Efficient Software Cost Estimation Using Artificial Intelligence

Abstract

Accurate cost estimation is desired for efficient budget planning and monitoring. Traditional approach for software cost estimation is based on algorithmic models expressing relationship among different project parameters using mathematical expressions. Algorithmic models are parameter-based models and produce the best accuracy when these parameters are well defined and predictable. The fundamental factor governing project cost within algorithmic models is the software size, quantifiable either in lines of code or function points. Analogy based estimation and expert judgment-based estimation falls under the category of non-algorithmic models. Both algorithmic and non-algorithmic models can estimate project cost and effort required but are unable to face challenges arising due to dynamic user requirements, latest technological trends, and impact of cost drivers on estimation process. Different machine learning based approaches like fuzzy modelling, regression models, optimization techniques, and ensemble methods can be used to predict an estimate nearest to the real cost of the project.