Application of Data Mining Techniques to Identify Patterns in Voluntary Counseling and Testing: A Case of Gondar University Hospital VCT Center, Gondar, Northwest Ethiopia

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.