Abstract:
The purpose of this study is to develop a knowledge-based system for blood donation and
transfusion using data mining techniques that support physicians and technicians. To overcome
the problems, the researcher used the KDD process model and data mining algorithms to identify
blood donors. For this study, the dataset was collected from Jimma Zone Blood Bank which has
12,442 instances and 15 attributes. The classification data mining technique was used and selected
algorithms are applied for identifying blood donors. The selected algorithms for the experiment
were NaiveBayes, J48, JRip, REPTree, and PART classification algorithms. For the test option,
the experiment used k-fold cross-validation and percentage of split using the whole and selected
attribute. The accuracy obtained from data mining algorithms using 10-folds cross-validation of
test options by whole attributes were, 99.6%, 99.7%, 99.49%, 99.54%, and 99.08% for Naïve
Bayes, JRip, PART, J48, and REPTree respectively. The developed model was tested by the same
algorithms and the accuracy obtained was 96.7%, 96.78%, 96.22%, 96.5%, and 96.82%
respectively. The algorithm with high accuracy was selected for generating rules. Hence,
knowledge was acquired from these generated rules and represented using rule-based knowledge
representation techniques. A knowledge-based system was developed from acquired knowledge
for donation and transfusion purposes. Finally, user acceptance testing was conducted, and a
developed knowledge-based system was got an acceptance of 87.2% via domain experts. For
future work, the dataset will be collected from the different blood banks and the research will be
applicable as a country at the different health centers and will use a hybrid form of the process
model for better accuracy. Additionally, the algorithms used will be increased and an attractive
user interface will be developed to fulfill users' needs.
