This project refers to the scripts used in the paper "A New Methodology for Automatic Cluster-Based Kriging using K-Nearest Neighbour and Genetic Algorithms"

These instructions will get you a copy of the project up and running on your local machine for testing purposes.

It is advisable that the user download the RSTUDIO Desktop Version, which is FREE and OPEN-SOURCE LICENSED, available in the link: https://www.rstudio.com/

The version 1.2.931 was used in the development of the scripts.

The installation of RStudio software is simple and can be done with all default selections by pressing the Next Button. After the installation is complete, it is necessary to install the packages included in the scripts of the project.

To install packages, select the "Tools" option in the menu and "Install Packages".

In the text field "Packages", type each of the following packages listed below and press the button "Install".

Here are the packages that must be installed:

• gstat
• sp
• automap
• GA
• reshape
• NISTunits
• SearchTrees
• fpc
• outliers
• scales
• RGeostats (Must be manually Installed)

The package RGeostats (Version used was 11.2.3), can be downloaded in the link: http://rgeostats.free.fr/forum/viewtopic.php?f=2&t=48 or in this github repository. It must be installed manually, since is not avaiable in the CRAN Repository.

To install the package, in the same "Install Packages" window, choose to "install from" Package archive file, and select the .zip/.rar RGeostats file and press "install".

After all packages are installed, the databases must be prepared/downloaded.

1. Meuse -> Available in the "sp" package
2. Coalash -> Available in the "gstat" package
3. Broomsbarn -> http://www.kriging.com/datasets/ (BroomsBarn.dat File) or in this repository
4. Wolfcamp -> http://www.kriging.com/datasets/ (Wolfcamp.dat File) or in this repository
5. Walkerlake -> Available in the "gstat" package

"Broomsbarn.dat" and "Wolfcamp.dat" must be included in the same path/directory of the script (GAClusterKriging.R) file

Ex:

• somePath/Files/GAClusterKriging.R
• somePath/Files/Broomsbarn.dat
• somePath/Files/WOLFCAMP.dat

After theses steps, you may run the code!

Open the script (GAClusterKriging.R) File in the RSTUDIO software.

BEFORE RUNNING ANY TESTS, MAKE SURE TO FOLLOW THESE STEPS:

• In the toolbar select "Session"-->"Set Working Directory"-->"To Source File Location";
• In the toolbar select "Session"-->"Clear Workspace";

To RUN the tests, first the database must be selected in the section bellow!

Type the name of the output file and the number of tests that will be applied to each cluster.

Select the size of the Genetic Algorithm Population and Number of Iterations. Next, select the number of neighbours considered in the KNN method. Note that the number is N-1, because the closest point, is the point itself. So if you want 3 neighbours, you must type 4. Finally, select the number of clusters of the K-Means Method.

Save the file and click on the "Source" Button to run the script.

The execution can be time demanding depending on the parameters and database.

The output text file will be created in the directory of the R script file.

At the header, informations about the selected parameters are shown.

Then, iterations for each number of clusters are listed.

Each column represent an algorithm, and every line represent a test or iteration. At the end, the mean and standart deviation are calculated. This is repeated from 1 to the selected number of clusters.

• Carlos Yasojima

[email protected]