AN IMPROVED META-HEURISTIC ORIENTED EARLY SIZE ESTIMATION UTILIZING ABC

Abstract: In order to channelize a team’s effort in a<br> fruitful direction any software oriented environment, preanalysis<br> of software components and the budget has<br> always been useful practice. Estimation of the size<br> becomes a tedious task when the machine is not trained<br> well enough to provide good classification accuracy. In<br> order to attain significant result when it comes to early<br> prediction, the selection of the attribute set plays a vital<br> role. Feature set selection has gained popularity in the<br> last couple of years to make the classification process<br> more precise. This paper utilizes and enhances the<br> current behaviour architecture of Artificial Bee Colony<br> (ABC), a meta-heuristic inspired algorithm is being used<br> for feature vector selection. The data is further trained<br> and classified by multiple multi-class classifiers. The<br> evaluation of the results has been made on the base of<br> quantitative parameter analysis and the co-relation of<br> effort and size has also been presented. The paper utilizes<br> dataset supported by NASA research frames.<br> Keywords: Software Size Estimation, Meta-<br> Heuristics, Artificial Bee Colony (ABC).