B-Maniac

Mobile Data Offloading permits users to use cheap communication media, whenever feasible, for delivering their personal data instead of using the infrastructure which is more expensive. Having good procedures to assess and compare the different possibilities a device has to send his data is crucial. In the following paper, we propose an evaluating approach that takes into consideration the topology changes history in order to provide an efficient way to calculate the quality of a specific medium. In particular Bayesian Networks are used as basis to provide our solution to the Maniac Challenge 2013. Bayesian Networks is a statistical model for the generation of inferences starting with very little information. In the next sections we will give a definition of the Bayesian Networks focusing on the simplest and computationally efficient of the Bayesian Networks versions, named Naı̈ve Bayes. Finally, we will show our method for the challenge to evaluate the medium and the bid strategy. I. BAYESIAN NETWORKS Bayesian Networks (BN) [3] aid the reasoning and decision making in an uncertain environment. BN are built on top of a graphical model representing a set of random variables and their conditional independence. The learning phase in BN consists in retrieving the probabilities and the inference phase consists in computing the probabilities a posteriori by applying the Bayes rules. BN are powerful in terms of inference, they are able to produce a decision pattern with very little information. Very efficient algorithms exist for computing the probability a posteriori. In the simplest BN model, all the variables are only dependent on the inferring variable. This model is called Naı̈ve Bayes (NB) and it is successfully applied to several domains, providing satisfactory results.