Revenues are vital sources of public infrastructure. The presence of collective consumption of properties and facilities necessitates putting some of our income into government hands. Nevertheless, gathering of tax is the main source of income for the government; it is challenging difficulties with fraud. Fraud encompasses one or more persons who deliberately act in secret to deprive the government income and use for their own benefit. Deception can take an unlimited variety of different forms. Fraudulent claims account for a significant portion of all dues received by auditors, and cost billions of birrs annually. In this study, experiments were performed by succeeding the six step Cios et al. (2000) KDD process model. It starts from a business understanding in the AMRA tax policy system and fraud, specifically taking place audit data set. By compelling the data from database of AMRA and understanding of the data with the help of domain expertise and literature. In data preprocessing; missing values, inconsistencies, outliers and related issue handled properly. Afterward that, construction of models and analysis of the result done to facilitate decision making in the business risk analysis. To gather the data, the researcher used interview and observation for primary data and database analysis for secondary data. To conduct this research, we used a total of 12000 records for training the classifier model. Experiment taking place on different classification algorithms such as Naïve Bayes, random forest and J48 algorithms were done. the result of the various models have compared to find the best model using percentage split (66/34%) and 10-fold cross validation evaluation methods. The study discovers that J48 classification algorithm achieves with superior accuracy than other testing mechanisms. J48 recorded an accuracy of 99% were 11880 instances are correctly classified out of 12000 test cases. As a result, further research directions are suggested in order to come up with a workable system in the subject field..