Studying Aging-Related Bug Prediction Using Cross-Project Models

Software ageing process as an evolving dynamic system

“…They treated ageing as non‐fatal performance degradation or as a fatal error condition. Many other model‐based studies also treated software ageing under various runtime scenarios as a stochastic phenomenon [1, 6, 11, 12 ]. To extend the runtime life of a software system, Machida et al [1 ] proposed a semi‐Markov model that can be used to derive the optimal trigger for starting life‐extension.…”

“…However, in the research area of software ageing, experimental studies are very limited. This is in contrast to the widely accepted importance of experiment‐based studies [6, 7 ], and has led to the practices where many of the ageing models proposed in the literature are based on some arbitrary assumptions about the root origins and distributions of software ageing. Consequently, most studies treat software ageing as a stochastic phenomenon, and discretise the ageing process in a series of phases or epochs, focusing on the optimal rejuvenation time based on some arbitrary assumptions [6, 8–13 ].…”

“…Although researchers have proposed many arguments for its causes and evolution [5–7 ], the root cause of software ageing is still not well understood. This fundamental question can only be answered through thorough experimental research.…”

Optimized Differential Evolution Algorithm for Software Testing

“…In the process of evolution of test cases, the newly generated test cases will fall in the vicinity of the points that have been inevitably searched in the solution space, which leads to unevenness of that distribution. If the test cases corresponding to the branches that have not yet been covered are distributed in the space that is far apart from the space of the individual convergence, especially in the boundary space of the solution space, new effective test cases may not be found even through repeated evolutions [23].…”

“…According to Refs. [24][25][26], the Siemens suite is most frequently used as standard experimental test programs, and the size of each program in the suite is small with less than 600 lines of code. Therefore, in this experiment, we selected the classic Siemens program set for the testing programs and downloaded this suite from Ref.…”