Hyper-quadtree-based K-means algorithm for software fault prediction

Abstract

Software faults are recoverable errors in a program that occur due to the programming errors. Software fault prediction is subject to problems like non-availability of fault data which makes the application of supervised technique difficult. In such cases, unsupervised techniques are helpful. In this paper, a hyper-quadtree-based K-means algorithm has been applied for predicting the faults in the program module. This paper contains two parts. First, the hyper-quadtree is applied on the software fault prediction dataset for the initialization of the K-means clustering algorithm. An input parameter D governs the initial number of clusters and cluster centers. Second, the cluster centers and the number of cluster centers obtained from the initialization algorithm are used as the input for the K-means clustering algorithm for predicting the faults in the software modules. The overall error rate of this prediction approach is compared with the other existing algorithms.