Research And Application Of A Combination Algorithm Model In Software Cost Estimation

With the rapid development of computer software industry, recent years the size of software system has become larger and larger, and the unsuccessful phenomenon, for example software project delay, cost overruns and loss, have also become more and more serious. As the urgent need of software development cost estimation, many organizations have studied and published a variety of estimation methods and estimation models, making this technology get mature day by day. However, these existing software development cost estimation technologies have many imperfections.For existing software cost estimation accuracy is low, estimated inflexible, can not effectively assist project management. This dissertation proposes a method which combines algorithm models and AHP group decision. As the algorithm model of clarity of purpose, objectivity, repeatability and high efficiency, judging from the development trend, algorithm models have more advantages. Because different algorithm models have different strongpoint and weakness and application scope, this approach integrates the advantages of all algorithm models, makes up their own shortcomings, and improves the estimation accuracy and flexibility.Firstly, the dissertation compares the advantages and disadvantages of current major software cost estimation methods (expert estimates, analogy estimates, regression analysis, each algorithm model) analyzes the major problems which software estimation models exist. An overview of algorithm models and AHP group decision-making, it lay a theoretical foundation for model construction.Then based on AHP group decision making, the dissertation establishes a program layer hierarchy ladder, integrated estimation as the target layer, software cost factors of production factors, computer factors, human factors, project factors and other factors as the criterion layer, algorithm models classification which include composite model, table model, product model, linear model as sub-criteria layer, algorithm models as project set. Use of expert pair wise comparison matrix gain relative weight, calculate weights of every level factors relative to the target layer from high-level to low-level successively, and finally get relative weights of the program level's algorithm models which is relative to the target level, then get estimated cost by weight synthesis.Finally, based on theoretical models, this dissertation makes a simple calculation example. In the instance, the absolute error and relative error between the actual cost data and the software cost estimation data which is obtained by AHP group decision-making weights method is least compared with other algorithm models. It verifies that the model greatly improves the estimation accuracy.