Predicting Code Change by Using Static Metrics

Abstract

Maintenance of software is risky, potentially expensive – and inevitable. The main objective of this study is to examine the relationship of code change, referred to as maintenance effort, with source-level software metrics. This approach varies from the typical approach of evaluating software metrics against failure data and provides a different angle on the validation of software metrics. The goal of this study is to show through exhaustive data mining that a relation between software metrics and code change exists. Once this connection is established, a set of software metrics is identified, which will be used in further studies to predict code change in problematic modules identified by the software metrics at an early development stage.