The impact of design properties on development cost in object-oriented systems

In the context of software cost estimation, system size is widely taken as a main driver of system development effort. But other structural design pro perties, such as coupling, cohesion, complexity have been suggested as addi tional cost factors. In this paper, using effort data from an object-orient ed development project, we empirically investigate the relationship between class size and the development effort for a class, and what additional imp act structural properties such as class coupling have on effort. We use Poisson regression and regression trees to build cost prediction mod els from size and design measures, and use these models to predict system d evelopment effort. We also investigate a recently suggested technique to co mbine regression trees with regression analysis, which aims at building mor e accurate models. Results indicate that fairly accurate predictions of class effort can be ma de based on simple measures of the class interface size alone (mean MREs be low 30%). Effort predictions at the system level are even more accurate as, using Bootstrapping the estimated 95% confidence interval for MREs is 3%-2 3%. But more sophisticated coupling and cohesion measures do not help to im prove these predictions to a degree that would be practically significant. However, the use or hybrid models, combining Poisson regression and CART re gression trees clearly improves the accuracy of the models, as compared to using Poisson regression alone.