A New Approach To Predict Stock Price By Combination Of Artificial Neural Networks And Stock Correlation

As a complex system which covers a great deal of information,the stock market reflects the development and changes in the national economy.The stockmarket is the open market for trading companies' stocks and derivatives,providing fair investment opportunities for individuals,brokers and companies.Hundreds of billions of capital in the stock market around the world exchange every day,the performance of the stock market directly affect the development of the national economy.The volatility of stock prices is a signal of capital and liquidity allocation,which plays a key role in the decision of investors.An accurate forecast of the stock trend can help investors make timely and accurate decisions on the stock exchange to gain benefits.The stock market is influenced by many factors,such as major political events,economic policies,industry development and traders' expectation.Research shows that there is a correlation between stocks,especially in the industry sector,the price fluctuations of stocks are affected by other stocks.Financial time series data,especially stock data,are highly non-linear,multi-dimensional and time dependent,and are non-stationary and unstructured data,which brings many problems to the analysis and prediction of financial time series.However,Artificial neural network(ANN),as a highly complex non-linear system,has extensive adaptability and learning ability in dealing with non-linear systems,so it has been thought as the best forecasting method in the fields of stock market forecasting.Therefore,this paper presents a forecasting model based on stock correlation,which includes the mutual influence of price between stocks in the model consideration.We proposed a hybrid neural network forecasting model which combined the stock correlation,selecting the relevant stocks as input variables by measuring the correlations between one stock and other stocks through copula function.The forecasting perfonnance can be improved by choosing the high correlation stocks as the input variables with copula function.The experimental results show that the prediction model improves prediction accuracy significantly,and the validity of the model is verified by both statistical and financial performance.