Topology design of bankruptcy prediction neural networks using Particle swarm optimization and backpropagation
Fatima Zahra Azayite, Saïd Achchab · 2018
In this paper 1, the contribution of Particle Swarm (PSO) Optimization is studied to define the topology of artificial neural network. Moreover, we evolve the convergence of the PSO-PSO algorithm by presenting a new approach based on PSO and Backpropagation. Through this second algorithm, we studied the contribution of Backpropagation, with its strong ability to find local optimistic results, as particles initialization algorithm, to accelerate the convergence of the PSO to global minima. This hybrid model shows a high performance even in a context of missing data.