Estimating Road Accidents of Turkey Based on Regression Analysis and Artificial Neural Network Approach

An artificial neural network (ANN) model and two new analytical models are proposed by this study for estimation of the number of Turkish accidents, injuries, and fatalities utilizing 1986-2005 historical data. Data for 1986-2000 were used for model development and data for 2001-2005 were utilized for developed model testing. A modified form of the Smeed accident prediction model is the first of the analytical models. A form of the Andreassen model adapted to Turkey is the second model. Model parameters were population, accidents, injuries, fatalities, and number of vehicles in model development. The sigmoid and linear functions in the ANN model were used, with feed forward-back proportion algorithm, as activation functions. There was comparison between model results and observations, and the ANN model was found to perform better than the other two analytical models. For model performance investigation for future estimations, a 2006-2020 15 year period was employed. Two possible scenarios were used to evaluate road safety strategies, considering that by 2020, Turkey is likely to enter the European Union. There are 1.7% and 7.5% respective assumed annual average population and number of vehicles growth rates (average 1986-2005 growth rates) in the first scenario. The average per capita number of vehicles, in the second scenario, is assumed to reach 0.45, representing, in 15 years, a three-fold increase. The suitability of current road safety application methods is revealed through results obtained from both scenarios.

• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/18245463
• Authors:
  • Akgungor, A P
  • Dogan, E
• Publication Date: 2008-11

Language

• English

Media Info

• Media Type: Print
• Edition: Sections A & B
• Features: Figures; References; Tables;
• Pagination: pp 11-22
• Serial:
  • Advances in Transportation Studies
  • Volume: 16
  • Publisher: University Roma Tre
  • ISSN: 1824-5463
  • Serial URL: http://www.atsinternationaljournal.com/

Subject/Index Terms

• TRT Terms: Estimation theory; Mathematical models; Mathematical prediction; Neural networks; Regression analysis; Traffic crashes
• Geographic Terms: Turkey
• Subject Areas: Data and Information Technology; Highways; Planning and Forecasting; Safety and Human Factors; I72: Traffic and Transport Planning;

Filing Info

• Accession Number: 01114089
• Record Type: Publication
• Files: TRIS
• Created Date: Oct 30 2008 6:52AM