Fast and Memory Efficient Spatial Interaction Models

Building on top of the 'RcppArmadillo' linear algebra functionalities to do fast spatial interaction models in the context of urban analytics, geography, transport modelling. It uses the Newton root search algorithm to determine the optimal cost exponent and can run country level models with thousands of origins and destinations. It aims at implementing an easy approach based on matrices, that can originate from various routing and processing steps earlier in an workflow. Currently, the simplest form of production, destination and doubly constrained models are implemented. Schlosser et al. (2023) .

cppSim

This aims at providing a set of fast, memory efficient functions to perform spatial interaction modelling, also called gravity modelling. Currently, the doubly and singly constrained models are implemented for canonical set of constraints. Future versions will aim to implement more origin and destination constraints as well. It was developed in the context of studying commuter flows by active travel (cycling & walking ) in Great Britain as part of a project at CASA, UCL.

Installation

Not yet on CRAN, so please install the development version of cppSim with:

# install.packages(C("devtools","pak"))

devtools::install_github("ischlo/cppSim")
# pak::pak("ischlo/cppSim")

Built in data sets

The package comes with sample data sets that allow to test the functions right away as well as see the type of input that is recommended.

• flows_test : using the official census data in England from 2011, it’s a 983x983 matrix representing the flows of cyclists and pedestrians from each to each MSOA in London.
• distance_test : the distances between centroids of MSOAs. Computed with the London road network from OpenStreetMap and using the cppRouting package.

Spatial interaction models

Refer to the vignette to find some theory on SIMs and a naive implementation in R.

Example

Using the built-in data sets flows_test and distance_test, we can run a test by following the example This is a basic example which shows you how to solve a common problem:

library(cppSim)
## basic example code

data("flows_test")
data("distance_test")


model_test <- run_model(
  flows = flows_test,
  distance = distance_test
)

Source

For an example of what can be done with this package, please refer to the publication on active travel spatial interaction models in London for which it was originally developed.

Schlosser, I., Maureira, V.M., Milton, R., Arcaute, E., Batty, M., 2023. Active-travel modelling: a methodological approach to networks for walking and cycling commuting analysis.

The accompanying code for the analysis is provided in the ischlo/quant_cycle_walk repository.

Dependencies

This package has some dependencies that might need manual installation, although the most important external ones have been provided with the source code.

External

The package uses the armadillo library, which is imported and linked automatically when the package is installed.

Other

On the R side, it uses Rcpp [@eddelbuettel2011] and RcppArmadillo [@eddelbuettel2014].

Performance

Compared to the equivalent functions implemented in pure R, it runs about x10 faster for a $\sim 1000 \times 1000$ OD matrix, the speed up is increasingly more significant as matrices get bigger.

[1] "/Users/cenv1069/Library/R/arm64/4.4/library/cppSim/extdata/benchmark_test.rds"

Citation

To cite package 'cppSim' in publications use:

  Schlosser I (2023). _cppSim: Fast and Memory Efficient Spatial
  Interaction Models_. doi:10.5281/zenodo.14367756
  <https://doi.org/10.5281/zenodo.14367756>,
  <https://ischlo.github.io/cppSim/>.

A BibTeX entry for LaTeX users is

  @Manual{,
    title = {cppSim: Fast and Memory Efficient Spatial Interaction Models},
    author = {Ivann Schlosser},
    year = {2023},
    doi = {10.5281/zenodo.14367756},
    url = {https://ischlo.github.io/cppSim/},
  }