Browsing by Author "Mohamed Suffian, Muhammad Dhiauddin"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Software defect prediction model for system testing using prior phases metrics
    (Universiti Teknologi Malaysia, 2020) Mohamed Suffian, Muhammad Dhiauddin
    The ability to predict defects before test execution start serves as an important element in ensuring that the defects could be contained within the system testing phase thus preventing them from escaping to the end-users. There have been many prior attempts in terms of approaches and techniques to predict defects in software via a model but there is limited knowledge on how the factors in phases prior to system testing could be used to predict defects in that phase. This research focuses on the development of prediction model for defects in system testing using metrics used in software development phases. The main aim is to establish a prediction model of defects for system testing phase using metrics used in phases prior to testing specifically for Vmodel development. Product and process metrics from actual software development projects using V-model were collected. The development and testing related activities metrics were then decomposed further into size, defect and effort-related metrics. Multiple regression analysis was utilized to analyse and determine the most significant metrics that demonstrate significant impact and serve as predictors to defect discovery in system testing. Prediction equation candidates were selected from the mathematical equations derived from the analysis with P-value of less than 0.05 for each metric used with R-squared and R-squared (adjusted) values of more than 90% respectively. The prediction equation candidates were then verified using new software projects. The equations that predicted defects within 95% prediction interval or prediction range of actual defects found was chosen as the active prediction equation as part of the prediction model. The selected prediction equation has successfully proven that the software defect prediction model is able to predict system testing defects using metrics in prior phases.