Demand modeling and forecasting is one of our core competencies. We have worked in this space for over 20 years and have specific expertise in the modeling of transport facilities such as highways, toll roads, ports, transit systems, and railroads.
We utilize the traditional demand modeling techniques (time series analyses, econometric modeling, and four-step transport planning models) as well as artificial intelligence-based techniques such as neural network-based machine learning to forecast demand, prices of stocks, etc.
Neural network-based back-propagation algorithms can be used to forecast the demand for toll facilities. Only the past period data is used to do the forecasts. This obviously creates a few issues.
- First, it is difficult to describe what is impacting the changes in the data.
- Second, the process does not take into account the structural changes on the forecasted variable.
Given these constraints, the advantage of neural network-based models are clear – they are inherently nonlinear, fast, and perform quite well when compared to traditional econometric models.
We would caution against using these models for long-term forecasts because in the longer-term one should take into account structural changes in the transport network, the economy, and other social conditions.
Try our demand forecasting application.