Cost Estimation of Software Project Development by Using Case-Based Reasoning Technology with Clustering Index Mechanism

• Main Authors: Zheng-Wei Huang, 黃政偉
• Other Authors: Deng-Yiv Chiu
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/45696991933647380028

碩士 === 中華大學 === 資訊工程學系碩士班 === 89 === In the business environment, it is common situation for many projects to use the same resources. Therefore, it becomes important to control the resources and avoid confliction. In other words, the allocation of resources becomes significant. Such planning involves that it has to estimate how much cost can be used so that the consequent resources planning be done smoothly. Therefore, the cost estimation is a critical process before a project is going to develop. In this thesis, we propose a CBR-Based with clustering index mechanism (CBR-C) to improve the predictive error and help manager to handle what significant features should be noticed. The object of CBR-C is to archive two points: (1) Reduce the estimative error and time. (2) Help the manager focus on the important resources and easily figure out the estimative process. Most models that applying case-based reasoning mechanism to estimate cost do not index their cases when they are applied to an unknown domain. Thus, in addition to provide numerical result, CBR-C introduces clustering method to index cases for finding optimal set of features that are used to help managers to figure what features are most important. Once a new case is going to be estimated, the classifier will first classify the new case to a cluster. Next, the classifier will retrieve the two mostly similar cases and input the two cases to the estimator with the new case. Afterwards, the estimator is responsible for outputting the estimative result and the generated optimal features to the user. Finally, CBR-C also utilizes the clustering method to generate the predictive error threshold for revising the averaged result. Through the experiment, it proves CBR-C providing suitable estimation accuracy and information.