Optimizing Predictive Mining Techniques in HIV-Related Opportunistic Infections: Case for Botswana


Botswana was one of the first countries to establish a national Antiretroviral therapy (ART) programme in Africa. “Masa,” a native word depicting “a new dawn,” is the programme name. The AIDS epidemic has led to the emergence of several disease entities which in the pre-AIDS era seemed innoxious. Some HIV AIDS positive patients under the ART programme continue to be at risk of contracting related Opportunistic Infections (OIs) and little evidence based research work has been carried out so as to apply preventative or mitigating factors. The use of Data Mining (DM) techniques is becoming more popular for investigating subtle relationships in Clinical data. This paper proposes to build a robust classification and prediction model by mining the historical data stored in a data-warehouse to determine which patients might be at risk of contracting the afore-mentioned infections. Four supervised learning algorithms viz Generalized Linear Model (GLM), Support Vector Machine (SVM), Decision Tree (DT) and Naïve Bayes (NB) were used for building the models. Their performances were analyzed and evaluated for their efficacy against the Confusion Matrix analytical performance, the Receiver Operating Characteristic (ROC) curve, the LIFT Cumulative and Profit analysis. Experimental results proved that the SVM exhibited superior performance and was therefore deployed in building the HIV related OI prediction model.

Extracted Key Phrases

7 Figures and Tables

Cite this paper

@inproceedings{Li2013OptimizingPM, title={Optimizing Predictive Mining Techniques in HIV-Related Opportunistic Infections: Case for Botswana}, author={Zhiyong Li and Clement K. Kirui}, year={2013} }