Impact of Real-time Data Filling Methods on Multi-source Data Fusion

The authers proposed three kinds of real-time data filling methods including time series,spatial correlation and history database methods to complement the missing data,which resulted from detection equipment failure and error data-elimination process.The data filling results based on those methods were applied respectively to an urban expressway multi-source data fusion model,in which,floating car data and remote traffic microwave sensor(RTMS) data were used,and the impact on data-fusion accuracy and the application priority of those data filling methods were analyzed.Resultsshow that the mean absolute percentage errors(MAPEs) of data fusion models based on time series and spatial correlation methods are both under 20%,and that the practicability of the proposed methods is verified.