Results

Results#

The results from POLARIS simulations are stored in a number of locations:

  1. Synthesized populations and their resulting travel demand (trips) are stored back into the <ModelName>-Demand.sqlite database. This database is well documented in the Demand Database section.

  2. Link and turn based attributes (such as volume, delay or speed) are stored in compressed binary format in a HDF5 datastore called <ModelName>-Result.h5. These are organized into sub-groups called link_moe and turn_moe respectively.

  3. Path sequences are also stored in the same h5 file, <ModelName>-Result.h5, in a group called paths

  4. Some higher level statistics (TNC servicing & repositioning, parking occupancy) will be saved into the Result SQLite database <ModelName>-Result.sqlite

In addition to the above low-level outputs, the standard convergence run process also builds a suite of KPIs for each iteration which are stored for quick access in a SQLite database called kpi.cache. These are generally just pandas DataFrames that have been serialized directly into SQLite (using a package called diskcache) and are very fast to retrieve later. These are discussed in more detail (including how to compare them across iterations or across models) in the KPI analysis section.

In the following pages we cover the low level file formats used to store results in more detail.