The dance of classes - A stochastic model for software structure evolution

In this study, we investigate software structure evolution and growth. We represent software structure by means of a generic macro-topology called Little House, which models the dependencies among classes of object-oriented software systems. We, then, define a stochastic model to predict the way software architectures evolve. The model estimates how the classes of object-oriented programs get connected one to another along the evolution of the systems. To define the model, we analyzed data from 81 versions of six Java based projects. We analyzed each pair of sequential versions, for each project, in order to depict a pattern of software structure evolution based on Little House. To evaluate the model, we performed two experiments: one with the data used to derive the model, and another with data of 35 releases, in total, of four open-source Java project. In both experiments, we found a very low rate of error for the application of the proposed model. The evaluation of the model suggests it is able to predict how a software structure will evolve.