Passenger Flow Prediction for New Line Using Region Dividing and Fuzzy Boundary Processing

• Published in: IEEE transactions on fuzzy systems Vol. 27; no. 5; pp. 994 - 1007
• Main Authors: Yu, Hai-Tao, Jiang, Chang-Jun, Xiao, Ran-Dong, Liu, Hang-Ou, Lv, Weifeng
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.05.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Airports; Buildings; Clustering algorithms; Fuzzy boundary processing; Mathematical models; Neural networks; passenger flow prediction; Passengers; Prediction algorithms; Prediction models; Predictive models; Public transportation; region dividing; Spatial data; Transportation networks; Urban areas
• ISSN: 1063-6706; 1941-0034
• DOI: 10.1109/TFUZZ.2018.2825950

Predicting the passenger flow of public transport in a newly developed area of a city is very urgent for designing a precise and efficient public transport network. This paper proposes a new prediction model by exploring the relationship between the passenger flow of a station and its surrounding area's factors. First, in order to obtain more accurate factors affecting the passenger flow, the city is divided into multiple regions with similar internal traffic properties and moderate spatial size using the data of urban road network and buildings. Second, to effectively solve the problem of fuzziness of the station's attraction scope, the concept of the membership degree and fuzzy processing method is proposed. Finally, the station's passenger flow prediction model is launched based on Xgboost. The experimental results on three districts in Beijing show that our method outperforms all baselines significantly, which improves the accuracy by more than 20%.