Research Of Software Reliability Model And Cost Model

With the rapid development of computer technology and the quick expansion ofcomputer applications, the scale of the software system has increased and the functionshave become greatly complex. Therefore, the requirement of software quality is alsoincreasing constantly. Software reliability, which is the most important factor of qualitymetric, has become a new development field of the computer technology. Thus, inrecent years, a large number of software reliability growth models have been proposedto analyze the reliability of software system during the testing phase. This thesisproposes a new software reliability growth model based on classic G-O model thenimproves the general cost model and presents a more realistic software cost model.Finally, on the foundation of the improved reliability model and cost model, thesoftware optimal release time can be solved according to minimizing the total testingcost.Main research topics of the thesis are as follows：①This thesis proposes a new software reliability growth model according tointroducing variable fault detection rate and the fault removal rate to the G-O modelbased on NHPP. The G-O model supposes that the fault detection rate is a constantvalue and the fault removal rate is100%. This thesis proposes the fault detection rate isa time-varying function, which considers the number of residual faults and the learningprocess of testers. And it is impossible to remove all the faults during software testing,so this thesis proposes the fault removal rate is a constant, which does not changedramatically with the testing time. The improved software reliability growth modelconsidering variable fault detection rate and the fault removal rate conjointly is moreaccuracy and realistic.②Then this thesis analyses the relationship between the software cost model andreliability model, and presents that software reliability model is the basis of cost model.Under the condition of improved software reliability growth model the thesis proposes anew cost model considering fault removal rate and risk cost of software failureconjointly according to improving the general software cost model. The new costmodel is more realistic.③The general cost model and improved cost model are established respectivelyunder the foundation of classic G-O model and the improved reliability model. Then the software optimal release time is solved under different models according to minimizingthe total testing cost.④Finally the numeric experiments are performed using the actual softwarefailure data set. For different failure data set, the thesis uses least square method andgenetic algorithm respectively to estimate the parameters in the software reliabilitygrowth model and verifies the improved software reliability growth model is moreaccurate for both the data fitting and predicting. Then the minimum testing cost and thesoftware optimal release time is solved for the general software cost model andimproved cost model. And the improved cost model is proved by actual experiment datacomparison that performance of the proposed cost model is superior to other models.Finally the thesis achieves the relationship between software optimal release time andthe parameters in improved cost model according experimental analysis.The research can satisfy the realistic requirement of research of software reliabilityand testing cost. The improved software reliability model and cost model in this thesisare accurate and universal. And they have important significance for related academicand application research.