Use-case based early Software Effort Prediction using Fuzzy Logic and Genetic Algorithms.

摘要

An important and challenging activity, Software Development Effort Prediction involves dealing with imprecision, uncertainty and dearth of information in the early stages of software development. With the focus shifting more towards the use of machine learning techniques, predicting effort using Fuzzy Logic, Neural Networks, Genetic Algorithms or a combination of these has also been heavily considered by the research community. This thesis presents an adaptive fuzzy logic based framework for use-case based effort prediction capable of handling imprecision and incorporating expert opinions. Additionally, a simplified framework is conceptualized and empirical evaluations regarding the impact of various objectives are investigated which show that the proposed frameworks are promising and produce acceptable results. Since prediction accuracy of a fuzzy logic based effort prediction system is highly dependent on the system architecture, the development of a genetic-fuzzy tool to evolve different architectures provides results pertaining to the impact of architectural differences on the accuracy of effort prediction systems.