Evaluating software reliability: Integration of MCDM and data mining

Although software reliability can be evaluated by applying data mining techniques in software engineering data to identify software defects or faults, it is difficult to select the best algorithm among the numerous data mining techniques. The goal of this paper is to propose a multiple criteria decision making (MCDM) framework for data mining algorithms selection in software reliability management. Through the application of MCDM method, this paper compares experimentally the performance of several popular data mining algorithms using 13 different performance metrics over 10 public domain software defect datasets from the NASA Metrics Data Program (MDP) repository. The results of the MCDM methods agree on top-ranked classification algorithms and differ about some classifiers for software defect datasets.