A Hybrid Approach for Software Fault Prediction Using Artificial Neural Network and Simplified Swarm Optimization

Abstract: The major task in the software developing is to provide a software which is free from any kind of defects. But this task is hard to accomplish by the developers. Fault prediction can be classified as one main region to forecast the possibility of the software containing faults. The aim of the fault prediction in software development life cycle is to categorize the software modules in fault-prone and non fault-prone modules as soon as possible. This classification of fault-proneness of a module is actually essential for reducing the cost and increasing the efficiency of the software development process. In this paper, we propose a hybrid model using artificial neural network (ANN) and Simplified Swarm Optimization (SSO) for fault prediction. ANN is used for categorization the software modules in fault-prone and non-fault-prone modules, and SSO is then used to reduce dimensionality of dataset. This approach is easy to implement as no expert knowledge is required. The attained results confirms a preferred performance of this approach for fault prediction and output rate or recognition. The results indicates the prediction rates of proposed method is more than 90 percent in best condition.

Keywords: Software fault proneness, fault prediction, artificial neural network, simplified swarm optimization.