Showing 1–1 of 1 results for author: Cismaru, A T
-
Improving Traffic Flow Predictions with SGCN-LSTM: A Hybrid Model for Spatial and Temporal Dependencies
Authors:
Alexandru T. Cismaru
Abstract:
Large amounts of traffic can lead to negative effects such as increased car accidents, air pollution, and significant time wasted. Understanding traffic speeds on any given road segment can be highly beneficial for traffic management strategists seeking to reduce congestion. While recent studies have primarily focused on modeling spatial dependencies by using graph convolutional networks (GCNs) ov…
▽ More
Large amounts of traffic can lead to negative effects such as increased car accidents, air pollution, and significant time wasted. Understanding traffic speeds on any given road segment can be highly beneficial for traffic management strategists seeking to reduce congestion. While recent studies have primarily focused on modeling spatial dependencies by using graph convolutional networks (GCNs) over fixed weighted graphs, the relationships between nodes are often more complex, with edges that interact dynamically. This paper addresses both the temporal patterns in traffic data and the intricate spatial dependencies by introducing the Signal-Enhanced Graph Convolutional Network Long Short Term Memory (SGCN-LSTM) model for predicting traffic speeds across road networks. Extensive experiments on the PEMS-BAY road network traffic dataset demonstrate the SGCN-LSTM model's effectiveness, yielding significant improvements in Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) compared to benchmark models on the same dataset.
△ Less
Submitted 31 October, 2024;
originally announced November 2024.
