Improved Time Performance of Adaptive Random Partition Software Testing by Applying Clustering Algorithms

Abstract

Software testing is generally accepted technique for evaluating and taming software quality. Random testing (RT) is a major software testing strategy and a basic testing technique randomly generates test cases from the set of all possible program inputs. The simplicity of this testing makes it likely the most efï¬cient testing strategy with respect to the time required for test case selection. Though very simple, RT is still considered as one of the state-of-the-art testing techniques, along with other more complicated and systematic testing methods. Its efficacy is notified to be less while considering its capacity of defect detection. This has been proven to pertinently overcome by Adaptive Testing (AT), on the other hand the methodology of AT is comprised of intricate complexity and high computational cost as its main constituents. Adaptive random testing (ART) is one major approach for enhancing RT. Another category of testing techniques is partition testing, which involves dividing the input domain up into a fixed number of disjoint partitions, and choosing test cases from within each partition. Partition testing has powerful, intuitive appeal, and analytical results show that even simple partitioning schemes may be more effective in fault detection than random testing. The existing hybrid approach is a combination of AT and RPT which is called as ARPT strategy, which enhances the AT. The objective of this proposed research is to improve random partition in ARPT strategies by utilizing clustering algorithms like Expectation Maximization (EM) algorithm and Nonnegative Matrix Factorization (NMF) clustering algorithms and the Self-organizing map (SOM) which can be efficiently utilized for partition. In this way random partitioning is improved to reduce the time conception and complexity in ARPT testing strategies.