Having effective intraday forecast for the level of trading volume is of vital importance to algorithmic trading and portfolio management since it attempts to minimize transaction costs by optimally scheduling and placing. The purpose of this project is to create dynamic statistical models of intraday trading volume prediction (in Python). By assuming the stable U shape distribution of intraday trading volume, we apply Deterministic blend, Lognormal Bayesian, Kalman filter and ARIMA model to estimate and generate out of sample forecast on 12 US equity sector ETFs. Results show that some of the proposed methods are able to obviously outperform common volume forecasting methods. - View it on GitHub