Research on Failure Model of Software Reliability Based on the Least Squares Support Vector Regression Machines and Simulated Annealing Algorithm

摘要

Building a software failure model (SFM) is an important means to compute the software reliability. While there are not enough faults data, the estimation for model parameters is not accurate, which would decrease the fitting precision. In addition, the estimation for parameters of complex models is very complicated. An SFM based on the least squares support vector regression machines (LSSVRM) could decrease the fitting error, and there are no parameters to estimate in that SFM. The LSSVRM optimized by simulated annealing algorithm (SA-LSSVRM) is introduced to overcome the problem of there being free parameters in LSSVRM to be modified manually. The SFM based on SA-LSSVGM could fit the faults data better, and the precision of software reliability could be improved.