Recognizing demand patterns from smart card data for agent-based micro-simulation of public transport

In public transportation the question of how to achieve a good match between demand and capacity is essential for operators to provide a high quality service level within reasonable costs. Agent-based micro-simulation is a promising method to evaluate the impact of operational decisions and selected tariffs at both the level of the individual passenger and the aggregate level of the operator. During recent years, this technique has been applied successfully to several large scale real life cases. However, the demand of the agent population in these simulations is usually derived from aggregated census data and surveys conducted among a relatively small sample of the travelers. With the advent of smart card ticketing systems new opportunities to generate an agent population have surfaced. We use a unique smart card dataset containing four months of individual mobility data from passengers among three modalities in an urban Dutch public transportation system to generate agent populations. We model the temporal flexibility of agents based on patterns observed in the check-in/check-out behavior of individual travelers. We then run simulations to study how these agent populations react to a discounted tariff in the off-peak hours. Finally, we discuss opportunities to improve our approach in the future.