그래프 컨볼루션 네트워크 모델을 활용한 역별 공간 특성을 고려한 자전거 수요 예측
Predicting Bicycle Demand at the Station Level by Incorporating Station-Specific Spatial Features in Graph Coutional Network Model
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study proposes an advanced Graph Convolutional Network (GCN) model to predict daily bicycle rental volumes at the station level, addressing imbalanced usage patterns in shared bicycle services. Bicycle stations are modeled as nodes and user interactions as edges, with spatial (e.g., location, infrastructure) and temporal (e.g., weather, time) features incorporated. Optimized through hyper-parameter tuning, the model surpasses traditional methods and GCN model without spatial features, effectively capturing station-level interactions and network structures. It enables precise demand predictions, dynamic bicycle redistribution, and improved operational efficiency, supporting data-driven urban transportation policies and enhanced user convenience. The model’s applicability extends to domains requiring network-based forecasting and resource allocation.
Abstract
This study proposes an advanced Graph Convolutional Network (GCN) model to predict daily bicycle rental volumes at the station level, addressing imbalanced usage patterns in shared bicycle services. Bicycle stations are modeled as nodes and user interactions as edges, with spatial (e.g., location, infrastructure) and temporal (e.g., weather, time) features incorporated. Optimized through hyper-parameter tuning, the model surpasses traditional methods and GCN model without spatial features, effectively capturing station-level interactions and network structures. It enables precise demand predictions, dynamic bicycle redistribution, and improved operational efficiency, supporting data-driven urban transportation policies and enhanced user convenience. The model’s applicability extends to domains requiring network-based forecasting and resource allocation.