Predict Example

This example demonstrates how to:

  • Read in points using Tracktable’s C++ command line factories

  • Assemble points into trajectories

  • Use boost program options to take parameters from command lines (in addition to the factories)

  • Conditioning trajectories based on length and object id

  • Use boost R-tree to locate similar trajectories based on cartesian distance in feature space

  • Output the predicted origin/destination/future of the trajectory

The full predict example source code can be found in the Tracktable source code distribution in the directory tracktable/Examples and the example can be executed by calling the predict program from the command line provided that the example is built and it’s location is exposed to the appropriate system path. Below you will find the execution command for the example as well as the source files in there entirety for quick reference.

Example Source Files

The page listed here contains a direct import of the source code for this example. This is provided for convenience and reference.

Command Line Interface

Note

This command is specific to Linux and Mac. Windows machines will have a different command line call.

The command to run the predict example is as follows.

Typical Command
$ ./predict --input=/data/SampleASDI.csv --delimeter=, --string-field=dest 30 --num-samples=10

Important

--string-field is required.

This command takes an input parameter of a tab separated value or comma seperated value file formatted as OBJECTID TIMESTAMP LON LAT of points to assemble, a delimeter parmater for the file delimiter, a destination string field parameter for the number of destinations, and a sample parameter for the number of samples to take.

Note

The default delimiter is tab, if you are using a CSV file you will need to set the --delimiter parameter. The default output is standard out.

The command line interface contains a --help option that will display all of the possible switches for the example.

$ ./predict --help

will display:

--help                            Print help
--num-samples arg (=10)           Number of samples

Point Reader:
 --input arg (=-)                  Filename for input (use '-' for standard
                                     input)
 --real-field arg                  Field name and column number for a
                                     real-valued point field
 --string-field arg                Field name and column number for a string
                                     point field
 --timestamp-field arg             Field name and column number for a
                                     timestamp point field
 --object-id-column arg (=0)       Column containing object ID for points
 --timestamp-column arg (=1)       Column containing timestamp for points
 --x-column arg (=2)               Column containing X / longitude coordinate
 --y-column arg (=3)               Column containing Y / latitude coordinate
 --delimiter arg (=    )           Delimiter for fields in input file

Assembler:
 --separation-distance arg (=100)  Set maximum separation distance for
                                     trajectory points
 --separation-seconds arg (=1200)  Set maximum separation time (in seconds)
                                     for trajectory points
 --min-points arg (=10)            Trajectories shorter than this will be
                                     discarded
 --clean-up-interval arg (=10000)  Number of points between cleanup