Urban Rail Transit Passenger Flow Forecasting Method Based on the Coupling of Artificial Fish Swarm and Improved Particle Swarm Optimization Algorithms

Yuan Yuan, Chunfu Shao, Zhichao Cao, Wenxin Chen, Anteng Yin, Hao Yue, Binglei XieSustainability 2019 · DOI

TL;DR

This paper introduces a novel hybrid algorithm, AFSA-PSO, that combines Artificial Fish Swarm (AFSA) and Particle Swarm Optimization (PSO) to improve the accuracy and efficiency of urban rail transit passenger flow forecasting. The AFSA-PSO algorithm demonstrates superior performance over individual PSO and AFSA methods, particularly in predicting passenger flow during peak hours and for large-scale tourist attractions, offering a more robust solution for transportation management.

What Problem Does This Paper Address?

Accurate forecasting of urban rail transit passenger flow is crucial for efficient station design, passenger management, and train operation planning. Existing methods often struggle with the inherent complexity and uncertainty of passenger behavior, especially during peak demand periods or when influenced by large events or tourist attractions. Traditional mathematical models can be too rigid to capture these dynamics, while some model-free algorithms face challenges like premature convergence to local optima or slow convergence speeds. This paper aims to address these limitations by developing a more robust and efficient forecasting algorithm.

Methods

The core contribution of this paper is the development of the Artificial Fish Swarm and Improved Particle Swarm Optimization (AFSA-PSO) algorithm. This hybrid approach leverages the strengths of both AFSA and PSO:

  • Particle Swarm Optimization (PSO): A well-established metaheuristic optimization algorithm that simulates the social behavior of bird flocking or fish schooling. It's known for its rapid convergence but can sometimes get stuck in local optima.
  • Artificial Fish Swarm Algorithm (AFSA): Another bionic algorithm inspired by fish foraging behavior. AFSA is noted for its ability to overcome local extrema and find global optima, and it doesn't require gradient information, making it adaptable to various search spaces.

The AFSA-PSO algorithm combines these by:

  • Dynamic Weighting: The algorithm uses dynamic weights to balance the influence of AFSA and PSO during the iteration process. In the early stages, AFSA's global search capability is emphasized, while in later stages, PSO's faster convergence is prioritized. This is achieved through a mutation rule that blends particles from both algorithms with weights that change based on the iteration number.
  • Particle Self-Regulation: The algorithm incorporates a mechanism for particles and fish swarms to self-regulate. This involves adjusting their positions based on local and global optimal values (for PSO) or the rate of change in fitness (for AFSA) to prevent premature convergence and improve exploration.

Datasets

The study utilizes historical passenger flow data from the Window of the World station on Shenzhen Metro Line 1 for its case study. This dataset includes:

  • Passenger flow data for normal working days (used for training and testing).
  • Passenger flow data for periods associated with large-scale tourist attractions (specifically, the evening peak hours from 5:00 to 7:00 p.m. on holidays).

The paper does not introduce new benchmark datasets but applies its algorithm to a real-world urban rail transit scenario.

Key Insights and Ideas

  • Synergy of Bionic Algorithms: Combining AFSA and PSO is more effective than using either algorithm alone. AFSA's global search and robustness complement PSO's speed, creating a more powerful forecasting tool.
  • Dynamic Weighting is Key: The dynamic adjustment of weights between AFSA and PSO is crucial for balancing exploration and exploitation throughout the optimization process, leading to better overall performance.
  • Addressing "Premature Convergence": The self-regulation mechanism and dynamic weighting effectively mitigate the common issue of premature convergence in swarm intelligence algorithms, allowing for more thorough exploration of the solution space.
  • Applicability Beyond Normal Peaks: The AFSA-PSO algorithm shows significant promise in predicting passenger flow during unusual demand periods, such as when large numbers of tourists leave a popular attraction, a scenario often overlooked by simpler models.

Key Results

The AFSA-PSO algorithm demonstrated superior performance compared to standalone PSO and AFSA across several metrics:

  • Prediction Accuracy:
    • Morning Peak: AFSA-PSO achieved an average Mean Absolute Percentage Error (MAPE) of 3.36%, significantly lower than PSO (4.13%) and AFSA (4.57%).
  • Evening Peak: AFSA-PSO achieved an average MAPE of 4.36%, outperforming PSO (7.14%) and AFSA (7.26%).
  • Large Tourist Attractions (Leaving Flow): AFSA-PSO achieved a MAPE of 4.47%, compared to PSO (6.87%) and AFSA (6.56%).
  • Convergence Speed: AFSA-PSO exhibited the fastest convergence speed and highest operational efficiency, reaching a stable MAPE in fewer iterations than PSO.
  • Correlation: The Pearson correlation coefficient between predicted and real values was consistently higher for AFSA-PSO, reaching an average of 0.998 for the morning peak and 0.990 for the evening peak, indicating a strong linear relationship.
  • Robustness: The algorithm showed strong robustness in predicting both normal passenger flow and the more complex scenario of large tourist attraction departure flow.

Limitations and Future Work

  • Limitations:
    • The study focused on a single station (Window of the World) and a specific metro line (Line 1). While the results are promising, generalizability to other stations or entire metro networks might require further validation.
  • The paper mentions that the prediction effect for the early peak was better than the late peak, suggesting potential variations in performance across different time segments.
  • Future Work:
    • The authors propose extending the forecasting to encompass all lines of the station collectively, rather than individual lines, to facilitate better overall passenger flow management and evacuation.
  • They also suggest incorporating the principle of entropy into the dynamic weight mechanism to further enhance prediction accuracy.

Summary updated

Abstract

Urban rail transit passenger flow forecasting is an important basis for station design, passenger flow organization, and train operation plan optimization. In this work, we combined the artificial fish swarm and improved particle swarm optimization (AFSA-PSO) algorithms. Taking the Window of the World station of the Shenzhen Metro Line 1 as an example, subway passenger flow prediction research was carried out. The AFSA-PSO algorithm successfully preserved the fast convergence and strong traceability of the original algorithm through particle self-adjustment and dynamic weights, and it effectively overcame its shortcomings, such as the tendency to fall into local optimum and lower convergence speed. In addition to accurately predicting normal passenger flow, the algorithm can also effectively identify and predict the large-scale tourist attractions passenger flow as it has strong applicability and robustness. Compared with single PSO or AFSA algorithms, the new algorithm has better prediction effects, such as faster convergence, lower average absolute percentage error, and a higher correlation coefficient with real values.
BibTeX
@inproceedings{yuan2019urban,
  author = {Yuan Yuan and Chunfu Shao and Zhichao Cao and Wenxin Chen and Anteng Yin and Hao Yue and Binglei Xie},
  title = {{Urban Rail Transit Passenger Flow Forecasting Method Based on the Coupling of Artificial Fish Swarm and Improved Particle Swarm Optimization Algorithms}},
  booktitle = {Sustainability},
  year = {2019},
  doi = {10.3390/su11247230},
}
Figure