DATA SIGNATURES FOR TRAFFIC DATA ANALYSIS

摘要

With the advent of the Information Age, interpretation of huge amounts of data always poses big problem to data analysts. Methodologies had been created to try to mine useful and, possibly, novel information from data sources of massive sizes. We use the concept of Data Signatures to accomplish these difficult tasks. Data sets are processed by combinations of algorithms with a common aim of creating smaller and more compact representation or summarization of the latent characteristics of the data sets. We investigate the effectiveness of Power Spectrum signatures to analyze traffic volume data. We show that this method is successful in data repre- sentation, detection of common patterns, determination of outliers, and realization of unexpected and novel information from the data set through the help of domain experts from the National Center for Transportation Studies (NCTS).