REQUIREMENT LEVEL IMPACT ANALYSIS WITH IMPACT PREDICTION FILTER

摘要

Requirement change is a common phenomenon in the software development lifecycle. Ineffective requirement change planning prior to actual change implementation could end catastrophically. Many approaches have been developed to support effective change planning including those for predicting the impact of the change using class interaction prediction. However this approach tends to generate unnecessary or false prediction results. In this paper we propose a new impact prediction filter that is able to rule out those false results which indirectly improves the accuracy of the prediction. We have also combined the new impact prediction filter with our previous work on predicting class interactions from requirement interactions to become a new requirement level impact analysis approach. The contributions of the paper are: (1) a new requirement level impact analysis approach; (2) a new impact prediction filter and; (3) evaluation results that show the filter improves the accuracy of the impacted class prediction.