Research And Implmentation Of Program Trading System Based On Entanglement Theory

Stock market investment has gradually developed into a popular investment method,and more and more investors are participating in it.However,most people are often unable to make correct trading decisions due to the influence of the environment and their own emotions.Therefore,an automatic trading tool that can maintain stable return can help investors get out of trouble,and program trading came into being.As one of the most popular technical theories in the stock market,Entanglement Theory can effectively help investors make correct decisions by decomposing the stock market trend.Therefore,this thesis implements the buy-and-sell point recognition algorithm based on Entanglement Theory,and improves it by combining in-depth learning in the computer field,and finally implements a program trading system based on Entanglement Theory.The thesis firstly digs deeper into the theoretical knowledge of entanglement theory,learns the idea of decomposing trend,proposes the concept of reversal pivot based on the complex and changeable reality of the stock market,designs and implements the buy-and-sell point recognition algorithm by combining the knowledge of entanglement theory morphology and dynamics,and verifies the correctness of the algorithm by applying the buy-and-sell point recognition algorithm to multiple stocks.Then,this thesis analyzes the problems of recognition lag and failure of the buy-and-sell point recognition algorithm according to the classification and morphological extension characteristics in the entanglement theory,and proposes corresponding solutions.By predicting the last K-line of the classification after the formation of the pivot,the purpose of identifying the trading point in advance is achieved,and further predicting the next classification type based on the prediction identification result to reduce the failure rate of the trading point.At the prediction model level,this thesis proposes an online fast stock price prediction model based on the existing deep learning temporal series prediction algorithm,which ensures the model prediction speed without reducing its prediction accuracy based on the powerful storage mechanism of this model.Then,according to the characteristics of classification data,this thesis proposes a classification prediction model based on classifier ensemble learning,and constructs an ensemble voting classifier through the stacking of various ensemble learning classification algorithms and the mechanism of soft voting.Finally,based on the minute-level K-line and typing data,a control experiment was conducted out on the above two models,and the effectiveness and feasibility of the model was verified by comparing the performance of different models.Finally,this thesis designs and implements a program trading system with the improved buy-and-sell recognition algorithm as the core,and specifies the trading strategy based on the buy-and-sell.After the back-test analysis of the system backtest module,the feasibility and stability of the trading strategy are verified,and it is able to can maintain a stable return.