Many software reliability studies attempt to develop a model for predicting the faults of a software module because the application of good prediction models provides important information on significant metrics that should be observed in the early stages of implementation during software development. In this article we propose a new method inspired by a multi-agent based system that was initially used for classification and attribute selection in microarray analysis. Best classifying gene subset selection is a common problem in the field of bioinformatics. If we regard the software metrics measurement values of a software module as a genome of that module, and the real world dynamic characteristic of that module as its phenotype (i.e. failures as disease symptoms) we can borrow the established bioinformatics methods in the manner first to predict the module behavior and second to data mine the relations between metrics and failures.

A software package evolves in time through various maintenance release steps whose effectiveness depends mainly on the number of faults left in the modules. Software testing is one of the most demanding and crucial phases to discover and reduce faults. In real environment, time available to test a software release is a given finite quantity. The purpose of this paper is to identify a criterion to estimate an efficient time repartition among software modules to enhance fault location in testing phase and to reduce corrective maintenance. The fundamental idea is to relate testing time to predicted risk level of the modules in the release under test. In our previous work we analyzed several kinds of risk prediction factors and their relationship with faults; moreover, we thoroughly investigated the behavior of faults on each module through releases to find significant fault proneness tendencies. Starting from these two lines of analysis, in this paper we propose a new approach to optimize the use of available testing time in a software release. We tuned and tested our hypotheses on a large industrial environment.

The purpose of this paper is to put forward a methodology based on discriminant statistical analysis, which, by evaluating a series of structural parameters of a program, is able to predict its risk level, namely how prone it is to containing faults.The metric was constructed in an experimental context in which the high number of available observations (almost 350,000 lines of code) allowed us to divide the body of data into two parts, one for the effective creation of the model, the other as an objective, statistical means by which the proposed methodology could be evaluated.The conclusions we have reached allow us to assert that the basic assumption holds true, and that this particular type of analysis can be used in a fixed environment, during the release and testing of software as a predictive metric for the early identification of dangerous programs.