The Research On Application Of LSTM Neural Network In Selecting Stocks

In recent years,Artificial Intelligence theories and techs are successfully used in many fields,such as speech recognition and the game of go between human fighters and computers.It is a new meaningful project on how to apply these AI techs to investment.On one hand,in the field of investment,choosing high quality stocks is one of the key for a good invest.It can help the investors save their efforts on researching plenty of stocks in the market.Therefore,properly using the computer techs into the stock selecting field is very important.On the other hand,BP neural network can not consider both time sequence feature and cross-sectional feature of the data.This weakness limit its application in stock market.Long-Short Term Memory neural net which get the function of sequence memory is a widely and successfully used Machine Learning tech in speech recognition.It is meaningful to use the LSTM neural net in stock selecting.On the application of the algorithm,this research firstly construct BP neural network and LSTM neural network which has the same structure and parameters such as training times and learning ratio,and use the same activate function in their common parts.Then I make the signals for selecting stocks based on the sharpe ratio.After that work,I input sharpe ratio of 5 years into two neural networks to finish the model's training step.Thirdly,I test the two algorithms and analyze the results.I find that LSTM neural network is better than BP neural network in selecting stocks.The high predicting accuracy make it posses strong applicability in real stock selecting process.Based on the above results,this research considers about the real process in stock choosing.From four sides including the feature of time sequence,nonlinear feature,stock selecting ability in amount of stocks in the market and different investors,I analyze the applicability of the LSTM neural network.Finally,after analyzing the results,,I give some advice to use it getting more profits when apply the LSTM neural network in selecting stocks.