Investigations on vector quantized and EM clusters with deep neural learning classifier and fuzzy analogy with firefly optimization for software effort prediction

Abstract

Software effort estimation is a procedure for estimating the quantity of effort desired to make the software system and its duration. It is one of the basic project management processes performed to support resource allocation activities effectually. Software effort estimation computes the project cost, time and quality necessary to achieve a specific task on the software development life cycle. Software cost estimation guides and supports the development of software projects. However, estimation of the accurate effort for developing any software project is difficult and may lead to project failure if estimation is not correct. Therefore, there is a requirement of efficient management of a software project which performs reliable estimates of effort to conduct the project and develop the software. This estimation can be done with the aid of data mining techniques and analogy based methods. Thus, the research work focuses on developing the effort estimation by performing classification, clustering and optimization technique in order to enhance the prediction based on analogy. At first, the proposed Multivariate Linear Regression for Software Effort Estimation (MLR-SEE) technique is introduced for knowing the classification accuracy in order to choose the best effort by using software projects. The proposed MLR-SEE technique is selected based on the correlation coefficient measure. If the correlation value is positive (i.e.,), then the relationship is stronger and thus obtains a better prediction model. Or else, the relationship is weaker when the correlation coefficient is negative (i.e., -1). According to the correlation coefficient results, the effort estimation is obtained with a lesser error rate and enhanced classification accuracy newline