某股份制银行软件成本估算模型研究和软件构建

发布时间：2018-02-08 19:22

本文关键词： 软件成本估算回归分析回归技术逐步回归　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：自二十世纪六十年代开始,以软件成本估计不准确、质量得不到保证和项目延期等为典型特征的软件危机频繁发生,因此,学术界和工业界都开始认识到软件成本估算的重要性。我国于2013年发布和实施了行业标准《软件研发成本度量规范》某股份制银行为响应国家相关部门的号召,要依据该标准研发一套软件研发成本估算系统。本文旨在完成该系统的研发,以协助企业方便地明确软件研发成本,其中对软件研发成本估算方法的研究成为本系统研发过程中的关键所在。近年来,越来越多的软件成本估算方法被相继提出,并且处于不断地完善中,主要包括统计分析法、模型法、专家法、基于机器学习的估算方法等。本文首先从公司实际出发,通过对各种估算技术的比较分析确定采用统计分析中的回归技术;然后详细介绍了回归技术的理论基础及方法原理,分析了回归分析中的一元回归、加权回归、多元回归、逐步回归、岭回归等核心方法,并仔细分析了以上方法在建模过程中的适用范围和注意事项;然后介绍了样本数据来源,研究并确立了软件研发成本影响因子;接下来进入建模过程:数据集的验证及预处理、采用不同的回归方法建立模型、模型校准、模型检验、模型性能分析比对;最后把最优模型应用到软件研发成本估算系统的研发过程中,并通过实际项目数据验证模型和软件研发成本估算系统的正确性。系统开发完成后,极大地方便了在新项目实施过程中确定软件成本预算和进度方面的一些参数,使软件的开发处于有效的监督之下,从而能够对软件的质量及进度等进行合理控制,还有助于企业做出合理的投资、招投标等商业决策等。系统目前在使用过程中运行状况良好,能够满足不同人员的使用需求,并且能够进行快速迭代方便地完成系统的更新和优化。
[Abstract]:Since 1960s, software crises, characterized by inaccurate software cost estimates, lack of assurance of quality and project extensions, have occurred frequently. Both academia and industry are beginning to realize the importance of software cost estimation. In 2013, China issued and implemented the industry standard "Software R & D cost Metrics", a joint-stock bank in response to the call of the relevant departments of the country. A software R & D cost estimation system should be developed according to this standard. The purpose of this paper is to complete the research and development of the system, in order to help enterprises to determine the software R & D cost conveniently. In recent years, more and more software cost estimation methods have been put forward one after another, and are in the process of continuous improvement, including statistical analysis. Model method, expert method, estimation method based on machine learning and so on. Firstly, the regression technology of statistical analysis is adopted by comparing and analyzing the various estimation techniques from the company's actual situation. Then the theory foundation and method principle of regression technology are introduced in detail, and the core methods of regression analysis, such as univariate regression, weighted regression, multivariate regression, stepwise regression and ridge regression, are analyzed. Then the source of sample data is introduced, and the influence factor of software development cost is studied and established. Then, the modeling process: validation and preprocessing of data set is introduced. Different regression methods are used to establish models, model calibration, model testing, model performance analysis and comparison. Finally, the optimal model is applied to the R & D process of software R & D cost estimation system. After the completion of the system, it is greatly convenient to determine some parameters of the software cost budget and schedule during the implementation of the new project. Make the software development under effective supervision, so that the software quality and progress of reasonable control, but also help enterprises to make reasonable investment, At present, the system runs well in the process of use, can meet the needs of different personnel, and can carry out quick iteration to complete the system update and optimization conveniently.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.5;O212.1

【参考文献】