Background: The likelihood of the occurrence of incontinence after successful surgicalrnrepair makes predicting urinary fistula surgical repair outcome important for decisionrnmaking during operation and for further follow up and treatment. rnrnObjective: The purpose of this thesis is to apply data mining techniques to build a modelrnthat can assist in predicting surgical outcome of urinary fistula repair based on clinicalrnassessments done just before surgical repair. rnrnMethodology: The six-step hybrid knowledge discovery process model is used as arnframework for the overall activities in the study. 15961 instances that have undergonernurinary fistula repair in Addis Ababa Fistula Hospital are used for both predictivernassociation rule extraction and predictive model building. Apriori algorithm is used tornextract association rules while classification algorithms J48, PART, NaÃ¯ve Bayes andrnmultinomial logistic regression are used to build predictive models. Support andrnconfidence are used as interestingness measure for association rules while area under thernWROC and ROC curve for each specific outcome is sequentially used to comparernperformances of models from the predictive algorithms. rnrnResults: Predictive association rules from Apriori have shown frequent co-occurrence ofrnless severity of injury with cured outcome. The predictive model from PART-M2-C0.05Q1rnschemernhas shown an area under WROCrncurve of 0.742. Area under the ROC curvernforrnresidual outcomern(ROCrn=0.822) from this algorithm is better than NaÃ¯ve Bayesrnand logistic, while the areas under the ROC curves for the other outcomes are greaterrnthan the model from J48.rnrnConclusion: Predictive model is developed with the use of PART-M2-C0.05-Q1. It is rnResidualrnbetter in detecting residual outcome than the logistic regression model. The predictive rnassociation rules and predictive model built with the use of data mining techniques canrnassist in predicting urinary fistula surgical repair outcome.