Brief overview of steps for tackling modelling process.
- What precisely is our problem and what do we want the model to do?
- Do we need a model at all? Perhaps it is too simple a task and we would sooner not waste time.
- Are similar models available? Can we buy one or use a pre-existing in-house one? Danger is discovering too late that some of its features were note accurate enough for current purpose.
- Choosing granularity and scope of models scale.
- What constitutes key part of system we are interested in?
- What constitutes basic time quanta of simulation (seconds, years, centuries) and how far in future do we want to predict behaviour we are investigating?
- Determine abstraction depth we want to model.
- Boundaries:
- How many aspects of reality should we include?
- How detailed will their description be?
- Also start with simple models and gradually add new features.
- Without simplicity, get stuck in too much data.
- Too simple a model loses realism and may miss peaks/troughs in performance.
- What are key processes and parameters of model?
- Choose correct modelling tools.
- Verification:
- Check if model behaves as expected and does what it is supposed to do.
- Validation:
- Check if model behaves realistically,
- For stochastic models, conduct statistical analysis of results.