Journal of Networking Technology Bahman Arasteh 9 3 2018 https://doi.org/10.6025/jnt/2018/9/3/94-101 http://www.dline.info/jnt/fulltext/v9n3/jntv9n3_3.pdf Nowadays, the role of software has becoming increasingly important in many safety-critical applications and the reliability is a key issue in the software systems. One of the ways for improving software Reliability is predicting its faults before tasting phase. Ability of predicting fault–proneness software modules can reduce software testing cost and consequently overall software project cost. In this paper, a combined method includes Neural Network and Naive Bayes algorithm are used to build a software fault prediction-model. Five traditional fault-datasets are used to construct and evaluate the prediction model using proposed method. The results of experiments indicate that the constructed model by the proposed method have higher prediction accuracy and precision than the other methods.