Bayesian Neuro-Rough model

This paper puts forward a neurorough model which is a combination of a multi-layered perceptron and rough set theory. The model is formulated using Bayesian framework and trained using Monte Carlo method and Metropolis criterion. The model is then tested on an ante-natal dataset and is able to combine the accuracy ofthe multi- layered perceptron model and the transparency of rough set model. The proposed model gives 62% accuracy compared to 62% for Bayesian multi-layered networks trained using hybrid Monte Carlo and 59% for Bayesian rough set models.