Bus ridership prediction

Abstract

During tenders in public transport, the bus network is reconsidered and adjusted. An accurate prediction of the effect on ridership is required. Due to the tender environment, time and input data are limited. This research focuses on deriving a relation between level-of-service (LOS) and ridership, and implementing this in a model suitable for the tender environment. A before-after study is performed based on two different levels of detail: OD-based and line-based. An OD-based analysis shows two issues for translation to model parameters: a poor fit of the trend line and a large share of a-typical data. A line-based approach provides more promising results. The range of elasticities and growth factors found is large, which is insufficiently acknowledged by current literature. Context is one of the main explanations for the large range. A regression analysis indicates a significant relation between LOS and ridership and indicates possible significance of additional predictors for ridership change. A comparison model is developed, which enables the user to match its request with an entry in the data base, providing the ridership development and context. The model is able to provide the user with growth factors for full day and peak/off-peak interaction and indicates sensitivity of the user groups. Future expansion of the model is recommended.