Skip to content

PresentPopulationEstimation Configuration

To initialise and run the component two configs are used - general_config.ini and component’s config present_population_estimation.ini. In general_config.ini all paths to the data objects used by PresentPopulationEstimation component shall be specified. An example with the specified paths is shown below:

[Paths.Silver]
...
grid_data_silver = ${Paths:silver_dir}/grid
cell_connection_probabilities_data_silver = ${Paths:silver_dir}/cell_connection_probabilities_data_silver
event_data_silver_flagged = ${Paths:silver_dir}/mno_events_flagged
present_population_silver = ${Paths:silver_dir}/present_population
...

Below there is a description of the sub component’s config - present_population_estimation.ini.

Parameters are as follows:

Under [PresentPopulationEstimation] config section:

  • data_period_start - string, format should be “yyyy-MM-dd HH:mm:ss“ (e.g. 2023-01-08 12:30:00). Determines when the first time point is generated.
  • data_period_end - string, format should be “yyyy-MM-dd HH:mm:ss“ (e.g. 2023-01-08 12:30:00). No time points can be generated after this time. A time point can be generated at this exact time.
  • time_point_gap_s - integer, in seconds. Determines the interval between two time points. Starting from data_period_start, one time point is generated after each time_point_gap_s seconds until data_period_end is reached.
  • tolerance_period_s - integer, in seconds. Determines the size of the temporal window of each time point. Only events within this distance from the time point are included in the results calculation of that point.
  • max_iterations - integer. Maximum number of iteration allowed for the Bayesian process for each time point.
  • min_difference_threshold - float. Minumum difference between Bayesian process prior and posterior population estimates needed to continue iterating the process.

Configuration example: grid-level aggregation

[Logging]
level = DEBUG

[Spark]
session_name = PresentPopulationEstimationGrid

[PresentPopulationEstimation]
data_period_start = 2023-01-01 00:00:00 # Starting bound when to create time points. The first time point is created at this timestamp. 
data_period_end = 2023-01-02 00:00:00 # Ending bound when to create time points. No time points are generated later than this timestamp. A time point can happen to be generated on this timestamp, but this is not always the case.
time_point_gap_s = 43200 # space between consecutive time points
tolerance_period_s = 3600 # Maximum allowed time difference for an event to be included in a time point
nr_of_user_id_partitions = 128 # Total number of user_id_modulo partitions. TODO should be a global conf value
nr_of_user_id_partitions_per_slice = 32 # Number of user_id_modulo partitions to process at one time
max_iterations = 20 # Number of iterations allowed for the Bayesian process
min_difference_threshold = 10000 # Minimum total difference between Bayesian process prior and posterior needed to continue processing