Research On The Urban Road Travel Time Prediction Model Based On Data Fusion

Road travel time is an important indicator of road traffic conditions. On the one hand, in real-life situations, it makes road network with time-varying traffic flow on spatial and temporal characteristicsdue, because traffic demand varies greatly of the day.Resulting in road sections running time is largely dependent on the traffic load changes. Therefore accurate prediction of road travel time is the basis of traffic guidance systems, traffic information service systems and traffic control systems coordination. On the other hand, due to the high complexity of road traffic flow, randomness and uncertainty, the traditional forecasting methods of road travel time have not been possible to obtain satisfactory prediction, such as based on the detection coil or based on GPS floating car, which affects the way for traffic control and road traffic-induced effects to a certain extent. Given their complementarity of advantages and disadvantages of the above prediction method, based on the research results of multi-source data fusion forecasting method of road travel time, this paper uses the improved BP neural network fusion of these two forecasting methods in order to improve the accuracy of the forecasts of road travel time.First, the obtained road traffic parameters data for fixed detector and GPS floating car have lost or error, so this paper proposes fault data recovery optimization based on the data of adjacent law of averageson. The improved method can better fix the fault data, thereby improving the quality of the data.Secondly, multi-vehicle GPS floating car data to predict the course of the road travel time has the problem that floating car are different in the length of passing on the road. This paper proposes an adaptive weighting factor line weighting fusion method for multi-vehicle GPS floating car data to predict road travel time.Finally, it is not accurate to predict road travel time by a single type of detector data. Based on improved BP neural network for road travel time data captured two detector fusion, this paper proposes road travel time prediction model based on data fusion for multi-source data to predict road travel time. Through the simulation of the above results, it is a preliminary study to achieve the desired goal.Urban road travel time prediction problem is relatively complex. Given the limited research capacity of the paper, there are many research to be further improved and continue to improve the follow-up study.