A Cooperative Data Mining Approach for Potential Urban Rail Transit Demand Using Probe Vehicle Trajectories

Promoting the use of public transportation is an important approach to develop sustainable mobility. However, lots of potential users of public transportation chose taxi, a semi-private mode for convenience. In this study, we first define this potential urban rail transit demand based on its spatiotemporal features. Then a novel data mining method is proposed to ascertain the potential urban rail transit demand from taxi trajectory data through considering spatial and temporal constraints simultaneously. Two features of the potential demand, i.e., the zero rates and volatility, are obtained by the combination of statistical and feature extraction (local neighbor descriptive pattern, LNDP) techniques. They are used to classify the urban rail transit stations into different categories which need different improvement measures to promote the attraction to the potential users. The effectiveness of the proposed method is tested using the GPS trajectory data of Shanghai collected from over 10,000 taxis in 12 consecutive days. We find that most urban rail transit stations have the potential to absorb the regular part of taxi ridership. Moreover, obvious imbalances exist between access and egress potential travel demands at these stations. The results show that metro stations can be classified into six groups according to the time-varying laws of potential travel demand, four of which need urgent measures. These findings provide useful insights for developing more effective and targeted strategies to encourage travelers to shift to public transportation. The estimated method of potential demand is the prerequisite for further optimization models.