Class complexity interaction point size measure for object-oriented software development effort estimation

• Main Author: Koh, Tieng Wei
• Format: Thesis
• Language: English
• Published: 2012
• Online Access: http://psasir.upm.edu.my/id/eprint/39339/1/FSKTM%202012%2034R.pdf; http://psasir.upm.edu.my/id/eprint/39339/

Software development efforts estimation is a practical process of predicting the most realistic use of effort required to develop or maintain software based on incomplete and uncertain user requirements. Effort estimates may be used as input to project plans, budgets, pricing processes and bidding rounds. Many estimation approaches have been proposed to address the problem of effort estimation for software development projects since 1960s. Up to date published surveys show that most of the research has focused on the construction and refinement of formal software effort estimation models such as parametric models and size-based estimation models. In this research, we focus on the size-based estimation models for object-oriented development paradigm effort estimation, where the quantification step is based on mechanical processes to overcome the four limitations found in most of the traditional size-based estimation models. As the name implies, software size measure is the core instrument for size-based estimation models, and its accuracy determines the reliability and usefulness of the effort estimation model. The Class Complexity Interaction Point (CCIP) size measure has been proposed for object- oriented development paradigm effort estimation from the three class level size measures. In particular, the Class Functional Weights (CFW) measure is derived from Number of Method (NOM) and Number of Attribute (NOA) defined in the classes, while the Interaction Weights (IW) measure is calculated based on the maximum number of possible interactions through method call between pair-wise classes. As much as 68 system modules which are collected from the six different object oriented software systems have been used in this research. The recruited software professional team from the local software industry has provided us the value of actual development efforts. Preliminary correlation coefficient analysis between both of these independent variables (CFW, IW, CCIP) and the actual effort (person hour) was investigated not only to identify the ability of these measures in predicting development effort but also to suggest the type of effort estimation model that should be proposed for CCIP size measure. The suggested models are further validated using formal statistical hypothetical tests. N-fold cross validation technique was used to measure the performance of the CCIP-based predictive models. The results show a significant accuracy for CCIP-based effort estimation models in predicting the development effort. The predictive accuracy results confirm that the CCIP size measure is reliable and useful as an indicator for early effort estimation under object-oriented development paradigm.