MCDA
Multi-Criteria Decision Analysis, or MCDA, is a structured process for evaluating options with conflicting criteria and choosing the best solution. This method is useful when multiple criteria need to be considered simultaneously.
Additionally, it is employed for retrieving thematic layers from our database to display on Application UI. MCDA tool is available on our platform.
MCDA allows users to combine multiple layers on the map.
Before combining layers, the following steps are applied to each layer:
- Determine global minimum and maximum values.
- Select the best dataset transformations for optimal layer visualization.
- Normalize values, with the minimum set to 0 and the maximum to 1.
The results are visualized on the map, where hexagons with the lowest values (0) are shown in red, and those with the highest values (1) are shown in green.
Users can adjust each layer’s options, such as:
- Setting a custom value range.
- Setting range outliers behavior.
- Updating the values considered the worst and best for the analysis.
This is useful when the area of interest has values that deviate significantly from the global averages, or when you need to focus on a specific value range.
For example, if you want to find areas far from electric vehicle stations, but the global maximum distance is around 7,000 km, it makes sense to set the maximum to a more realistic 70 km for this case.
Changes hexes score behavior with layer values out of considerable range.
Possible values:
- Clamp
- Don’t modify
- Hide
For example, consider the distance to power lines in Haiti from OSM data, with the upper limit set at 3000 meters, and the sentiment set that 0 is good and the longer distance is worse.
Outliers: Clamp
Meaning: Values that exceed the limits contribute equally to the analysis, no matter how far they are from those limits.
Use Case: Useful for identifying people living more than 3,000 meters from power lines, without needing to account for the exact distance.
Outliers: Don’t modify
Meaning: Contribution to the analysis increases with the distance from the acceptable limit.
Use Case: Useful when analyzing locations for building a solar station, where distances beyond the limits are still considered but involve additional costs.
Outliers: Hide
Meaning: Hexagons with values beyond the limits are excluded from the analysis.
Use Case: Useful if non-electrified areas are not considered, for example, when opening a store.
Sentiments determine how the layer’s values impact the analysis:
- Bad → Good: Higher values are considered positive.
- Good → Bad: Higher values are considered negative.
Purpose:
This is useful when combining both negative and positive criteria, such as population density and the number of markets, to identify areas lacking food shops.
For example, you can set:
Food shops to Area (n/km²) bad (0) → good (21), n/km²
Population (ppl/km²) (ppl/km²) good (0) → bad (4242), ppl/km²
In this case, red hexagons will highlight areas with high population density and low food shop availability.
By default, all layers contribute equally to the analysis using a weighted average. You can increase the weight of a specific layer (e.g., 2x, 3x) to give it more importance in the analysis.
Also you can set 0 to prevent layer values infuence to analysis final score. If you adjust layer range and set outliers hide for excluding some areas from your analysis, but you don’t need to consider these layer values in the analysis.
Apply mathematical transformations to the values before normalization. These transformations help create a more linear distribution, providing clearer and more detailed insights for analysis.
Possible values:
- No transformation
- Square root: sign(x)⋅√|x|
- Cube root: ∛x
- log₁₀(x - xmin + 1)
- log₁₀(x - xmin + ε)
Population distribution shows many areas with low population and a sharp rise in cities.
Low-contrast visualizations like this one lose a lot of information and nuance. To recover this information and make the map more contrastive, GIS specialists employ mathematical transformations that make the distribution more linear-like. Here’s an example where we transform the distribution with Log(x):
This map presents much more information, making mountains, small towns, and city outskirts distinguishable. Each layer has its own transformation function determined by the nature of the distribution. Previously, users had to do this manually for each layer. Now, Atlas automatically chooses the best transformation, making it easier to get a great analysis.
All layers (criteria) can be rescaled to have a range in [0, 1], also known as rescale or min-max normalization. In case you want to get original values, set "no" value here.
Hexagons are colored based on color legend. Green = good place to open new food shop, red = not recommended. Interpretation of analysis is "let's assume that hexagons with fewer stores and higher population are more suitable for opening a new food shop".
Hexagons are clickable. To view calculation details, click on any hexagon, and pop-up will show layer values, weights, normalization and the final MCDA result.
MCDA is not saved between user logins. To save analysis for future use, click download button near MCDA name. All settings, sentiments, weights, etc. will be saved as single JSON file.
To continue working with previously exported MCDA, upload the file by clicking "Upload MCDA" button on the toolbar.
MCDA file structure describes details of exported JSON file.