United States Department of Transportation
[PDF-1.98 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report
-
Corporate Publisher:
-
Abstract:Traffic volume data is crucial in many applications, including transportation operation analysis, congestion management, accident prevention, etc. Yet an extensive capture of accurate volume information on a large-scale network can be difficult and costly. This research focuses on hourly traffic volume prediction in a statewide network using spatial-temporal features and heterogenous data sources. We present a classic machine learning technique - support vector machine (SVM) and compare its efficiency for traffic volume prediction with traditional estimation method. Further, the study develops an innovative spatial prediction method. The method is built off a state-of-the-art tree ensemble model - extreme gradient boosting tree (XGBoost) - to handle the large-scale features and hourly traffic volume samples. Moreover, spatial dependency among road segments is considered using graph theory. Specifically, we created a traffic network graph leveraging probe trajectory data and implemented a graph-based approach - breadth first search (BFS) - to search neighboring sites in this graph for computing spatial dependency. The proposed spatial dependency feature is subsequently incorporated as a new feature fed into XGBoost. The proposed methods are applied to 101 continuous count station (CCS) sites in the State of Utah. Prediction accuracy and training time are compared across the proposed models.
-
Format:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
