ENHANCING SOFTWARE PROJECT EFFORT ESTIMATION (SPEE) USING NEURO-FUZZY SYSTEM

Abstract

Accurately estimating software development effort (SDE) is important in effective project management processes like budgeting, project planning, and control. To realize an accurate estimate some algorithmic estimation techniques proposed to eliminate the inaccuracies of estimation. Constructive Cost Model (COCOMO) is a parametric or algorithmic model wont to estimate software effort & cost. However, thus far no model has proven successful to effectively and consistently predicting software effort & cost. Parametric models are considered vulnerable when faced with the matter of non-linearity of the complex within the parameters. In recent years, some estimation techniques appear using intelligent systems to predict software effort. The soft computing techniques were rarely applied for such a problem and their performances haven't been well investigated employing a systematic procedure. This research looks into the accuracy and stability of a selected soft computing method for the matter of effort estimation. ANFIS (Adaptive Neuro-Fuzzy Inference System) belongs to the family of a fused neuro-fuzzy system in which the fuzzy system is incorporated in a framework that's adaptive in nature. In this paper, we've performed neuro-fuzzy computing through ANFIS using Cocomo_Nasa & Cocomo81 benchmark datasets by comparing the anticipated and actual data, results indicate that the modeling techniques have comparable performance metrics, and be able to efficiently used for estimating the software development effort. Thus, one can use this neuro-fuzzy framework to urge better leads to comparison to ANNs and symbolic logic alone.