Development of a crash risk prediction model using the k-Nearest Neighbor algorithm

Min Ji Kang, Oh Hoon Kwon, Shin Hyoung Park

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

This study aims to create a crash risk prediction model using k-Nearest Neighbor, one of the machine learning algorithms. Based on the traffic flow information collected by an advanced traffic management system (ATMS) and the corresponding crash historical information and weather information, this model derives the probability of a crash occurrence by looking for the most similar conditions at the time of a past accident. The predicted results of the model were evaluated using the metrics of the receiver operating characteristic (ROC) curve and area under the curve (AUC), which indicated that model performance belongs to the good side. The results of this study are expected to upgrade the safety management system of the ATMS further and contribute to reducing crash occurrence by giving preemptive notification to drivers.

Original languageEnglish
Title of host publicationAdvanced Multimedia and Ubiquitous Engineering - MUE/FutureTech 2018
EditorsJames J. Park, Kim-Kwang Raymond Choo, Gangman Yi, Vincenzo Loia
PublisherSpringer Verlag
Pages835-840
Number of pages6
ISBN (Print)9789811313271
DOIs
StatePublished - 2019
Event13th International Conference on Future Information Technology, FutureTech 2018 - Salerno, Italy
Duration: 23 Apr 201825 Apr 2018

Publication series

NameLecture Notes in Electrical Engineering
Volume518
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

Conference13th International Conference on Future Information Technology, FutureTech 2018
Country/TerritoryItaly
CitySalerno
Period23/04/1825/04/18

Keywords

  • Advanced traffic management system (ATMS)
  • Big data
  • Crash risk prediction
  • Intelligent transportation system (ITS)
  • k-Nearest neighbor

Fingerprint

Dive into the research topics of 'Development of a crash risk prediction model using the k-Nearest Neighbor algorithm'. Together they form a unique fingerprint.

Cite this