Modeling development effort in object-oriented systems using designproperties

摘要

In the context of software cost estimation, system size is widelyntaken as a main driver of system development effort. However, othernstructural design properties, such as coupling, cohesion, andncomplexity, have been suggested as additional cost factors. Using effortndata from an object-oriented development project, we empiricallyninvestigate the relationship between class size and the developmentneffort for a class and what additional impact structural properties suchnas class coupling have on effort. The paper proposes a practical,nrepeatable, and accurate analysis procedure to investigate relationshipsnbetween structural properties and development effort. Results indicatenthat fairly accurate predictions of class effort can be made based onnsimple measures of the class interface size alone (mean MREs below 30npercent). Effort predictions at the system level are even more accuratenas, using Bootstrapping, the estimated 95 percent confidence intervalnfor MREs is 3 to 23 percent. But, more sophisticated coupling andncohesion measures do not help to improve these predictions to a degreenthat would be practically significant. However, the use of hybrid modelsncombining Poisson regression and CART regression trees clearly improvesnthe accuracy of the models as compared to using Poisson regression alone