Should Function Point Elements Be Used to Build Prediction Models?

摘要

Schedule and cost managements are indispensable to develop large software, and predicting of project results such as development effort is the basis of the management. So, high accuracy of project prediction is needed. Many researches about software project prediction use software size as a major explanatory variable. To estimate software size, function point method is often applied. In the method, five fundamental function elements are identified and scores called function point (FP) are given to each element. After that, the sum of FP is calculated. Generally, the FP elements are not used as independent variables directly, and only the sum of them is used as an independent variable. As long as we know, it is not clear to what extent FP elements affect quality, cost, and delivery of software project. Also, it is not clear whether using the elements as independent variables improves prediction accuracy of quality, cost, and delivery or not. So, this study analyzed the influence of the elements to quality, cost and delivery. Also, we evaluated the accuracy of software project prediction when the elements were used as independent variables. The goal of the study is to support data collection and variable selection of the prediction. Experimental results showed that when duration and effort were estimated by analogy based estimation, using FP elements as independent variables was effective to enhance the estimation accuracy.