Skip to content

Latest commit

 

History

History
36 lines (24 loc) · 2 KB

README.md

File metadata and controls

36 lines (24 loc) · 2 KB

CIM-2017

This is the R code associated to the article:

Rodríguez-Fernández, Víctor, Antonio Gonzalez-Pardo, and David Camacho. “Modelling Behaviour in UAV Operations Using Higher Order Double Chain Markov Models.” IEEE Computational Intelligence Magazine. In press (2017)

You can find a link to the article here: [https://doi.org/10.1109/MCI.2017.2742738]

Requirements

  1. Download R 3.4.1 and RStudio (a version of RStudio supporting R 3.4.1 is needed)
  2. Open the file CIM-2017.RProj with RStudio
  3. Type install.packages("packrat") to install package packrat, for managing dependencies
  4. Type packrat::restore() to load all the project dependencies from remote repositories. This process will take several minutes.
  5. Type install.packages("./MmgraphR_0.1.tar.gz",repos = NULL, type = "source") to install the package ``"MmgraphR" from a local source tarball.

Usage

The project structure has been created using the R package ProjectTemplate. For more details about ProjectTemplate, see http://projecttemplate.net

  • The input data can be found in the data folder
  • Every analysis script can be found in the folder src.
  • The output data from the scripts is stored in the output folder

The scripts found under the src folder are named with numerical prefixes (01-, 02-, ...) to indicate the proper order of execution in order to reproduce the results of the paper. Below are briefly described the contents of each script:

  • 1-Markovian_Comparison-InputData.R: Load the data from the data folder to be used in the rest of the scripts.
  • 2-Markovian_Comparison-DCMM: train and evaluate DCMMs using the march package.
  • 3-Markovian_Comparison-Rank_Aggregation: performs a Rank Aggregation process over a grid of evaluated DCMMs.
  • 4-Markovian_Comparison-CIM-All_in_one_Rank_Aggregation: Study of the predictability/interpretability importance in the rank aggregation.

Contact

For any questions about the use of the code please contact me by email or add a new issue in this repository.