VWAP Algorithm Of Dynamic Intraday Volume Percentages Forecasting Based On Machine Learning Methods

In recent decades,the rapid development of computer technology and internet technology has led to the rapid development of electronic transaction execution systems,which is called algorithmic trading.Algorithmic trading refers to the trading method in which in the financial market the investor issues a trading order through a computer program and uses a computer algorithm to determine the trading timing,price,number of orders,etc.The problem studied in this paper is VWAP algorithm trading strategy,which is the most widely used in the algorithm trading strategy.The execution effect of VWAP algorithm trading strategy depends largely on the intraday volume percentages forecasting.There-fore,the focus of this paper is to study the forecasting method of the intraday volume percentages.The intraday volume percentages forecasting model proposed in this paper uses the random forest and feedforward neural network method.The input of the model includes the volume of the historical trading days and the volume of the same interval in the past period,and the volume of the interval in the past several intervals.The output of the model is the intraday volume percentages of the interval.The data used in this paper is the 5-minute K-line data of the main contract of the Shanghai and Shenzhen 300 stock index futures of China Financial Futures Exchange.On this data,the model proposed in this paper is verified.Compared with the traditional rolling meaning method,the mean square error of the feedforward neural network method is improved by 9.90%,which proves the validity of the intraday volume percentages forecasting model proposed in this paper.Next,the paper uses the data of the main contract of the Shanghai and Shen-zhen 300 stock index futures to verify the correlation between the proportion of volume and the rate of return.In this paper,the Pearson correlation coefficient and the Spearman correlation coefficient of between the proportion of volume and the rate of return are calculated respectively,and then the p-value is calculated by the hypothesis test,which proves that between the proportion of volume and the rate of return not only have significant linear correlation,but also have sig-nificant nonlinear correlation.Then the paper adds the rate of return and the volatility calculated by the rate of return to the prediction model proposed in this paper.Relative to the rolling meaning method,the mean square error of the random forest method is improved by 12.90%.When using the feedforward neural network method at the same time,compared with the prediction model without the rate of return and volatility,the mean square error(relative to the rolling meaning method)of the new prediction model increases from 9.90%to 13.27%,which proves that using of the rate of return and the volatility to predict the proportion of the volume is effective.Finally,the paper calculates the VWAP of the forecasting model of the intra-day volume percentages with the rate of return and volatility and the VWAP of the traditional rolling meaning method,which are compared with the real VWAP of the market.Using the mean absolute percentage error as the evaluation stan-dard,compared with the traditional rolling meaning method,the mean absolute percentage error of the VWAP tracking effect of the prediction model with the rate of return and the volatility increases by 38.36%,which proves that the VWAP tracking effect of the model proposed in the paper is better.By analyzing the actual situation of the trading day with bigger VWAP tracking error,the paper finds that the transaction price begins to drop sharply at the middle of the trad-ing day,which leads to a surge in volume.At this time,the difference between the volume percentages forecasted by the traditional rolling meaning and actual situation is large.The volume percentages forecasted by the prediction model with the rate of return and the volatility can fit the actual volume percentages better,so VWAP tracking effect is better.At the same time,the feedforward neural network method is used to predict the intraday volume percentages,which lays a foundation for the application of more complex neural network methods in this field.