Novel Machine Learning Methods for Accident Data Analysis

Details

• Creators:
  Lin, Lei ; Wang, Qian ; Sadek, Adel W.
• Corporate Creators:
  State University of New York at Buffalo. Dept. of Civil, Structural, and Environmental Engineering
• Corporate Contributors:
  Transportation Informatics University Transportation Center ; United States. Department of Transportation. University Transportation Centers (UTC) Program
• Subject/TRT Terms:
  • Algorithms
  • Crash analysis
  • Crash data
  • Crash risk forecasting
  • Data mining
  • Machine learning
  • Traffic crashes Complex Network Analysis
  • Frequent Pattern tree (FP tree)
  • Fuzzy C-means clustering (FCM)
  • Bayesian network
  More +
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Edition:
  Final: June 2014 – January 2018
• Corporate Publisher:
  Transportation Informatics University Transportation Center
• Abstract:
  The field of traffic accident analysis has long been dominated by traditional statistical analysis. With the recent advances in data collection, storage and archival methods, the size of accident datasets has grown significantly. This in turn has motivated research on applying data mining and Machine Learning algorithms, which are specifically designed to handle datasets with large dimensions, to traffic accident analysis. This project explores three specific applications of Data Mining and Machine Learning algorithms to traffic accident analysis. The first application explores the potential for using a modularity-optimizing community detection algorithm and association rules learning algorithm, to identify important accident characteristics. The second application proposes a novel Frequent Pattern tree (FP tree) based variable selection method, and then develops models for the real-time prediction of traffic accident risk. Finally, the third application proposes a novel approach to developing accident duration prediction models. The approach improves on the original M5P tree algorithm through the construction of a M5P-Hazard-Based Duration Model (HBDM).
• Format:
  PDF
• Funding:
  DTRT13-G-UTC48
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/40076/dot_40076_DS1.pdf
• File Type:
  [PDF - 1.91 MB]
• Collection(s):
  University Transportation Centers ; US Transportation Collection
• Main Document Checksum:
  urn:sha-512:878ffb1eef5c62937a410175329fec1e68001c14aeb3785b6c2a8c5d7c2917c37d51f7808a0b60b04976d69b7cb3b0f3842f6585e26aa84e217455494e662552