Algorithm And Its Application In Stock Forecasting Mixed Copula EDA-BP Optimization

Artificial Neural Network is a departure from the imitation of biologicalpoint.A neural network system have human perception and learning ability,withnonlinear mapping ability and self-learning ability to adapt well,it has beenwidely used in many fields.There are many intelligent optimization algorithm which is used tooptimize weights and threshold of the neural network,such as particle swarmoptimization algorithm,artificial fish swarm algorithm, genetic algorithm, andthe EDA which is a new evolutionary algorithm developed based on the geneticalgorithm,and get the new individual by establishing a probability model.in theEDA,the PBIL algorithm, the UMDA algorithm,the PBILc in continuousdomain,the MIMICc and the EGNA algorithm are feasible in optimizing theweights and threshold of the neural network, but the EDA in estimating theoperation of probability model is very complicated and have huge computation.The Copula theory is applied in the EDA that is a hot research now, the jointdistribution function of outstanding is decomposed into a Copula function andone-dimensional distribution function with multiple variables.The Copula EDAsimplify the process of probability distribution model and improve theefficiency and accuracy of the estimation.The Copula EDA optimize during thewhole solution space with characteristics of global search, but there are somedifficulties in finding the exact location of the solution. BP algorithm use thegradient descent method to search, especially suitable for finding exactsolutions of the local area.This paper studied the Copula EDA of artificialneural network and BP algorithm.This article’s main research work are shown as follows:1.Copula EDAoptimize neural network’s weights and threshold and are tested by the UCIdatabase standard classification datasets with comparing with the BPalgorithm,it can be seen the Copula EDA optimization is more stable, but havethe lack of accuracy.2.Based on The comprehensive advantages of Copula EDAand BP algorithm, the Copula EDA and BP algorithm are combined,and establish binding mode and cooperative mode to optimize neural network’sweights and threshold, the feasibility and effectiveness of two optimizationmodels are verified,The results show that the two mode have a goodoptimization results, and can improve the convergence speed and the accuracy.3.The two model which the Copula EDA and the BP algorithm are combined(beassisted binding mode and cooperative binding mode)are applied predict thestock’s closing price of the Shanghai A stock market and the stock of Lu’an.