Improving estimation accuracy of the COCOMO II using an adaptive fuzzy logic model

Software development time and cost estimation are the process of estimating the most realistic use of time and cost required for developing a software. It is one of the biggest challenges in the area of software engineering and project management, in the last decades. The software estimates are difficult to obtain due to incomplete software information is available in the early phase of software development process. Insufficient software information causes inaccuracy in software attributes. Thus, the vagueness and uncertainty of the software attributes is the main reason of inaccuracy of software estimates. Software cost estimation models such as regression model, expert judgment, SLIM, and COCOMO require accurate software attributes and long term estimation process, which are not completely achievable in early phase of software development process. Soft computing techniques such as fuzzy logic can reduce the vagueness and uncertainty of software attributes. Therefore, it may consider as alternative to decrease the inaccuracy of software estimates. This research aims to utilise an adaptive fuzzy logic model to improve the accuracy of software time and cost estimation. Using advantages of fuzzy set and fuzzy logic can produce accurate software attributes which result in precise software estimates. The Two-Dimension Gaussian Membership Function (2-D GMF) was used in the fuzzy model to make software attributes smoother in terms of the range of values. The COCOMO I, NASA98 data sets; and four project data from a software company in Malaysia were used in the evaluation of the proposed Fuzzy Logic COCOMO II (FL-COCOMO II). The evaluation of the obtained results, using Mean of Magnitude of Relative Error (MMRE) and PRED(25%) evaluation techniques, showed that the FL-COCOMO II produced the MMRE less than the original COCOMO and the value of PRED(25%) in the Fuzzy-COCOMO II is higher than the original COCOMO. Furthermore, the FL-COCOMO II showed 8.03% improvement in terms of- - estimation accuracy using MMRE when compared with the original COCOMO. Using advantages of fuzzy logic such as accurate estimation; adaption; understandability, and etc., can improve the accuracy of software estimates.