Research On Software Cost Estimation Method Based On Requirement Identification

With the continuous development and popularization of information technology,various software has become an indispensable part of modern society and involves an increasingly wide range of applications.However,the cost and complexity of software development is also increasing,and it is becoming more and more important to conduct scientific and reasonable software cost estimation.Software cost estimation is very professional and complicated,and the traditional software cost estimation mainly relies on expert experience,which puts high demands on the business ability of the reviewer and has problems such as large subjective errors and low efficiency.At the same time,software development involves many factors,such as development time,staffing,technology selection,etc.These factors are interrelated and affect each other,making it difficult to conduct accurate quantitative analysis and to guarantee the accuracy of cost estimation.In particular,the task of conducting software cost estimation for government information system projects requires consideration of multiple factors such as data privacy,policies and regulations,and specific operation and maintenance requirements,and also has higher requirements in terms of estimation time and accuracy.Therefore,this thesis uses natural language processing techniques to optimize the software cost estimation process,aiming to assist experts in improving the efficiency and accuracy of software development project cost estimation.To address the issues related to software cost estimation,this study optimizes the whole estimation process.First,a requirements identification model based on graph attention network is used for software requirements identification,and functional and non-functional requirements are identified from the complicated software requirements documents.Second,based on the results of requirements identification,software size estimation is performed for functional requirements,and functional points in functional requirements are identified using event extraction techniques based on the function point analysis method.Classification statistics are performed for non-functional requirements to assist the determination of scale degree factors and cost driving factors in the subsequent cost estimation steps.Finally,the obtained function point data are input into the optimized COCOMO II model to estimate the software workload and software cost.By using natural language processing technology,this study reduces the dependence on manual work for software requirement identification and function point extraction,and optimizes the cost driving factors in the COCOMO Ⅱ model,which has important research significance and practical application value for reducing software estimation cost and improving estimation efficiency and accuracy.To verify the effectiveness of the software cost estimation process proposed in this study,a series of experiments,including software requirement identification experiments,software size estimation experiments and software workload and cost estimation validation experiments,were conducted based on the software cost estimation data of the government information system.The experimental results show that the software requirement identification model based on graph attention network can effectively distinguish functional requirements from non-functional requirements,and also performs well in the task of classifying non-functional requirement subcategories.The structured extraction of functional points using event extraction method can predict software scale well and makes the evaluation metrics of software scale metrics achieve good results.The improved COCOMO II model is used to estimate the workload and software cost of government information system projects,and the error of workload estimation results are verified to be less than 15% by actual projects.From the comprehensive experimental results,the method proposed in this study can effectively improve the accuracy and efficiency of software cost estimation,and is expected to be promoted and applied in the actual software project management.