only those instances registers more than 99% accuracy were used to develop a
knowledge base for the CBR development for a better efficiency. Then, implement the prototype
by using jCOLIBRI version 1.1. Finally, testing of the developed prototype CBR system is done
to evaluate the performance of the system. The prototype is evaluated using system testing and
user acceptance testing. System testing performed in terms of recall, precision and F-measure
registered 96%, 89% and 92.36%, respectively. User acceptance testing also performed by
involving domain experts and an average of 94% acceptance was achieved. This shows the
system has registered a promising result. However, case-based reasoning system needs to be
supported by rule-based reasoning for providing a complete advice for the problem, increasing
number of cases and including other significant attributes improve the performance of the
developed system which is forwarded as future work.
