Soft Computing Approaches for Software Project Effort Estimation (SPEE) - A Review
The main idea behind writing this paper is to exemplify and signify the systematic review or survey of software project based effort or cost estimation techniques and prototypes which relies on soft computing approaches. Basically, there are lots of models and frameworks available for computation of precision based calculation of software effort estimation or cost using different soft computing or machine learning approaches but in the clear term there is “No single” model exists which can generate accurate results that can tell exactly how many resources are required in stipulated time frame to build a software for various needs. In this research article, we have offered the Soft Computing Based Intelligent techniques to compute the software estimated effort in terms of cost from the analysis and exploration of various research papers from the reputed conferences and journals i.e. ProQuest, IEEE, ACM, Elsevier, Springer, and from other international journals of repute published since last two decades. This survey paper is broadly classified into the following mentioned techniques viz. Artificial Neural Networks (ANNs), Symbolic or fuzzy logic, Genetic Algorithms (GAs), Combinational techniques, Case base reasoning, decision tree, Regression and soft computing based techniques. These techniques or approaches are the primary source of computing the software effort or cost and thus our focus is on the importance of benchmark datasets that were used for training and testing of all the mentioned techniques. We find that, the benchmark datasets are Cocomo81, Cocomo Nasa, Cocomo Nasa2, Cocomo SDR, China, Desharnais. Also, the evaluation metrics MRE, MMRE, RMSE, R2, MARE and PRED (prediction level) are the prominent assessment metrics based on total count and used for assessing the final outcome. Furthermore, we found that each and every technique has its own merits and demerits and the hybridization of several approaches can be seen as a substitute to acquire realistic estimates. Additionally, this review or survey is going to be useful for researchers as beginners and will provide them the future directions. Besides, this would eventually lead to better predict, in the field of Software Development & Cost Estimation.