Chaos-Based Estimation Method for Software Reliability

Reliability is one of the essential attributes of the dependable software, and an important factor for quantitatively characterizing software quality. Conventional methodology is Software Reliability Growth Model (SRGM), which specifies the form of a random process that describes the behavior of software failures with respect to time. In this paper, we propose that the behavior of software failures possesses determinacy as well as randomness. So we apply chaos theory to software reliability assessment. Based on chaos theory, we can estimate software reliability according to the objective law hidden in the data, unlike the SRGM which usually makes a number of assumptions. This approach avoids the subjectivity in the SRGM. Also, we analyze the actual software failures data sets with the add-weighted one-rank local-region method (AOLM) based on chaos theory and compare the results with the conventional stochastic SRGM JM (Jelinsky-Moranda) model and NHPP (nonhomogeneous poisson process) mode. Comparison results show that the proposed method fits better than the stochastic ones.