CLUSTERING APPLICATION FOR RELEASE PLANNING

Abstract

The software projects are characterized by a long planning horizon, which entails the policy
of release management. A short-term version of this problem is known as the Next Release Problem. A central issue release planning is determining which features should be included in which releases. This problem
is NP-hard and thus cannot be solved analytically. To reduce the complexity, it is proposed to apply a simple
clustering algorithm. The similarity coefficient combines precedence based similarity, predecessor based
similarity and successor based similarity. The resource constraints for the particular release define a clear
cutting point for the cluster size. Proposed approach reduces the complexity of the problem from O(n!),
where n is the number of features, to O(m2), where m is the number of releases. One example is considered.
There is pointed that to improve the results of features allocation to releases the priorities of features should
be taken into account