Research And Application Of Software Price Surveys Method Based On Linear Regression Analysis Technology

Since software development is much influenced by user's requirement and development method, it would cause software project spending out of control and then lead to software crisis if without clear estimation of software cost. Thus how to make a good estimation of software cost is one of the researchable problems that is urgent to be solved in software industry.In this paper collected"authorized"software pricing data from the domestic commonly used software pricing method, Cost Plus Method, analyzed distribution and linear rule in software pricing and related cost; based on linear regression analysis method, established a software price model. We carried out in practice that the model may very well work out the cost of software project, effectively lower the difficulty and workload of estimation, and further be applied in software pricing.This paper has mainly done these works:First, Pricing process forward software costs. Deeply analyzes the software cost comparison authoritative domestic cost-plus law constitutes a method in the paper, the software research and development process, producing eight charges set to basic response factor, the method based on linear regression analysis, a suitable for domestic industry use software costs estimates process. Second, establishing a software price model. Data sets used by modeling werederived from 11 authorized software project quotation program. In modeling, selected attribute factor with step regression analysis method, followed by validation, optimization and predicted performance analysis of the model, then derived the final model.Third, processing incomplete data. As incomplete data existing in original data sets, paper discussed the cause and principle, then proposed five methods to fill in the missing value, in which filling with regression sample mean was firstly proposed.Fourth, optimizing the model. Paper separately established price models associated with those five methods, and then derived the optimum model by comprehensive screening. The optimum price model is the optimum tri-variable model about salary cost, management charges and design cost, which interprets 97.77% changes of software cost.In the future, the model are supposed to be optimized in: collecting more software pricing projects for regression analysis to improve its prediction accuracy, detailing software cost to find and bring in more variables about software cost, trying to model with other regression analysis methods, etc.