A Hybrid Gaussian Membership Function (GMF) and Fuzzy based Cost Drivers for Effective Software Cost Estimation: An Application Software Perspective

Dr. Abdullah Al Hussein

Software Cost estimation and evaluation is an important and essential process in the SDLC. The availability of models such as COCOMO have revealed a lot of possibilities in arriving at a optimized cost function. It is a challenging process for understanding the SRS with appropriate time bound and staffing. Also, the use of Trapezium member function results in a situation where the attributes are assigned a maximum compatibility where this should be minimum at most of the times. In order to overcome these issues, this paper presents a hybrid model comprising of the Efficient Cost Predictive Fuzzy Model (ECPFM) and the GMF for addressing the cost drivers is proposed. Unlike the conventional neural approach, the proposed model is easy for interpretation and validation by non-professionals. The model is intended to manage the in-efficiency of the un-accuracy of certain inputs of the application software and improves the Application software consistency by optimized cost estimation. Continuous variables are used as the input for the application and also use linguistic variables which avoid the issue of application software having large variation in evaluated costs. The experimental evaluation of the proposed model minimizes the MMRE by 15.66% and MRE by 31.2%. The prediction (PR) is also improved by 12.56% and the Evaluation Function by 6.89%. It is also found that the GMF performs better than the other geometrical member function as it efficiently demonstrates smoother transitions in the regular intervals which achieve more closer results to the actual cost.

Volume 12 | Issue 8

Pages: 56-66

DOI: 10.5373/JARDCS/V12I8/20202446