Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

A Time Series Approach to Removing Outlying Data Points from Bluetooth Vehicle Speed Data

Roth, Jennifer M.

Abstract Details

2010, Master of Science in Engineering, University of Akron, Civil Engineering.
The idea of using Bluetooth technology to obtain vehicle speed data is currently in the early stages of development. Recently, much work has been done to refine the data collection methodology and to verify its validity. However, little has been done to develop a data cleaning methodology for the Bluetooth vehicle speed data beyond visual inspection of the data set. A data cleaning methodology was developed in this study based on a time series approach. The collected Bluetooth vehicle speed data were modeled using an ARIMA model to determine what speeds should be expected based on preceding data points. In order to meet the fundamental requirements of the time series modeling procedure, the data must be evenly spaced in time prior to modeling. While the nature of traffic data does not lend itself to being naturally evenly sampled, the data can be grouped in order to create an evenly sampled time series. In this study, several time groupings were considered and it was determined that using the median value of a 30 second time group produced the best initial model fit. The residual values were then evaluated using common statistical outlier detection tests, such as the Modified Z-Test, Grubbs’ Test and Chauvenet’s Criterion, to determine how much each data point differs from the expected value. The number of outliers detected by each method was compared, and it was determined that the Modified Z-Test detects the most outliers. Once all outliers were removed from the data set, the initial ARIMA model was applied in order to create a direct comparison between the original time series and the outlier-free time series. It was determined that using the Modified Z-Test to detect outliers provides the most improvements in model fit.
William Schneider, Dr. (Advisor)
178 p.

Recommended Citations

Citations

  • Roth, J. M. (2010). A Time Series Approach to Removing Outlying Data Points from Bluetooth Vehicle Speed Data [Master's thesis, University of Akron]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=akron1289758088

    APA Style (7th edition)

  • Roth, Jennifer. A Time Series Approach to Removing Outlying Data Points from Bluetooth Vehicle Speed Data. 2010. University of Akron, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=akron1289758088.

    MLA Style (8th edition)

  • Roth, Jennifer. "A Time Series Approach to Removing Outlying Data Points from Bluetooth Vehicle Speed Data." Master's thesis, University of Akron, 2010. http://rave.ohiolink.edu/etdc/view?acc_num=akron1289758088

    Chicago Manual of Style (17th edition)