Medium-term public transit route ridership forecasting: What, how and why? A case study in Lyon

Abstract

Demand forecasting is an essential task in many industries and the transportation sector is no exception. In fact, accurate prediction of future demand is an essential components of intelligent transportation systems [Vlahogianni et al., 2014, Koutsopoulos et al., 2019] and a fundamental aspect of any rationale planning process [Bonnel, 2002, Ortúzar and Willumsen, 2011]. Thus, it is not surprising that passenger demand forecasting is a widely studied subject. To summarize this area of research, we must take into account the domain of application and the type of planning issues the forecast intend to addressed [Hyndman and Athanasopoulos, 2018]. In many organisation including public transport, they are three commonly accepted level of planning and organisational control [van de Velde, 1999, Pelletier et al., 2011]. Strategic level deal with long-term decisions and objectives. Tactical level focus on decisions that take place in medium-term and aims to guarantee that the means to reach long-term goals are in place. Operational planning is concerned with short-term decisions that ensure the efficiency of the production. In agreement with this hierarchical order of decision-making activities, operators and transport agencies need to generate different forecasts. Short term forecasting is a very active field [Vlahogianni et al., 2014] that deal with models that predict demand from few minutes to few hours into the future [Vlahogianni et al., 2014, Noursalehi et al., 2018]. In the context of public transit (PT), authors argue that it can enable the design of better control strategies and improve passenger experience by proactively adjust services and customer information [Koutsopoulos et al., 2019, Noursalehi et al., 2018, Ma et al., 2014, Wei and Chen, 2012]. Long term ridership forecasting deals with models that predict demand for time horizon ranging from 5 to 15 years ahead. In contrast to short-term forecasting, it is often a onetime exercise and forecasts are rarely generated continuously. Forecasts are produced to assess and evaluate future scenarios and support long-range strategical planning. The most common method for long-term ridership forecasting traditionally relies on four-step travel demand model [Boyle,