Token based approach for cross project prediction of fault prone modules

A fault in a module may cause failure of the system. Faults pose a great challenge for the testing and maintenance phase. Fault prediction is of utmost importance to reduce the costs involved in the software development life cycle. Most of the studies use past history of the projects for predicting fault prone modules. Not much higher results are recorded in case of cross project prediction. In this paper, researcher tries to explain the various metrics and classification techniques used for fault prediction. It also describes a list of metrics to improve the results for cross project prediction. Further, an integrated approach is proposed which uses a metric computed from tokens along with the conventional metrics.