Analyzing the effectiveness of the gaussian mixture model clustering algorithm in software enhancement effort estimation

Abstract

Background: The influence of data clustering on the effort estimating process has been studied extensively. Studies focus on partitioning and density-based clustering, and some use hierarchical clustering, but most focus on software development effort estimation. Aim: We focus on the Gaussian Mixture Model algorithm's effectiveness in the software enhancement effort estimation. Method: We used the Gaussian Mixture Model clustering algorithm to cluster the dataset into clusters and then applied the IFPUG FPA method for effort estimation on these clusters. The ISBSG dataset was used in this study. The number of clusters is determined using the Elbow method with the Distortion score. Besides, the k-means algorithm was also used as the comparative algorithm. The baseline model was determined by using the FPA method on the entire dataset without clustering. Result: With the number of clusters selected as 4, on six evaluation criteria, MAE, MAPE, RMSE, MBRE, and MIBRE, the experimental results show the estimated accuracy using the FPA method on clustered data significantly better when compared with no clustering. Conclusion: the software enhancement effort estimation can be significantly improved when using the Gaussian Mixture Model clustering algorithm.