Description
Taxi services are a crucial mode of transportation in urban areas and their demand forecasting is an important challenge for intelligent transportation systems. Accurate taxi demand prediction can help transportation authorities allocate resources more effectively and mitigate issues like empty taxis and traffic congestion. Traditional forecasting methods often fail to capture the complex nonlinear spatiotemporal relationships in taxi demand data. Recent deep learning-based approaches have shown promise, but they often overlook the impact of data sparsity on model performance. In this work, we propose a spatiotemporal graph neural network model for predicting sparse taxi demand, leveraging pedestrian traffic data as an additional input feature. Our model utilizes convolutional layers and gated recurrent units to extract multi-view spatiotemporal features of taxi demand. By incorporating population density information, the model is able to improve prediction accuracy in the face of data sparsity. This framework represents a novel approach to addressing the challenging problem of forecasting sparse taxi demand in smart cities.Period | 24 Apr 2024 |
---|---|
Event title | Postgraduate Seminar Series |
Event type | Public Lecture |