Calibration Methods
Routing Summary
This folder covers approaches to calibrating agent-based models. Contains 8 notes across two approaches: GA-based point estimation (Ben Said, Karakaya, Bonabeau) and uncertainty quantification-based calibration (McCulloch et al. 2022).
- Need an overview of ABM calibration challenges and approaches? → ABM Calibration Overview
- Need the GA-based calibration procedure? → Genetic Algorithm Calibration for ABM
- Need the fitness evaluation mechanism? → GA Fitness Evaluation and the RAM
- Need the combined HM+ABC framework for calibration with uncertainty quantification? → HM-ABC Calibration Framework
- Need History Matching methodology (implausibility score, waves)? → History Matching for ABMs
- Need ABC rejection sampling for ABMs? → Approximate Bayesian Computation for ABMs
- Need the three ABM uncertainty types (model discrepancy, ensemble variance, observation)? → Uncertainty Quantification for ABM Calibration
- Need worked examples on SugarScape, birds, and RISC? → ABM Calibration Case Studies
Concept Map
| Concept | Note | Type | Depends On | Key Result |
|---|---|---|---|---|
| Calibration overview | ABM Calibration Overview | concept | ABM Methodology and Principles | Three approaches: GA, controlled experiments, analytical baselines |
| GA calibration | Genetic Algorithm Calibration for ABM | concept | ABM Calibration Overview | 6-gene chromosome, roulette wheel selection, 85% crossover, 1% mutation |
| Fitness evaluation | GA Fitness Evaluation and the RAM | concept | Genetic Algorithm Calibration for ABM | Dual macro/micro evaluation comparing simulation to observed data |
| HM+ABC pipeline | HM-ABC Calibration Framework | overview | ABM Calibration Overview | 4-step pipeline: define space → quantify UQ → run HM → run ABC; 3,185 runs vs 11,000+ for ABC alone |
| History Matching | History Matching for ABMs | concept | HM-ABC Calibration Framework, Uncertainty Quantification for ABM Calibration | Implausibility ; Pukelsheim rule |
| ABC for ABMs | Approximate Bayesian Computation for ABMs | concept | History Matching for ABMs | Rejection sampling with HM-informed prior; |
| Uncertainty types | Uncertainty Quantification for ABM Calibration | concept | ABM Calibration Overview | 4 sources: parameter, model discrepancy, ensemble variance, observation uncertainty |
| Case studies | ABM Calibration Case Studies | example | HM-ABC Calibration Framework | SugarScape (2 params), birds (comparison to alternatives), RISC (16 binary model variants) |
Notes
- ABM Calibration Overview — CONTAINS: why ABM calibration is hard (5 challenges), three calibration approaches comparison, general calibration workflow
- Genetic Algorithm Calibration for ABM — CONTAINS: chromosome encoding (6 genes), GA procedure (elite selection, roulette wheel, arithmetic crossover, mutation), parameter table, convergence results
- GA Fitness Evaluation and the RAM — CONTAINS: RAM architecture, Category 1 (macro curves), Category 2 (micro probes), fitness computation, integration with GA loop
- HM-ABC Calibration Framework — CONTAINS: 4-step pipeline overview, efficiency comparison table, birds 3,185 vs 11,000+ runs result, accuracy trade-off (coverage vs precision)
- History Matching for ABMs — CONTAINS: implausibility score (Eq. 1), model discrepancy variance (Eq. 2), ensemble variance (Eqs. 3–4), wave structure, LHS sampling, stopping criteria, SugarScape 10-wave example
- Approximate Bayesian Computation for ABMs — CONTAINS: rejection sampling algorithm, HM-informed prior, threshold setting (Eq. 5), posterior refinement, accuracy table vs ABC alone
- Uncertainty Quantification for ABM Calibration — CONTAINS: 4 uncertainty sources (parameter uncertainty, model discrepancy , ensemble variance , observation uncertainty ), total uncertainty budget, why traditional calibration fails
- ABM Calibration Case Studies — CONTAINS: SugarScape step-by-step, birds model (3 fitting criteria, 3 waves, efficiency comparison vs simulated annealing/EA/ABC), RISC Scottish cattle farms (16 binary model variants, MASE error metric, POM application)
Sources
- abm_human_behaviour.pdf — GA calibration and RAM (Ben Said et al. 2002)
- abm_consumer.pdf — Experimental calibration approach (Karakaya et al. 2011)
- abm_word_of_mouth.pdf — Analytical baseline comparison (Bonabeau 2002)
- calibration_ABM.pdf — HM+ABC framework (McCulloch et al. 2022, JASSS 25(2))