Stock High Frequency Tranding Strategy Based On Price Change Decomposition Model

High-frequency trading data are generally securities trading data with a short time interval(less than one day).In contrast,high-frequency trading data are characterized by unequal recording interval,data discretization,daily existence mode and simultaneous occurrence of multiple transactions.In the analysis of high-frequency data,it is an important method to decompose the data with complex variation patterns,improve the accuracy and reliability of the analysis by integrating the models.The purpose of this paper is to study the high frequency trading data by apply the price change decomposition model.First of all,on the basis of introducing the basic principle of price change decomposition,the appropriate methods of modeling the decomposed parts are discussed.The study showed that Logistic regression was used to establish the model for the indicator variable of price change and the direction of price change.As for the size of price change,this paper conducted a detailed study on it.Three models were established by using 1+ geometric distribution,1+ poisson distribution and Logistic regression.Through comparison,it was found that Logistic regression had the best fitting effect on different types of data and had stronger compatibility with data.Then,the methods of estimation,verification,evaluation and variable selection of the integrated ADS model are presented,which systematically solves the problem of studying high-frequency data based on the price change decomposition model.At the same time,this paper used the Logistic regression model to fit the five-minute trading data of the bank of China,and calculated the correlation probability of whether the price changed,the direction of change and the size of change according tothe parameters obtained by the fitting,so as to develop a stock trading strategy based on the decomposition model of price change.In this strategy,the buying and selling points are partly determined by the indicator variable of price change and the direction of price change,and the proportion of funds used in buying is controlled by the magnitude of price change.Finally,in order to verify the feasibility of the strategy,this paper use the out-of-sample data of bank of China shares and the five-minute data of sinopec shares to verify the strategy.The results show that the trading strategy is profitable for different types of stock data.Based on the empirical research results,this paper summarizes the price change decomposition model and its trading strategy,points out the technical problems in the model and strategy,and points out the future improvement of the model and strategy as well as the research direction.