Research On Stock Pairing Trading Strategy Based On CS-LSTM

Pairs trading is a popular trading strategy that exploits market inefficiencies to make a profit.The idea of pairs trading is very simple.First,find two stocks with similar price movements.When the price difference between the two stocks exceeds a certain threshold,take a long or short position on each stock.Until the price difference between the two stocks returns to a normal level,then the opposite operation is performed on the two stocks.Neural networks show great superiority in prediction because of their unique nonlinear adaptive processing capability.In general,financial time series prediction needs to be completed based on the dependence between historical data and time series,while traditional neural networks cannot do this.Using previous events to infer subsequent events is usually done using a recurrent neural network(RNN).In addition,due to the problems of gradient disappearance in the traditional RNN model,a special RNN model,namely long and short term memory(LSTM)model,is proposed.In this paper,cuckoo search algorithm(CS)is also used to optimize the LSTM neural network model.In other words,the time series is input into the LSTM neural network,and then the optimal hyper-parameters of the LSTM neural network based on the cuckoo search algorithm are found.In this paper,CSI 300 are taken as the research object of pairs trading strategy.Firstly,the correlation coefficient interval was divided based on the correlation analysis,and then the paired stocks with good co-integration relationship were selected in different intervals.The LSTM neural network model is used to predict the price difference sequence of paired stocks,in which the cuckoo search algorithm is used to select the hyper-parameters in the LSTM neural network model.According to the prediction results of the model,the empirical analysis of pairs trading was conducted,and the results showed that compared with the BP neural network model,using CS-LSTM neural network model to forecast,the root mean square error of the prediction was the smaller,and its accuracy was higher,so the CS-LSTM neural network model of prediction ability was superior to the BP neural network model,andthe pairs trading strategy of stocks based on CS-LSTM neural network model could obtain certain benefits.At the end of this paper,the effectiveness and robustness of the pairs trading strategy based on the CS-LSTM model was tested.The results showed that,based on different periods,the trading strategy can obtain certain returns,and the trading strategy is also feasible when paired stocks are selected from different markets,such as a-share market and h-share market,and the trading strategy has certain robustness.