Improving effectivenss and efficiency of regression testing techniques using nature inspired algorithms

Abstract

Generally regression testing is used to validate the modified versions of software newlinein maintenance phase. Regression testing incurs significant cost and also considered as newlinefrequently executed process. Cost of regression testing increases when changes are newlinemade in later stages because such changes also impact the existing functionality of the newlinesoftware. newlineThe major difference in development and regression testing is that during newlinedevelopment testing, we have well planned schedule or we can say that proper time is newlineallocated to development testing during project planning. While during regression newlinetesting we do not have enough time available for it. So, we have to rely on the well newlineestablished test suite (established during the development testing) and its execution newlinehistory. This strategy is very basic and known as retest-all-test-strategy. Re-Test-alltest-strategy may consume significant time and resources which definitely increase the newlinecost of regression testing as well as the time needed for the regression testing. Other newlinetechniques used to reduce the cost and times required for regression testing are: test case newlineselection, test case minimization and test case prioritization. These techniques reduces newlinethe overall time as well as the resources required for regression testing by selecting, newlineminimizing, prioritizing some subset of pre existing test suite. newlineThe proposed work addresses the issues related to existing regression testing newlinetechniques and suggest a novel process to generate the modification traversing test cases newlinefor regression testing using nature inspired algorithms. As a test case generation newlineproblem is considered as NP-hard problem so nature inspired algorithms produce newlineimproved solutions in terms of efficiency and efficacy. Here, two different nature newlineinspired algorithms have been proposed (i.e. EAM and MRDE) and their performance newlinehave been tested on different benchmark functions and the proposed algorithms newlineoutperforms the existing algorithms. Further these algorithms are applied on the test newlinecase generation problem in regress