Fuzzy expert system models for software size and effort predictions

Predicting an accurate cost for system development is critical to the success of large scale information system projects. Software cost estimates are used to plan, schedule, manage resources, monitor and control the project. Realistic estimates are important to both customer and developer. Fundamental to software cost models is the need to accurately predict the size of an application. Software cost models use these sizing estimates along with project specific attributes to then predict the amount of effort required to complete the project. Two widely used software costing models are Function Points Analysis for predicting size and the Constructive Cost Model (COCOMO) for predicting effort.; Software project sizing is done early in the project life cycle when the least is known and project uncertainty is greatest. Current software cost modeling techniques do not provide for uncertainty in their inputs, processing or outputs. Fuzzy set theory is used to capture imprecision in inputs and outputs of models, and fuzzy logic/expert systems is used as a method of reasoning with imprecision.; This research develops new models for fuzzy function points and fuzzy COCOMO, which demonstrates the use of fuzzy logic as a means of capturing and reasoning with uncertainty in software cost models and demonstrates their benefits.