An Improved Fuzzy Approach for COCOMO’s Effort Estimation using Gaussian Membership Function

摘要

In software industry Constructive Cost Model(COCOMO) is considered to be the most widely used modelfor effort estimation. Cost drivers have significant influenceon the COCOMO and this research investigates the role ofcost drivers in improving the precision of effort estimation.It is important to stress that uncertainty at the input level ofthe COCOMO yields uncertainty at the output, which leadsto gross estimation error in the effort estimation. Fuzzylogic has been applied to the COCOMO using thesymmetrical triangles and trapezoidal membershipfunctions to represent the cost drivers. Using TrapezoidalMembership Function (TMF), a few attributes are assignedthe maximum degree of compatibility when they should beassigned lower degrees. To overcome the above limitation,in this paper, it is proposed to use Gaussian MembershipFunction (GMF) for the cost drivers by studying thebehavior of COCOMO cost drivers. The present work isbased on COCOMO dataset and the experimental part ofthe study illustrates the approach and compares it with thestandard version of the COCOMO. It has been found thatGaussian function is performing better than the trapezoidalfunction, as it demonstrates a smoother transition in itsintervals, and the achieved results were closer to the actualeffort.