A Grey-based Rough Set Approach For Evaluation Of Software Cost Estimation Methods

With the rapid development of the computer software industry, software cost estimation methods have become an important research direction of software engineering and project management. However, because of the limitation of acknowledgment, complexity of software system and interference of irrelevant information, traditional software cost estimation methods have been faced with the challenge of poor and uncertain inputs, which leads to different cost estimation methods varying greatly on estimation accuracy and effectiveness. Therefore, it is crucial to perform objective and reasonable evaluation of software cost estimation methods. Based on the analysis of the features and complementarities of rough set and grey system, this paper presents a grey-based rough set comprehensive evaluation model in order to evaluate software cost estimation methods.The main work of this paper is as follows:Firstly, the related concepts of rough set theory and grey system theory are described, and the similarities and complementarities of two theories are summarized and analyzed. Then a grey-based rough set comprehensive evaluation model is proposed. The model considers grey evaluation information on the whole, uses interval grey number to describe attribute values, and then builds the original grey decision table. The model makes use of the grey system theory to compute the grey interval correlation coefficient and establish the whiterization weight function. Besides, attribute weights are computed based on attribute importance and grey relational grade, grey-based rough approximation operators are made to divide the decision attributes. Then the results of evaluation are obtained.Secondly, the grey-based rough set approach for evaluation of software cost estimation methods is studied. The approach uses different formulas standardize benefit-type or cost-type attributes, and then computes grey interval correlation coefficient. According to the grey interval correlation coefficient, the method of determining the whiterization weight function threshold, that is mean value method, is proposed, which greatly improves the accuracy and objectivity of evaluation. The approach uses grey clustering to make a clustering analysis of attribute values, and exploits grey-based rough approximation operators to approximately classify decision objects with limited and fuzzy cognition. Taking a development project of Shenyang Blue Torch Corporation as the data source, this paper uses the approach to evaluate four software cost estimation methods such as Bayesian Network(BN)、COCOMOⅡ、 COBRA and COCOMO 81. Then the features and advantages of grey-based rough set approach are summarized.Finally, the grey-based rough set model and the approach for evaluation of software cost estimation methods are verified by simulation experiment, which proves the model’s validity and reasonability. The results of evaluation of four cost estimation methods are obtained by way of analyzing the experiment. The results indicate that grey-based rough set model is an effective approach to evaluate software cost estimation methods when the given information is insufficient and uncertain.