Portforlio Selection Based On Recurrent Neural Network

Portfolio's optimization selection and investment income by means of recurrent neural network have been discussed in this paper.First,we select 100 companies in Shanghai Stock Exchange,and reconstruct 10 technical indexes based on the 7 technical indexes we selected,we obtain 3 principal component data by means of data mining and principal component analysis.Nextly,we simulate the operation rule of securities market through recurrent neural network.By letting 9 principal component data of 3 days as input of network and rate of return for successive 2 days as out.put of network,we haven't stop training net,work until it satisfied our demand.As a result,we.gained the prediction model based on the recurrent neural network,then it was used to predict stock rate of return.In the end,we use a group of new exchange data as input of network;Then we take investment simulation on the basis of yield rate predicted.Results show that both rate of return and Sharpe ratio are all better than Shanghai index yield-Moreover,from the point of fundamental analysis,we use 11 financial index that obviously affected stock yield rate,predicting the medium term rate of return of the stock by means of recurrent neural network.Result show that,in the matter of yield rate and Sharpe Ratio,the predicted model we constructed equally better than Shanghai index yield rate.