Author: Muhammad Dhiauddin Mohamed Suffian | Suhaimi Ibrahim
Item Type: Article
Item Date: 2012
View Repository Item

The importance of knowing total number of defects to be found at the start of system testing phase has become significant element in software engineering practices towards achieving zero known post release defects of the software released to users. Defect prediction for testing demonstrates the initiative for test process improvement and optimization while allowing independent testing team to discover as many defects as possible, to determine when to stop testing, put comprehensive test coverage while ensuring all defects are contained within testing phase. The research aims at demonstrating the industry-centric approach of predicting defects and controlling them within system testing to prevent any escapees to the next testing level and later, to the end users. Despite various prior efforts and approaches in building defect prediction models, this research attempts to focus on formulating a mathematical model for specifically predicting functional test defects to be found in system testing phase before it is actually started. Six Sigma is chosen as the approach in building up the mathematical model for predicting functional defects in system testing phase. Design for Six Sigma (DFSS), one of the Six Sigma methodologies is used for the research where it involves 5 main phases: Define, Measure, Analyze, Design and Verify (DMADV). The study illustrates the motivation on having this prediction model and its benefits to test process management. V-Model software development process is the area of focus in which the software projects adopting this model are selected for the research. The activities start with the analysis of factors narrowing to phases prior to testing phase that have direct impact to the discovery of defect in system testing. These factors are concerned with the product and process metrics collected from requirement phase until construction and unit test phase. The metrics in those phases which are significant to system testing defects are collected and analyzed statistically using multiple regression analysis. The proposed prediction model for functional defects in system testing phase is formulated using the best result generated from the regression analysis, which is a combined metrics of development and testing metrics in phases prior to testing phase. The proposed model is then validated to ensure it is fit for actual implementation of predicting functional defects in system testing phase. Control plan is also incorporated as a reference on what action to be taken when the actual functional defects found do not fall within the prediction interval. (Abstract by authors)