An Approach to Outlier Detection of Software Measurement Data using the K-means Clustering Method