Application Of Deep Learning In Quantitative Transaction

There are a large number of historical data in the field of financial transactions.These data resources provide a good data base for quantitative transactions.The financial foreign exchange market is a complex nonlinear dynamic system.The traditional machine learning algorithm is used to mine financial data.Its ability to express complex functions is limited under the condition of limited samples and computing units,and there are bound to be many limitations and deficiencies.However,deep learning can transform the characteristic representation of the data in the original space into the new characteristic space layer by layer.It has higher self-learning ability,stable performance and abstract simulation ability.It is more suitable for solving complicated nonlinear problems such as financial markets.Therefore,this paper uses the depth learning model to predict the foreign exchange price and the rising and falling trend.In the regression model of forecast price,the existing foreign exchange price and technical indexes are taken as the input data of the model,and the price is taken as the output data of the model.By changing the super parameters,the influence of them on the accuracy of the model is studied.Then,the optimal super-parameters are selected to determine the LSTM neural network model so that the prediction effect of the model is optimal.Finally,the LSTM model and the traditional RNN model are used to predict foreign exchange prices,and the prediction effects of the two models are tested.The experimental results show that both LSTM neural network model and traditional RNN neural network model have certain prediction effect on foreign exchange price,and LSTM model always has smaller prediction error and better prediction effect than RNN model.Therefore,LSTM neural network can be preferred when establishing regression model for price time series data by using depth neural network.In the classification model for predicting the price rise and fall trend,the existing basic historical price data of foreign exchange is taken as the input of the model,and the price trend(rise,shock and fall)on a certain day in the future is taken as the output of the model.The LSTM neural network model is determined by selecting the optimal hyper parameters through experiments.Then use this model to forecast the price trend of the first day in the future and record the accuracy of the model.Finally,on the basis of forecasting the trend of the first day in the future,the model is expanded to forecast the price trend from the second day to the fifth day in the future to compare the accuracy of the model.Experiments show that in the classification model for predicting the future price rise and fall trend,the accuracy of LSTM model prediction is relatively high in the short term and shows an upward trend,but will gradually decline in the later period.