Risk Classification in Global Software Development Using a Machine Learning Approach: A Result Comparison of Support Vector Machine and K-Nearest Neighbor Algorithms

Abstract

Software development through teams at different geographical locations is a trend of modern era, which is not only producing good results without costing lot of money but also productive in relation to its cost, low risk and high return. This shift of perception of working in a group rather than alone is getting stronger day by day and has become an important planning tool and part of their business strategy. In this research classification approaches like SVM and K-NN have been implemented to classify the true positive events of global software development project risk according to Time, Cost and Resource. Comparative analysis has also been performed between these two algorithms to determine the highest accuracy algorithms. Results proved that Support Vector Machine (SVM) performed very well in case of Cost Related Risk and Resource Related Risk. Whereas, KNN is found superior to SVM for Time Related Risk.