Entropy-Based Measures of Heterogeneity of Software Structural Profile

Following the growing research interests in complex networks, in recent years many researchers treated software static structures as complex networks and revealed that most of these networks follow a scale-free degree distribution. Different from the perspectives adopted in these works, our previous work found that the networks of software dynamic execution processes may also be scale-free. Scale-free degree distribution demonstrates that during the execution process the methods being invoked only a few times are far more abundant than those being frequently invoked, i.e. the software structural profile is heterogeneous. Software structural profile describes the probabilities of software modules or states being invoked. Since many unique properties of complex networks are due to the heterogeneity, the quantitative measure of it is important and desirable. This paper proposes two quantitative measures of heterogeneity of software structural profile based on entropy. Three case studies are presented to show the effectiveness of the proposed measures.