An Efficient Medical Diagnosis Algorithm Based on a Hybrid Neural Network with a Variable Adaptive Momentum and PSO Algorithm

Neural Network has been used in a number of scientific fields including medical diagnosis. It is clear that classification is one of the main challenges in any medical diagnosis. However, the drawbacks of conventional neural network classifier are summarized as slow convergence and tendency to be trapped in local minima. The aim of this study is to present a hybrid Back-Propagation algorithm with a variable adaptive momentum (BPVAM) and particle swarm optimization (PSO) (BPVAM-PSO). The PSO is an efficient training algorithm as it does not require any complex calculations. The BPVAM variable momentum, in this case, will start with a high value, which increases the convergence rate; the momentum value will decrease as the error rate is decreased. The best weights of the BPVAM are passed to the PSO. The PSO uses the weights to determine the best set of parameters and as a result, the number of hidden neurons is reduced. This will improve performance further. To verify the efficiency of the BPVAM-PSO algorithm it will be used for the classification of a number of different medical datasets. The results of the proposed algorithm are compared against the performance of the conventional BP Neural Networks and BPVAM algorithm.