Towards Understanding the Behavior of Classes Using Probabilistic Models of Program Inputs

We propose an approach to characterize the behavior of classes using dynamic coupling distributions. To this end, we propose a general framework for modeling execution possibilities of a program by defining a probabilistic model over the inputs that drive the program. Because specifying inputs determines a particular execution, this model defines implicitly a probability distribution over the set of executions, and also over the coupling values calculated from them. Our approach is illustrated through two case studies representing two categories of programs. In the first case, the number of inputs is fixed (batch and command line programs) whereas, in the second case, the number of inputs is variable (interactive programs).