Developing a model for targeted transit advertising using smart card data

Abstract

The great potential of exploiting Smart Card (SC) data in developing transit advertising models has been neglected. These valuable sets of data reveal comprehensive travel behaviour of passengers in the public transit network, which was not feasible before. SC data can reconstruct trips and activities of passengers. Also, advertising models moved forward toward targeted advertising techniques that effectively target groups of audiences with common interests. SC data provide the opportunity to develop a targeted advertising model in the public transit system. This thesis proposes a model for the targeted advertising in the public transit network using SC data; in other words, it addresses the gap between the targeted advertising techniques and public transit medium.This thesis develops a targeted advertising model in the public transit network. The model targets passengers by maximizing the number of passengers who will watch an advertisement relevant to their trip. The model is developed on location, time, and purpose types of passengers’ trips. The model differs with traditional transit advertising in considering not only time of the day and location of the advertisements but also passengers’ actual activities. The targeted advertising model determines location and time slots for desired advertisement types, and then public transit authorities can sell location and time slots in the network to interested marketing companies. Outcomes of the targeted advertising model can turn the public transit network into a dynamic and competitive medium for the marketing companies and gain more revenues for the authorities. Main steps of the methodology behind the developed model in this thesis are as follows. In simple words, the model discovers groups of passengers with similar activities and trips, and displays relevant advertisements to the passengers, imputed with the sociodemographic characteristics, according to their trip purpose type. Firstly, trips and activities of passengers are reconstructed from SC data. Secondly, similarity measures for the activity and trip of passengers are elaborated. Thirdly, groups of passengers with similar activities are first clustered and then further re-clustered into groups of passengers with similar activities and trips. After discovering the groups of passengers who travel together and perform their activities together, the targeted advertising problem in the public transit network is formulated in a linear optimisation framework. The (passenger-based) targeted advertising model maximises the number of passengers exposing to relevant advertisements during their trips and activities. In the end, a classification method is developed to enrich outcomes of the targeted advertising model with sociodemographic characteristics of the passengers.The model is implemented on one-day SC data from South East Queensland (SEQ). Also, components of the model are validated on Household Travel Survey (HTS) data from SEQ. Validation outcomes present that the developed model successfully meets the expected research objectives. 69% of passengers are exposed to relevant advertisement type at the right time and location during their trips and activities.