Using Combinations of Bio-inspired Feature Selection Algorithms in Software Efforts Estimation: An Empirical Study

Feature selection algorithms are used to extract the most relevant features from a dataset and filter those which may affect the accuracy of the estimation techniques. A variety of features selection algorithms are investigated in the context of software engineering; however, the use of the bio-inspired algorithms has become very popular in the last few years, particularly in the domain of the software efforts estimations. However, sometime a single and standalone bio-inspired algorithm is not fit enough to select the most relevant features from a particular dataset and hence, the researchers have started to employ the hybrid of these algorithms. In this paper, we have investigated the use of various combinations (hybrid) of bio-inspired algorithms with the variety of datasets used and have compared their performance with the state-of-the-art standalone bio-inspired feature selection algorithms. The datasets employed in this case are Albrecht, China, COCOMO, Finnish, Kemerer, Maxwell, Miyazaki and NASA and the estimation techniques used are Support Vector Regression (SVR) and Random Forest (RF). We have obtained different results based on the datasets and estimation techniques.