A Researth On The Improved VWAP Strategies Based On The Volume Decomposition

As the rapid development and widely used of Algorithmic trading in the stock market, the competition between the institutional investors in the algorithmic trading strategy became more and more intensely. VWAP(volume weighted average price) strategy as the most widely used trading strategies on the market is aimed to get the market average price strategy, is one of the hot research topic in the algorithmic trading strategies. Using VWAP strategy to trade big orders can reduce impact cost while hidden the intention of big deals, forecasting the distribution of the stock’s intraday volume is a prerequisite for the implementation of VWAP strategies. The classical VWAP strategy use static average prediction method, the improvement of the research literature in recent years shows that most of the VWAP strategy using real-time information of high-frequency data dynamic adjustment the predictions.The major forms to forecast the volume distribution is decomposed the volume into the market part and special part, subsequently predict the distributions, and then can according to real-time information dynamic adjustment the predictions. In detail investigate the modeling method and result of volume prediction model of the existing VWAP improvement strategies, we find the existing model is limited by the sample and is difficult to apply to most stocks, the reason is while decompose the volume is not effectively separate the periodic structure of stock’s intraday volume, and affect the subsequent modeling prediction. To solve this problem, we assume that the volume of orders in the arrival process is the basis of a doubly stochastic Poisson point process, from the view of the relationship between volume changes of stock itself, we deduced a relationship with stock market turnover, in order to separate the impact of intraday periodic structures by the market turnover, we propose a new volume decomposition model to divided the volume into market portions forecasting with the historical average datas and special portions using ARMA model to prediction. There are two ways to build volume sequence representate the market trends, the principal component analysis and the sample weights, By dynamically adjusting the predicted values, Designed dynamic VWAP strategy-one and strategy-two.The empirical results show that our volume decomposition model can effectively separate the "U" type periodic structure of stock’s intraday volume. Comparing with the traditional VWAP strategy, our dynamic VWAP strategy has not only broader applicability but also smaller tracking error, and effectively reduces the market impact cost. Further analysis of the data found that in the aspect of forecasting the intraday volume distribution, Strategy One(PCA) has the smaller standard deviation of error, Strategy II(capital weights) has the smaller average error and the higher probability to overcome the traditional strategy. The model with smaller standard deviation is better at predicting the samples which are difficult to predition; the model with the higher dominant probability is better at predicting the samples which are easy to predition. The tracking error of the exercise price is mainly affected by the standard deviation of the volume prediction model, the tracking error of Strategy One is smaller. The accurary of volume prediction inversely proportional to the standard deviation, the degree of improvement on the accurary is proportional to the average error and standard deviation of classical VWAP strategy and also proportional to the occupation probability; the tracking error with a similar linear relationship, and the sample has the larger mean and standard deviation of volume, the more reduction in tracking error.