A Radial Basis Function Neural Network approach to detect novelties: Applications on health datasets

In this paper we propose a novelty detection approach based on radial basis function neural networks (RBFNN), Markov chains and entropy. We employed it to model novel states and temporal relationships of health datasets.We conduct experiments with one synthetic dataset and three real-world ones available at the UCI repository. For every experiment we present accuracy, precision, recall, specificity, false positive rate, false negative rate and f-measure. Results are promising and confirm the benefits of the proposed approach.