Predicting Faultiness of Program Modules Using Mamdani Model by Fuzzy Profile Development of Software Metrics

This research seminar proposed and implemented a new approach toward reliability and quality measurement of software systems by building a fault prediction model and faultiness degree estimation before starting the testing phase. The main goals of this model were to support decision making with regard to testing phase which leads to reduce the testing efforts, and to optimally assign the needed resources for testing activities. This research used KC2 dataset originated from National Aeronautics and Space Administration (NASA) project to evaluate the predictive accuracy of the proposed model. Software metrics in this dataset are of fuzzy nature, consequently, this work used MATLAB system to build a Mamdani fuzzy inference model. Then, this research applied and validated a published methodology for fuzzy profile development from data as an important requirement to build the model. Moreover, the proposed model utilized the capabilities of k-mean clustering algorithm as a machine learning technique to extract the fuzzy inference rules that were also required to build the model. Finally, this paper used suitable approaches to validate and evaluate the model. Accordingly, the results show that the proposed model provides significant capabilities in fault prediction and estimation.