Evaluation strategy for software reliability based on ANFIS

With improving of the complexity level of the large software system, the number of code line shows geometrical multiples growth, so software reliability become more and more difficult to be guaranteed. The traditional test security measure must spend a lot of manpower and time to complete. And the system itself needs more reliability to run. In some key fields, such as aviation, spaceflight, its reliability requirements is especially important. The reliability of the software in the system is directly related to their own reputation and survival development competition ability. This paper presents a method to evaluate the software reliability using Fuzzy-Neural network. In order to improve the accuracy of the evaluation, this paper established a reliability prediction model based on adaptive-network based fuzzy inference system (ANFIS). The model use the reliability data(defect counts of every thousand code lines) of one software project as input data, and use the prediction of reliability as output data, training Adaptive-Fuzzy neural network, get the membership function of defect counts of every thousand code lines.