Development of accident prediction model by using artificial neural network (ANN)

• Main Author: Ramli, Mohd Zakwan
• Format: Thesis
• Language: English; English; English
• Published: 2011
• Subjects: HE Transportation and Communications; HE5601-5725 Automotive transportation Including trucking, bus lines, and taxicab service
• Online Access: http://eprints.uthm.edu.my/2728/1/24p%20MOHD%20ZAKWAN%20RAMLI.pdf; http://eprints.uthm.edu.my/2728/2/MOHD%20ZAKWAN%20RAMLI%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/2728/3/MOHD%20ZAKWAN%20RAMLI%20WATERMARK.pdf; http://eprints.uthm.edu.my/2728/
• Institution: Universiti Tun Hussein Onn Malaysia

Statistical or crash prediction model have frequently been used in highway safety studies. They can be used in identify major contributing factors or establish relationship between crashes and explanatory accident variables. The measurements to prevent accident are from the speed reduction, widening the roads, speed enforcement, or construct the road divider, or other else. Therefore, the purpose of this study is to develop an accident prediction model at federal road FT 050 Batu Pahat to Kluang. The study process involves the identification of accident blackspot locations, establishment of general patterns of accident, analysis of the factors involved, site studies, and development of accident prediction model using Artificial Neural Network (ANN) applied software which named NeuroShell2. The significant of the variables that are selected from these accident factors are checked to ensure the developed model can give a good prediction results. The performance of neural network is evaluated by using the Mean Absolute Percentage Error (MAPE). The study result showed that the best neural network for accident prediction model at federal road FT 050 is 4-10-1 with 0.1 learning rate and 0.2 momentum rate. This network model contains the lowest value of MAPE and highest value of linear correlation, r which is 0.8986. This study has established the accident point weightage as the rank of the blackspot section by kilometer along the FT 050 road (km 1 – km 103). Several main accident factors also have been determined along this road, and after all the data gained, it has successfully analyzed by using artificial neural network.