Estimation Method And Application Research Of E-government System R&D Cost Interval

The interval estimation of software system development costs are complex.At present,China has undergone a one-year trial run on the pre-budget approval of the e-government system.Due to the lack of consistent review rules by multiple expert groups,or the complexity of the software product development process and the ability of users to feedback,generated a larger review gap.In order to eliminate the evaluation inconsistency factor,the main work of this paper is as follows:this paper completes the rapid cost interval estimation based on the preliminary design and the technical characteristies of the e-government system,and uses the machine learning methods to further correct the evaluation result interval with the CAC's 2017 review results.On the one hand,this paper designs a table for collecting project characteristics.Firstly,the preliminary design of the declaration party fills in and extracts the required software feature data,and then obtains the project feature value through the machine automatic extraction method.Base on the theory of the software architecture,through the quantitative calculation of system complexity and component reusability factors,and according to the cost interval estimation principle,constructs a cost structure cost interval calculation model based on J2EE architecture development method,that is,the COCOMO model formula is used to calculate the research and development cost.The interval solves the problem that the traditional COCOMO model can only obtain a single value.Finally,the system is speculated and verified,which ensures the validity of the R&D cost interval estimation.On the other hand,through the data extraction and analysis calculation of the declaration document,the obtained results are roughly consistent with the actual disbursement of funds,but there are still some cases where the error of some items is large.Therefore,the characteristics extracted from the project declaration and the actual disbursement amount are fitted as BP network training data,and the final data convergence effect is better,which can be used as reference data for expert approval.Trained data can be accumulated as experience points to enrich the project library of the e-government software system.