Deviation of peak hours for metro stations based on least square support vector machine

Area of Science:

• Transportation Engineering
• Urban Planning
• Data Science

Background:

• Station-level ridership is crucial for metro design, but peak times often differ between stations and the city.
• Current forecasting methods, using city peak as reference, lead to underestimations and potential disruptions at stations with differing peak demands.

Purpose of the Study:

• To investigate the factors influencing peak deviation coefficient (PDC) in metro ridership.
• To develop a more accurate station-level ridership forecasting method that accounts for differing peak times.

Main Methods:

• Introduced the peak deviation coefficient (PDC) as the ratio of station peak ridership to city peak ridership.
• Employed the least square support vector machine (LSSVM) model to analyze ridership determinants affecting PDC.
• Proposed and utilized a novel land-use function complementarity index to describe local commute land use relative to the network.

Main Results:

• The LSSVM model effectively revealed nonlinear relationships between land use and metro ridership temporal distribution.
• The proposed land-use function complementarity index demonstrated superior explanatory and predictive power for peak deviation compared to simple commute land-use ratios.
• The study successfully provided a method for fine-grained, station-level ridership forecasting, addressing peak deviation.

Conclusions:

• Metro ridership forecasting needs to incorporate the peak deviation phenomenon for accurate planning.
• The land-use function complementarity index is a valuable tool for understanding and predicting peak deviation.
• LSSVM is an effective technique for analyzing complex, nonlinear factors in metro ridership.