Research And Application Of Software Cost Estimation Method Based On Classification And Prediction

With the rapid development of software industry, software products become more and more complicated, and developing costs become more and more difficult to control. Software cost estimation has become an improtant part of the software engineering. Current software cost estimation techniques are mainly model-based estimation techniques, expert determination technology, estimation techniques based on machine learning, regression analysis, using a combination of several methods.This paper studies the software cost estimation method based on classification and prediction , which is one of estimation techniques based on machine learning.Software cost estimation method based on classification and prediction, which basic idea is based on extensive historical project data to build the classification model, and then uses the classification model to predict the target attribute values of a new software project. Classification and prediction for software cost estimation is easy to understand, easy to implement, and its estimated time is short, easy to accept, estimates objectively, without too many project background and so on.Firstly, this paper studies the common software cost estimation techniques, analyzes the principles and characteristics of the various estimation techniques, summarizes some problems in the current software estimation, and studies the basic theory of classification and prediction, focuses on analyze the choice of classification in the software estimation and the significance of decision tree classification for estimating software. Based on these work, a process of software cost estimation which based on classification and prediction is presented.Then, by estimating the specific set of software project, the paper describes the process of software cost estimation based on classification and prediction in detail. The main steps are: 1) Data preprocessing. For historical data sets, we select data, clean data, select feature and classify the target property. The classification of target attributes to adapt the software cost estimation is presented, and the type of the target attribute is mapped to specific predictive values. 2) Construction of decision tree classification model. We compare and analyze the different decision tree algorithm, and build a decision trees classification model of the software cost estimation. 3) The post-processing of the Model. Decision tree is pruned and the model is evaluated and optimized. Then the software cost estimation model is optimal. 4) Estimating software project. By the optimal model, we predict the target attribute value of a new software project.Finally, we analyze two different types of software project data sets using the software cost estimation process based on classification and prediction. We calculate the specific performance and predictive ability of the model.