Reliability Evaluation and Prediction of Improvable Information and Communication etworks

This paper targets to propose an efficient reliability evaluation and prediction scheme based on concepts of improvable information and communication networks (ICN). In this order of ideas, several black-box models of reliability growth are proposed in the literature. Remarking that a wide variety of methods, in the literature, suppose restrictive assumptions on the manner to discover ICN failures, we propose an alternative approach in accordance with observable dysfunction and improvement profiles of the analyzed network. Thus, we define a non-parametric model of ICN reliability and we consider a nonparametric statistical way to evaluate and predict the reliability during improvements with different strategies of correcting ICN malfunctions. This method allows us to study the evolution of the reliability growth without introducing too restrictive hypotheses. We show that the maximum likelihood estimate under improvement restrictions is equivalent to the estimate obtained by the antitone regression and we confront our model to others by using the u-plot method