Optimisation routines used for demand management in transport systems strongly depend on accurate forecasts. Outliers caused by systematic shifts in demand cause erroneous forecasts for both current services and future services whose forecasts are based on historic demand. Transport service providers often rely on analysts to identify outlier demand and make adjustments accordingly. However, previous research on judgemental forecasting shows that such adjustments can be biased and even superfluous. Literature on automated detection and evaluation of outlier demand in this context is scarce.

To date, most literature on forecasting and optimisation in transport planning does not account for demand outliers despite the negative impacts it can have. This thesis presents a novel methodology, which combines network clustering with functional data analysis and time series forecasting, to detect outliers in demand for transport systems. This thesis also contributes a simulation framework for evaluating the performance of the proposed outlier detection procedure and for quantifying the effects of outlier demand on different optimisation routines. The use of such a method as a decision support tool for analyst adjusted forecasts, and how the outlier alerts may be best communicated, is also considered. Computational studies highlight the benefits of different adjustments that analysts may take after the identification of outlier demand. Multiple empirical studies will demonstrate how the method can be applied in practice to different types of transport systems, with analyses of Deutsche Bahn railway booking data and Capital Bikeshare usage data.