A Fuzzy Approach for Integrated Measure of Object-Oriented Software Testability

Abstract

For large software systems, testing phase seems to have profound effect on the overall acceptability and quality of the final product. The success of this activity can be judged by measuring the testability of the software. A good measure for testability can better manage the testing effort and time. Different Object Oriented Metrics are used in measurement of object-oriented testability but none of them is alone sufficient to give an overall reflection of software testability. Thus an integrated measure considering the effect of all these measures is required to well define the testability. The paper combines OO software metric values into a single overall value (called Testability Index) that can be used to calculate the testability of a class. The approach uses fuzzy techniques and concepts (fuzzification of crisp metric values, inference and aggregation, defuzzification of fuzzy output). We include empirical data of testing time of 25 different Java classes, which proves that individual metric values are not sufficient to arrive at the testability of a class and validates the testability index as a good integrated measure for arriving at the testability of the class.