Mining Features of Software Reliability using Multimethod Data-Mining Approach

M. Mertik, M. Lenič, M. Zorman (Slovenia), and M. Pighin (Italy)

Keywords

Software, Reliability, Classifiers, Code metric, Data mining, Multimethod

Abstract

Software development is a complex and complicated process, which is nowadays maintained trough different levels of the software lifecycle. During every phase of the cycle there is a threat of "the faults insertion phenomena", which is especially observed during the coding level. It has been shown that the pattern of "the faults insertion phenomena" can be dramatically minimized with the use of software assessment metrics that are related with measurable features of the code. In this paper we introduce the research on the extracted features of software reliability using dynamical combination of classifiers implemented in a novel Multimethod approach. We will present the results of extracted features for the particular industrial software environment based on the large database of modules in C programming language and compare them with the measurement of standard Halstead's and more modern alpha metrics of complexity. At the end we will present some related viewpoints and ideas for the future work.

Important Links:



Go Back


IASTED
Rotating Call For Paper Image