Investigating the Impact of Functional Size Measurement on Predicting Software Enhancement Effort Using Correlation-Based Feature Selection Algorithm and SVR Method

摘要

The software Functional Size Measurement (FSM) is one of the major factors affecting the effort estimation. Several FSM methods have been proposed since they are useful when development/ enhancement effort must be estimated. However, the greatest challenge for the project managers and other stakeholders is how to identify the effectiveness of an FSM method and select an accurate enhancement effort prediction model. There is only one 2nd generation FSM method- the Common Software Measurement International Consortium (COSMIC) method and four first-generation FSM, including the International Function Point Users Group (IFPUG) method. The main goal of this paper is to investigate the effectiveness of the first and the second FSM generations, respectively, IFPUG and COSMIC methods for sizing functional changes, and their use for predicting software enhancement maintenance effort. In this paper, the Correlation-based-Feature Selection (CFS) algorithm is combined with the Support Vector Regression (SVR) model. The dataset used for training and testing the prediction model is obtained from the International Software Benchmarking Standards Group (ISBSG) Release 12. To make comparisons between the impact of enhancement functional size generated from COSMIC and IFPUG methods on the enhancement effort prediction, two types of experiments were conducted, one with the use of the IFPUG and the other with the use of the COSMIC. Results show that the use of COSMIC functional change size measurement as input for predicting enhancement effort provides significantly better results than IFPUG in terms of MAE (Mean Absolute Error) = 0.0382 and Root Mean Square Error (RMSE) = 0.1082.