Action-based in-process software defect prediction software defect prediction techniques based on software development activities

Abstract

An action-based in-process software defect prediction (ABDP) applies classifying records of a performed action to predict whether subsequent actions cause defects in a project. A performed action is previously defined herein as an operation performed based on tasks in Work Breakdown Structure (WBS) of the project. Rather than focusing on the reported defects, ABDP discovers the patterns of the performed action that may cause defects composing a historic data set, and uses analytical results to predict whether the subsequent actions are likely to generate defects composing another historic data set. Once actions with high probability of causing the defects are identified, stakeholders review these actions carefully and take appropriate corrective actions to create a fresh performed action. The fresh performed action is continually appended to the historic data sets of amending the defects to construct a new prediction model for further subsequent actions.