Polynomial prediction of neurons in neural network classifier for breast cancer diagnosis

Post hoc evaluation mechanisms are utilized for determining the configuration of classifiers. Heuristic approaches mean that sub-optimal configurations could be used; resulting in lost training time, sub-optimal performance and can result in inappropriate results especially for large complex datasets. This paper proposes a new technique to determine the number of neurons in feed forward neural network on two large-scale breast cancer datasets. Classification accuracy of 86% and 89.17% was achieved and the technique predicted the upper and lower bounds for neurons in the feed forward neural networks.