A systematic approach towards development of universal software fault prediction model using object-oriented design measurement

Abstract

A new systematic approach towards development of a software fault prediction model for trustworthiness and excellence enhancement of software systems by predicting fault-proneness in different severity level of module after coding has been performed in this paper. The classification ability of data mining technique and important object-oriented metrics of object-oriented software system are used to develop this fault prediction model. This model predict the severity level of object-oriented software module based on the number of faults in nominal, low, mid, high and not fault prone (NFP). A different decision tree like Hoeffding tree (VFDT) is created for existing project data Eclipse 2.0 in order to gain information for the purpose of decision-making whether a particular module is faulty or not, and if faulty then in which category. The rules that are obtained from VFDT decision tree algorithm are transformed into fuzzy rules and join together with the fuzzy inference system (FIS) of soft computing approach to predict fault-proneness of object-oriented module in different categories of fault or also not fault free for target data that is other two version of existing project data Eclipse 2.1 and Eclipse 3.0. The main aim of developing this fault prediction model is that to give a direct help to software manager in identifying fault-prone module by including this model before testing phase and just after coding phase so that testing effort can be reduced. And indirectly, this model helps the common people in getting reliable and quality-based system.