Abstract:
Background: - Voluntary Counseling and Testing (VCT) is an important intervention
and entry point in the prevention, control and management of HIV. But the
documentation process of VCT logbooks and dataset has given little attention on the
identification of the VCT user’s and test result patterns in Ethiopia. So, assessment of
patterns in VCT outcomes using classification and association techniques of data
mining is important for proper intervention.
Objective: - To apply data mining in identifying patterns of VCT dataset to discover
knowledge that enables to design proper counseling and prevention strategies.
Method: - The study design was record review, hybrid model of data mining on Gondar
University VCT service. The research is based on 12,033 recorded data. Redundant,
noisy, misclassified, inconsistent, outlier data cleaned and transformed into the
appropriate format used for data mining classification and association rule discovery
algorithms by using WEKA software. Missing, misclassified, noisy values were handled
appropriately during pre-processing step.
Results
The maximum correctly classified percentage obtained is 85.15% with j482
classification algorithm. And, the optimal attributes of the dataset that has been shown
in the experiments to classify HIV test result are physical status of the clients, reasons
to test, marital status, history of STIs, age and sex. And as of the domain experts and
the researcher, there are some unraveling rules from the classifier. In addition, Apriori
association technique experimented with selected variables and positive and negative test results shown various interesting patterns.
