Portforlio Selection Based On Long-short Term Memory Neural Network

This paper discusses the optimal selection of investment in China's securities market from the view of technology analysis and basic analysis respectively.In terms of technical analysis,firstly,this paper sele.cts 7 transaction data and 10 fre.quently-used technical analysis indicators from 100 listed companies in Shanghai Stock Exchange.Secondly,three comprehensive characteristic variables are obtained bt using principal component analysis to reduce the dimension of 17 indicators.Then,the principal components are used as input of long-short term memory neural network.We gain a prediction model through constant training the network,which is based on the long-short term memory neural network.At last,the cumulative return rate reaches 0.1007 through simulation investment for 32 days,which is 158.2%and 21.4%higher than the results of previous studies.In terms of of basic analysis,in the same way,we can get that the cumulative return rate and sharp ratio reaches 0.7938 and 0.4721 respectively by two and a half years of simulation investment.Compared with the results of previous studies,the cumulative return rate is similar,but the risk is much lower.In a word,the prediction model established in this paper improves the results of previous studies and has important value.