Improving estimation accuracy by using case based reasoning and a combined estimation approach

Several models of software estimation are available in the industry today. Models are available for estimating project cost, schedule, defects, personnel required etc. Available estimation models can be classified as being analogy based, parameter based, expert based and size based. Experimental results show that different models yield different results of estimates for the same project. In this paper, we demonstrate a technique that combines two available techniques. Our approach permits adding parameters for estimation dynamically based on the context of the project using case based reasoning. These parameters are used to compute a similarity index, which in turn, is used for estimation. Estimates can also be revised based on delay causing incidents that occur during the execution of the project. Data from about 10 projects executed in the past shows that estimates developed using our approach can converge to actual values over a period of time.