Research On Application Of Shanghai Stock Index Prediction Model Based On PSO-BP Neural Network

With the rapid development of the national economy and the continuous improvement of the market economy,stock investment has become an important part of modern people's investment and financial management.The volatility of stock prices directly affects the stability of the stock market and the healthy development of finance and economy.Successful forecasting of stock prices and trends help investors to make profits,and it also helps government departments provide timely and reasonable market guidance and supervision.However,the stock market is an extremely complex dynamic system,which is not only influenced by the internal factors of the company's own development,but also by the complex external factors such as macroeconomics and investor psychology.Its nonlinearity,turbulence,high noise and other factors determine that forecasting stocks is a difficult and complicated process.So,how to predict the stock market is an important and valuable issue in the financial community.The neural network simulates the structure of human neurons.Because it has strong self-learning,self-organization,and memory,it can predict the future development trend of stock prices from the historical data and relevant information of the stock market.Back-propagation neural network(BPNN)is a model of neural network with infinite approximation to nonlinear system and good fitting goodness.Therefore,BP neural network is suitable for dealing with stock price prediction problems with high complexity.This paper elaborates the basic theory of BP neural network in detail,and analyzes that the gradient descending algorithm it relies on has strong local search ability,but it has weak global search capability,and the weight easily falls into the local minimum.The Particle Swarm Optimization(PSO)algorithm is proved to be fast convergent and effective at the initial stage of global search,but its disadvantage is that the local search ability is weak.Based on the advantages and disadvantages of the two,this paper proposes a hybrid algorithm that combines PSO algorithm with BP algorithm and is used to train the weights of feed-forward neural network,and it is called PSO-BP algorithm.The obtained hybrid algorithm not only can use the powerful global search ability of PSO algorithm,but also can utilize the strong local search ability of BP algorithm.This paper uses PSO-BP neural network to make empirical analysis ofShanghai Stock Index'closing price,and discusses the selection of sample indexes and data;the topology structure of the network;the number of hidden layer nodes,learning factors,and the selection of activation functions.At the same time,we also give the GA-BP neural network algorithm that is also a smart hybrid algorithm and compare with it.The experimental results show that the hybrid PSO-BP neural network algorithm is superior to the GA-BP neural network algorithm and BP neural network algorithm in the simulation and prediction and can accurately predict the closing price of the Shanghai Stock Index and the future ups and downs in the near future.The model has a good application prospect.