Software Reliability Prediction by Using Ant Colony Optimization Technique

Software reliability prediction is very challenging in the starting phases of software development. In the past few years many software reliability models have been proposed for assessing reliability of software but building accurate prediction models is hard due to the recurrent changes in data in the domain of software engineering. As a result, the prediction models built on one dataset show a significant decrease in their accuracy when they are used with new data. The objective of this paper is to introduce a new approach that optimizes the accuracy of software reliability predictive models when used with raw data. We propose Ant Colony Optimization Technique (ACOT) to predict software reliability based on data collected from literature. An ant colony system with an accompanying TSP algorithm has been used, which has been changed by implementing different algorithms and extra functionality, in an attempt to achieve better software reliability results with new data. The intellectual behavior of the ant colony framework by means of a colony of cooperating artificial ants are resulting in very promising results. The method is validated with real dataset using Normalized Root Mean Square Error (NRMSE).