A distributed incremental aggregation method combined with traffic flow data cleansing rules is proposed, and it can provide more accurate and reliable data for traffic flow forecast analysis. Through the correlation analysis of traffic flow in road network, the authors used the multi-allocation of turning rate in the intersection to build the spatial weight matrix, and improved the STARIMA traffic flow forecasting model. The experiment result proves that this method can meet the needs of traffic flow big-data forecasting in the efficiency and accuracy, and provide the basis for the traffic routing information.