This research presented the concepts of knowledge breakthrough based on the BayesianrnNetworks technology to extract valid models of knowledge. The application domain ofrnthe research is the health sector, one of the potential sectors to apply Bayesian NetworkrnTechnology. The research generally aimed at investigating the potential applicability ofrnBayesian Network technology in developing a model that can support the prediction ofrnART clientsâ€™ adherence tends in Ethiopia. The research was conducted in selectedrnhospitals of SNNP Regional Health Bureau.rnThe methodology used to conduct this research consists of four major steps: Data SourcernIdentification, Data Collection, Data Preprocessing and Model Building/Testing. A totalrnof 1561 records, having 15 attributes each, were used for building, training and testing arnBayesian Network model.rnThe Bayesian Network learning process was done for complete data (the data for whichrnthe training data containing no missing values) which again applied constraint-basedrnapproach. This approach performs conditional independence tests on the data. Then itrnwill search for a network that is consistent with the observed dependencies andrnindependencies (applying d-separation concept). Conditional independence relationshipsrnamong the attributes can serve as constraints to construct a Bayesian Network.rnThe belief network modelling software employed for the purpose of training and testingrnthe BN model was the Belief Network PowerSoft, which applies a constraint-basedrnXIIrnapproach. Three-Phase-Dependency-Analysis (TPDA) in BN PowerConstructor wasrnemployed in developing the model. The BN PowerPredictor, on the other hand, wasrnused to evaluate the prediction accuracy of the model. BN PreProcessor was also therernto preprocess the data so as to make it ready for model building purpose.rnModel testing was implemented in two phases- the first phase without involvement ofrnexpert knowledge (i.e., without node ordering, in which, the algorithm learns bothrnstructure and parameters), and the second phase by eliciting domain expert knowledgern(i.e., involving node ordering, in which, the algorithm learns only the parameters). Forrnboth cases, to ensure consistency across the data during the selection of the test andrntraining sets, experiments were carried out by splitting the data into 10 partitions, i.e., arnpercentage split (10-fold) was used to partition the dataset into training and test data.rnEach partition, in turn, was used for testing while the remainder was used for training.rnThis process was repeated ten times for the learning algorithm and, at the end, everyrninstance was used exactly once for testing. Finally, the average result of the 10-fold crossrnvalidation was considered.rnAccordingly, the average predictive accuracy for the model without an expertrnintervention (Experiment I) was 72.80% at 95% confidence level. According to thisrnmodel, the adherence (to the medication) of an ART client is directly affected by twornfactors: Addiction (drug or alcoholic behaviour) and loss of job due to ill-health.rnNext, TPDA algorithm (Experiment-II) was implemented by allowing elicitation ofrnexpert knowledge. Accordingly, the predictive accuracy of the modified model wasrn75.9% which is a better result.rnXIIIrnSignificant enhancements in prediction and reduction in error rates in the modified modelrnwas taken as the indication of the significance of a domain expertsâ€™ intervention duringrnmodel building.rnAccording to the later model (Experiment II), adherence of an ART client is directlyrnaffected by six factors: Addiction (drug or alcoholic behaviour), loss of job due to illhealth,rnResidence of a patient, Knowledge the client has concerning HIV,rnEmployment status of the client, and Family dependence (independence).rnFrom the model developed, it was observed that Bayesian Network is a powerfulrnpredictor even in the absence of a domain expert. With a proper intervention of domainrnexperts, it was noticed to perform even better.