Classifying Travelers from Connected-Vehicle Data: Visitors or Commuters?

Abstract: Traditional travel behavior analyses using aggregated data, such as the data collected by traffic detectors, often fail to distinguish between visitors and commuters, which can lead to biases in decision-making and policy design for urban planning, transportation management, and economic development. This distinction is especially important in regions where tourism plays a significant economic role. Tourism accounts for over 10% of global GDP and supports one in ten jobs worldwide, highlighting the need to accurately understand tourists' travel behaviors to inform effective policies and operations.
With recent technological advancements, connected-vehicle data has gained attention for providing high-resolution vehicle location data, offering detailed insights into travelers' micro-movement patterns. However, connected-vehicle data also does not inherently distinguish between visitors and commuters. To address this limitation, we will introduce a deep-learning method designed to classify travelers from connected-vehicle data into visitors and commuters. This approach differentiates travelers based on distinct patterns such as visit frequency and duration at various locations.
We will apply this method to Hawaii, a popular tourist destination, to analyze and compare the travel behaviors of visitors and commuters, focusing on their destination preferences and travel patterns. The insights from this analysis will provide valuable guidance to local transportation agencies, helping them better understand and differentiate between these groups, ultimately improving transportation management, planning, and operations.