Test Set Generation and Reduction with Artificial Neural Networks

Mark Last, Prachi Saraph, Abraham Kandel

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review


Reducing the number of test cases results directly in the saving of software testing resources. Based on the success of Neural Networks as classifiers in many fields we propose to use neural networks for automated input-output analysis of data-driven programs. Identifying input-output relationships, ranking input features and building equivalence classes of input attributes for a given code are three important outcomes of this research in addition to reducing the number of test cases. The proposed methodology is based on the three-phase algorithm for efficient network pruning developed by R. Setiono and his colleagues. A detailed study shows that the neural network pruning and rule-extraction can significantly reduce the number of test cases.
Original languageEnglish
Title of host publicationArtificial Intelligence Methods in Software Testing
EditorsMark Last, Abraham Kandel , Horst Bunke
PublisherWorld Scientific
ISBN (Electronic)9789814482608
ISBN (Print)9789812388544
StatePublished - Jun 2004


Dive into the research topics of 'Test Set Generation and Reduction with Artificial Neural Networks'. Together they form a unique fingerprint.

Cite this