Application of Improved LM-BP Neuron Network in stock prediction

摘要

The prediction of stock closing quotation is very difficult because the time series seem random but not random completely. So a Back-Propagation Neurons Network based on Improved Levenberg-Marquardt algorithm was presented in this paper. The nonlinear mapping characteristic of the Back-Propagation Neural Network makes it can approach to every function in any precision. The Improved Levenberg-Marquardt algorithm can accelerate the speed of convergence, so it is easy to approach to the global optimal solution, not the local optimal solutions and. The prediction experiment shows that the Improved Levenberg-Marquardt algorithm works better than traditional Auto-regressive Moving-Average model.