Author(s)

Lin Jinjian

Corresponding Author

Lin Jinjian

Abstract

Unsupervised algorithms, such as clustering algorithm, could be used on the fault tag position for fault prediction of software module. A software fault prediction algorithm based on quadtree k-means clustering algorithm was proposed in the paper. The purpose of adopting quadtree mainly included two aspects: the first was to seek for clustering center required by k-means clustering algorithm using quadtree, and the second was fault prediction of software module using quadtree. In this algorithm, input threshold parameter decided the initial clustering center. Through changing the threshold parameter, users could get the expected center of clustering. The performance of the algorithm was measured using such a new standard as “clustering earnings”. Through simulation and comparison, it was discovered that the algorithm proposed in the paper had highest clustering earnings. Moreover, in most cases, the total error rate of the algorithm proposed in the paper was lower than that of other algorithms, which indicated the effectiveness of the algorithm proposed in the paper in the prediction of software fault.

Keywords

K-Means Clustering, Recommendation Algorithm, Software Prediction, Diagnosis