A metadata-supported distributed approach for data mining based prediction in wireless sensor networks

This paper introduces a new approach to the problem of data mining in wireless sensor networks. The problem is fast emerging, due to recent technological advances in both data mining and wireless sensor networks. The basic, axiomatic, assumption for this research is that a typical decision making process converges faster if some positive knowledge is incorporated into the process, i.e., the knowledge that decreases the probability of wrong decision. The above implies that any data mining process can be improved, if metadata with positive knowledge are added to it. Consequently, the same applies if data mining is performed on top of some wireless sensor network infrastructure. The proposed algorithm is compared with the best one from the open literature, with the final goal to compare the two algorithms analytically and by simulation. It is expected that the proposed algorithm behaves better in all conditions, and much better in conditions of the suddenly changing environment.