How do you test the strength of your travel models?

1 Data sources

The first step in testing your travel models is to collect and analyze relevant data sources that can help you compare your model outputs with observed travel behavior. Depending on the type and scope of your model, you may need data on traffic counts, transit ridership, travel surveys, land use, socio-economic factors, and other variables that affect travel demand and supply. You can obtain data from various sources, such as government agencies, private vendors, online platforms, or your own surveys and sensors. However, you should also consider the quality, availability, and compatibility of the data you use, as well as the potential biases and errors that may affect them.

2 Validation methods

The next step is to apply different validation methods to assess how well your model matches the data. There are several types of validation methods, such as face validity, internal consistency, external validity, and sensitivity analysis. Face validity checks if your model results are reasonable and consistent with your expectations and assumptions. Internal consistency checks if your model components are coherent and logical within the model structure. External validity checks if your model results are consistent with independent data sources that are not used in the model development. Sensitivity analysis checks if your model responds appropriately to changes in input parameters and scenarios.

3 Calibration techniques

The final step is to adjust your model parameters and coefficients to improve the fit between your model and the data. This process is called calibration, and it involves using various techniques, such as trial and error, optimization algorithms, or statistical methods, to find the optimal values for your model variables. Calibration can be done at different levels of aggregation and disaggregation, depending on the level of detail and accuracy you need. For example, you can calibrate your model at the regional, sub-regional, or zonal level, or at the individual or household level.

4 Challenges and limitations

While testing the strength of your travel models is important and beneficial, you should also be aware of the challenges and limitations that may affect your model performance and validity. Some of these challenges include data scarcity, uncertainty, variability, and heterogeneity, as well as model complexity, instability, and non-linearity. Moreover, you should recognize that no model can perfectly represent the reality of travel behavior and choice, as there are always trade-offs between simplicity and realism, accuracy and precision, and generalization and specificity. Therefore, you should always use your models with caution and transparency, and acknowledge their assumptions and limitations.