Research On Cost Estimation Model Of Software Project Based On

The scale of software project gets more and more large, the development and applicationenvironment gets more complex, especially at the beginning of starting a project, the projectinformation is not comprehensive, the demand is unclear, its all cause inaccurate estimation of theproject’s cost, ultimately cause project overruns it’s cost, and because of the improper adjustmentof the cost monitoring, leading to more and more fail cases that fail to deliver on time or even cannot put in use. Therefore, in order to improve the accuracy estimates of software project’s upfrontcosts, study the method of cost estimation of a software project at it’s initial stage, and build arelatively perfect cost estimation system, to make sure the software project meets customer’srequirements in the expected cost and time, has a significant theoretical and practical value.The main research contents are as follows:First of all, on the base of analyzing the software project characteristics and the existingsoftware cost estimation research, pay more attention on studying project’s initial stage’s costestimate method, compare the analysis of analogy estimation, expert judgment method, theparameter model method, genetic algorithm and the advantages and disadvantages of the othercommonly used estimation model and the applicable conditions, combine with the characteristics ofsoftware project and the principle of CBR model, and analysis of the feasibility and theeffectiveness of constructing the software project cost estimation model of Case-BasedReasoning (CBRCEM).Secondly, in order to fully describe the attributes of a software project cost, the article establishthe software project case attributes based on software project characteristics, project technologyfactors and the factors of the developers; combine with the cost driver theory based on COCOMOmodel; establish two level index system of the cost attributes of software, index calculation formulaand the agreed range estimation, and design the structure of the software project cost case library.The software project has been delivered on schedule, finishing the same type or application ofrepresentative cases, classify and integrate the key factors affecting software cost estimation, thenestablish the source case library.Third, in order to search the completed software project information which case base withcharacteristics similar to the target case, normalize the utility function, use the AHP and theimproved optimal gradient descent method to determine the weight of the key attribute of the software cost; Because of the software project case is relatively small, grey correlation is not limitedby the number of samples and sample rules, establish the improved grey relational analysis softwareproject case matching algorithm based on similarity theory, similarity calculation target case andsource case, retrieval of similar cases.Fourth, in order to improve the software project cost estimation accuracy, according to the casesimilarity ranking selection of multiple software projects similar to the target case, the softwareproject cost estimation value for solving multiple attribute decision making problems withincomplete information on attribute weights, the article optimize the software project cost attributeweights; combine with the feature weight of software project, apply particle swarm optimizationalgorithm, code the weight of cost estimate into a particle of PSO, select the cost estimation errorfunction as fitness function, the iterative updating particle position and velocity information, searchfor optimal weight, get more close to the actual target software project cost.In the end, the article applies the CBRCEM to the cost estimation of the marketingmanagement information system of China Southern Power Grid. Through the comparative analysisof various algorithms, we know the CBRCEM calculation software cost estimate and the estimationerror is small relative to the actual cost value, The evaluation results are more close to the actualcost of the project, which can effectively avoid the influence of experience of experts andprofessional differences due to limitations of cost estimation accuracy, therefore, CBRCEM has acertain practical value.