steffanv/mcda4arcmap

An MCDA add-in for ArcMap.

MCDA4ArcMap

This add-in for the ArcMap geographic information system (GIS) offers multi-criteria decision analysis (MCDA) and visualization functions for vector data. The MCDA process is highly interactive and the results can be processed within ArcMap.

PLEASE NOTE THE DISCLAIMER BELOW AND THE STIPULATIONS UNDER THE LICENSE TAB.

Features

The add-in supports the following MCDA methods:

• Weighted Linear Combination (WLC)
• Ordered Weighted Averaging (OWA)
• Locally Weighted Linear Combination (LWLC)

The LWLC method includes the following neighborhood definitions:

• Queen
• Rook
• Distance
• K-Nearest Neighbors (KNN)
• Automatic (increases KNN until the result can be calculated)

Additional add-in features include:

• Maximum-score standardization procedure
• Score-range standardization procedure
• CSV export of the result table and parameters
• Classified and unclassed choropleth maps with diverging colour scheme
• Three modi to define the map rendering frequency

Limitations

• Only polygon geometry is supported
• Criterion values must be numeric
• Rook contiguity for large data sets (> 1,000 polygons) is slow (several minutes processing time)
• Sessions are not persistent, MCDA results are lost when MCDA4ArcMap is closed - use right-click | Data | Export to save map layer containing results

System Requirements

• Developed for ArcMap 10.1, ArcMap 10.2 or later and the .Net Framework 4.0 or later
• Download should appear as "ESRI AddIn File" - if ArcMap runs, this add-in will also work! Downloads can be found in the releases section

CREDITS

• The add-in (version 1.0) was developed by Steffan Voss, Institute for Geoinformatics, University of Muenster, Germany, during his research visit in the Department of Geography, Ryerson University, Toronto, Canada, from August 2012 to January 2013.

• The add-in is further developed by Steffan Voss.

• This project was initiated and supervised by Dr. Claus Rinner, Dept. of Geography, Ryerson University.

• Partial funding for Steffan's research from Dr. Rinner's NSERC Discovery Grant is gratefully acknowledged.

• An earlier version of the LWLC tool for vector data in ArcMap was developed by Brad Carter for his Master of Spatial Analysis (MSA) degree at Ryerson University.

• The project was moved from codeplex.

DISCLAIMER

The MCDA4ArcMap tool has garnered considerable interest from researchers and practitioners. While we are very pleased to see this, we need to remind all users that the tool is provided "as is" and that we cannot guarantee its fitness for a particular use. If you are using MCDA4ArcMap for a specific purpose, such as for a thesis or for real-world decision-making, it is your responsibility to verify that the MCDA algorithms are implemented in the way you expect and need them. As an open-source software project, MCDA4ArcMap allows you to do this directly in the source code. Please see the Apache License under the license tab above for the full legal details.

mcda4arcmap's Issues

allow point and line geometry for certain methods

WLC/ OWA will work with point or line geometry

new classification feature

Translating German text below to English:

Instead of quantiles by unit number, use a second variable to classify the first. For example, classify average household income by quintiles of population count as follows: (1) sort units (e.g. neighbourhoods, districts) by avg hhld income; (2) add units from sorted list keeping a cumulative total of the population variable until 20% of total population of study area is reached; (3) start second class and reiterate, etc.. The result is a classification that includes the "poorest quintile" of the population (instead of the set of admin units). The second variable that determines the quantiles has to be a raw-count variable.

fuer die quantile klassifikation koennte man eine zweite variable (e.g., total population per unit) waehlen, und die quantiles waeren nicht abgezaehlte raeumliche einheiten (e.g., neighbourhoods) sondern gleiche kumulative summen (z.b. fuenftel) von der zweiten variablen - d.h., units sortiert nach erster variable (criterion - e.g., household income) aber klassifiziert nach je ein 1/n der zweiten variable (e.g., population). damit koennte man dann sowas ablesen wie: wo leben die reichsten 20% aller torontoer haushalte? nicht: wo sind die 20% reichsten stadtviertel, was man normalerweise sehen wuerde.

introduce local OWA

based upon: Malczewski, J. and Liu, X. 2014 ‘Local ordered weighted averaging in GIS-based multicriteria analysis’ Annals of GIS (in press)

Saving Project Parameters: Weights or Decision Rules

It would be beneficial if within the project if a file could be created with the selected criteria and decision rule, OR selected criteria and weights. Therefore, using the "saved" parameters users can explore the differing decision rules or weights without having to reset the project parameters.

Introduce automated testing

Cover large parts of the code with tests.

extend rook and queen contiguity

second, third ... order

Spatial Decision Support

The ability to create graphics/charts/graphs within the MCDA4ArcMap package would assist in its abilities to behave as a Spatial Decision Support System and improve analytical and visualization capabilities.

The ability to create an output chart that ranks the WLC, OWA scores with each location, as opposed to exporting the data into excel for manipulation or creating a new shapefile with results and manipulating it in an edit session. ie. a chart with the ten highest WLC scores with corresponding spatial attribute (ie. neighbourhood name)

Create a processing extent option?

Dear Developer,

I have been experimenting with the MCDA4ArcMap tool with a fairly large dataset (>20,000 polygons) and this would be on the small size of any future processing I intend to do. It would be nice to see a use display as current processing extent option. This could be a check box in your config section. This would limit the analysis to the current display extent. Ideal for testing and exploring the weights before attempting to apply them to the entire dataset.

This option becomes significant when one is exploring large datasets and is interested in the current area before attempting to roll out the weighting to the full dataset. It would allow the various MCDA methods to execute significantly faster as they would only ever be processing a subset of a much larger dataset.

Hope this idea makes sense?

normalization with second variable

... very similar to ArcMap