Study On Dynamic VWAP Strategy Based On Support Vector Machine

The development of computer science and technology has brought about great changes to trading in financial markets.The algorithmic trading as an important part of programming trading is more and more important for institutional investors,which in order to get the market average price for the purpose of VWAP algorithm is the most widely used.The core idea of VWAP algorithm is putting a large order split into multiple small orders according to certain rules,and each order in the set time slice to hide large market shocks in the process of trade and reduce trade cost.Including the intraday volume distribution in VWAP strategy is the core of the strategy to construct prediction effect directly affect VWAP strategy executive pay.The traditional VWAP strategies tend to adopt static average prediction method to predict its intraday volume step by step,which only consider the historical data and not included in the real-time trading data,makes the forecast effect is poorer.So the improvement of this paper is focus on using high-frequency data for dynamic forecast volumes.In this paper,based on the static VWAP strategy,the introduction of support vector machine(SVM)using machine learning methods to construct A dynamic model to forecast the market trading volume,and use A stock market high frequency data for empirical testing,have certain reference value for investors.This paper is divided into four parts in structure:the first part for algorithmic trading related history and literature review,mainly introduced the definition of algorithmic trading,development and application in our country.Focus on static VWAP strategy and its improvement strategy,lead to trading in the related theory and support vector machine(SVM)modeling groundwork for the following.The second part is the mathematical modeling,this paper based on static VWAP iterative dynamic forecasting model is constructed,and intraday volume distribution by support vector machine(SVM)regression prediction,by market signals and adjust the volume factor to control each time period in order to reach the purpose of optimizing VWAP strategy.The third part is the empirical test:First of all,this paper selected the 10 stocks of Shanghai stock exchange from January 2017 to April’s high frequency data as experimental samples,using AR,MA and static prediction model based on support vector machine(SVM)respectively to predict its intraday trading volumes and comparison,the distribution of prediction effect of the latter is the better results were obtained.Then use the predicted results as the initial deal curve generation into the dynamic prediction model,the adjusting factor and the dynamic model of the market signals to make optimal model.Finally this paper use the improved dynamic simulated trading VWAP strategy,A detailed evaluation of the proposed model in the a-share market operation.The fourth part is the article conclusion:summarizes the work,this paper illustrates some problems in the process of building model and points out that dynamic volume prediction model for the future improvement direction.