ahp-mcda.md

Multi-Criteria Decision Analysis (MCDA) & AHP in Scenario Modeller

The Scenario Modeller library is designed to natively support MCDA and AHP (Analytic Hierarchy Process) through its Prioritise stage. This guide explains how to leverage the built-in multi-criteria plugin to rank any spatial unit (buildings, parcels, H3 cells) based on weighted stakeholder preferences.

1. Core Concept: Ranking vs. Filtering

In a simulation, MCDA typically operates at the Prioritise stage. While Filter determines who is eligible (binary), Prioritise determines the order of merit (ranking).

• Analytic Hierarchy Process (AHP): A method to derive weights from pairwise comparisons of criteria.
• Weighted Sum Model (WSM): The actual math applied to each entity using those weights to calculate a "Priority Score."

2. Using the Multi-Criteria Plugin

The library provides a specialized plugin createMultiCriteriaPrioritiserPlugin designed for generality. It does not assume specific field names, making it compatible with any data schema.

Basic Implementation

import { createMultiCriteriaPrioritiserPlugin } from 'interactive-scenario-modeller/plugins/optimization';

const prioritiser = createMultiCriteriaPrioritiserPlugin({
  criteria: [
    { 
      name: 'carbon', 
      weight: 0.5, // Derived from AHP
      getValue: (e) => e.carbonPotential,
      direction: 1 // Higher is better
    },
    { 
      name: 'cost', 
      weight: 0.3, 
      getValue: (e) => e.estimatedCost,
      direction: -1 // Lower cost is better
    }
  ],
  // Optional: property name to store the final calculated priority score back on the entity
  scoreProperty: 'priority_score'
});

const intervention = new Intervention('AHP-Driven Rollout', {
  prioritise: prioritiser.prioritise,
  // ...
});

3. Dynamic AHP Weights

One of the library's strengths is the ability to adjust weights mid-simulation. By default, the plugin looks at context.state.mcdaWeights.

Year-by-Year Weight Steering

Scenario managers can shift priorities over time (e.g., focusing on early carbon wins, then shifting to equity later):

const intervention = new Intervention('Steered Rollout', {
  prioritise: prioritiser.prioritise,
  initYear: (year, context) => {
    if (year >= 2030) {
      // Shift priority to equity after 2030
      context.state.mcdaWeights = {
        carbon: 0.2,
        cost: 0.2,
        equity: 0.6
      };
    }
  }
});

4. Domain Examples

Urban Planning (Parcels)

Rank parcels for green infrastructure, balancing flood risk reduction, air quality improvement, and land cost.

• Criteria: floodMitigation (max), pm25Reduction (max), landValue (min).

Public Health (LSOAs)

Rank neighborhoods for health program rollout, balancing infection rates and accessibility.

• Criteria: prevalenceRate (max), distToClinic (min), avgAge (max).

Energy (H3 Cells)

Rank hexagonal cells for EV charging hubs based on proximity to main roads and existing grid headroom.

• Criteria: distanceToA_Road (min), availableCapacity (max), trafficFlow (max).

5. Technical Notes

• Generality: Always use functional getValue getters. Avoid hardcoding properties like entity.cost.
• Normalization: The standard Weighted Sum Model works best when values are on similar scales (e.g., 0-1 or 0-100). If your data units are drastically different (e.g., £Millions vs 0.1 Carbon), pre-normalize your data in the FacetAdapter or inside the getValue function.
• Performance: The plugin uses standard JavaScript sort(). For very large datasets (50k+ entities), ensure your getValue functions are optimized (e.g., O(1) property lookups).

---

This document reflects the implementation as of version 0.1.0.