2019-10-222019-10-22Oct-17FHWA-OK-17-02Oklahoma Department of Transportation State Planning and Research Item Number 2300https://hdl.handle.net/11244/321501This report presents research detailing the use of the first version of the National Performance Management Research Data Set (NPMRDS v.1) comprised of highway vehicle travel times used for computing performance measurements in the state of Oklahoma. Data extraction, preprocessing, and statistical analysis were performed on the dataset and acomprehensive study of dataset characteristics, influencing variables, outliersand anomalies was carried out. In addition, a study on filtering and removing speed data outliers across multiple road segments is developed, and a comparative analysis of raw baseline speed data and cleansed data is performed. A method for improved congestion detection is investigated and developed. Identification and a computational comparison analysis of travel time reliability performance metrics for both raw and cleansed datasets is shown. An outlier removal framework is formulated, and a cleansed and complete version of NPMRDS v.1 is generated. Finally, a validation analysis on the cleansed dataset is presented. In the end, research affirmsthat understanding domain specific characteristics is vital for filtering data outliers and anomalies of this dataset,which in turn is key for calculating accurate performance measurements. Thus, careful consideration for outlierremoval must be taken into account when computing travel time reliability metrics using the NPMRDS.118 pages5,002,161 bytesapplication.pdfTechnical ReportITSTraffic AnalyticsData AnalyticsMachine LearningTravel TimeNPMRDS