An empirical analysis of data preprocessing for machine learning-based software cost estimation

摘要: ContextDue to the complex nature of software development process, traditional parametric models and statistical methods often appear be inadequate model increasingly complicated relationship between project cost features (or drivers). Machine learning (ML) methods, with several reported successful applications, have gained popularity for estimation in recent years. Data preprocessing has been claimed by many researchers as a fundamental stage ML methods; however, very few works focused on effects data techniques. ObjectiveThis study aims an empirical assessment effectiveness techniques context estimation. MethodIn this work, we first conduct literature survey publications using techniques, followed systematic analyze strengths weaknesses individual well their combinations. ResultsOur results indicate that may significantly influence final prediction. They sometimes might negative impacts prediction performance methods. ConclusionIn order reduce errors improve efficiency, careful selection is necessary according characteristics machine datasets used