Kairos

Kairos is a module responsible for forecasting road traffic and transport demand (so-called displacements or travel arrangements). In our solution, we use several sources of information, including the current decisions made by fleet management systems and models of making these decisions. Accurate reflection of displacements in the analyzed or managed area is a key point in creating a simulation model and traffic forecasts. Forecasts prepared in this way allow for optimal use of fleets and reduction of operating costs.

As part of this module, we have solutions for the implementation and comparison of any Machine-Learning (ML) models in order to achieve high precision of prediction. We have base models as well as advanced road network forecasting models, including the ARIMA model used as the base model and models based on artificial neural networks, including Stacked Auto Encoder (SAE) models and models based on graphs: Temporal Graph Convolutional Network ( TGCN), Adaptive Graph Convolutional Recurrent Network (AGCRN). The results of training and validation of each model are monitored using tools that allow, among others, to track model convergence and compare metric values.