How do you model travel behavior with uncertainty?

1 Define the sources of uncertainty

Uncertainty can arise from different sources in travel behavior modeling, such as data quality, model specification, parameter estimation, model validation, scenario analysis, and model application. Each source can introduce errors, biases, or variations that affect the accuracy and reliability of the model outputs. Therefore, it is important to identify and define the sources of uncertainty in your model, and to assess their magnitude and direction. For example, you can use sensitivity analysis, error analysis, or confidence intervals to measure the impact of data errors, model assumptions, or parameter estimates on the model results.

2 Use probabilistic or stochastic models

Probabilistic or stochastic models are types of models that incorporate randomness or variability in the model inputs, outputs, or processes. They can capture the uncertainty and heterogeneity in travel behavior by using probability distributions, random variables, or simulation techniques. For example, you can use discrete choice models, mixed logit models, or agent-based models to represent the probabilistic nature of travel choices, the heterogeneity of travelers' preferences, or the interactions of travelers and the transportation system. Probabilistic or stochastic models can provide more realistic and flexible representations of travel behavior, and can generate a range of possible outcomes and their likelihoods.

3 Incorporate behavioral theories and insights

Behavioral theories and insights are concepts and findings from psychology, sociology, economics, or other disciplines that explain how people make decisions and behave in different situations. They can help you understand and model the uncertainty and complexity in travel behavior by accounting for factors such as attitudes, beliefs, values, norms, habits, emotions, biases, heuristics, or learning. For example, you can use behavioral economics, prospect theory, or social network analysis to model how travelers perceive and evaluate risks, losses, or gains, how they are influenced by social norms or peer effects, or how they adapt or change their behavior over time.

4 Update and calibrate your model regularly

Updating and calibrating your model regularly is a good practice to reduce the uncertainty and improve the performance of your model. Updating your model means collecting and incorporating new or updated data, information, or knowledge into your model, such as travel surveys, traffic counts, census data, policy changes, or technology trends. Calibrating your model means adjusting or optimizing your model parameters, coefficients, or functions to fit the observed or expected data or behavior. For example, you can use calibration methods, such as maximum likelihood estimation, least squares estimation, or genetic algorithms, to find the optimal values of your model parameters that minimize the error or maximize the likelihood of your model.

5 Communicate and visualize your model results clearly

Communicating and visualizing your model results clearly is essential to convey the uncertainty and the implications of your model to your stakeholders, clients, or decision-makers. Communicating your model results means explaining and interpreting your model outputs, assumptions, limitations, and uncertainties in a clear, concise, and relevant way, using appropriate language, terminology, and formats. Visualizing your model results means presenting and displaying your model outputs, scenarios, or alternatives in a clear, attractive, and informative way, using appropriate charts, graphs, maps, or tables. For example, you can use communication tools, such as reports, presentations, or webinars, to summarize and highlight your model findings, recommendations, or trade-offs. You can use visualization tools, such as dashboards, animations, or interactive maps, to illustrate and compare your model outcomes, probabilities, or uncertainties.