A Comparison of the Performance of Ensemble Tree and Neural Networks for The Prediction of Traffic Accident Duration

Hüseyin Korkmaz; Mehmet Ali Ertürk; Mehmet Adak

Article Details

A Comparison of the Performance of Ensemble Tree and Neural Networks for The Prediction of Traffic Accident Duration

Authors:

Hüseyin Korkmaz | Mehmet Ali Ertürk | Mehmet Adak

Publisher:

Istanbul University

Main Topics:

Ulaşım | Lojistik ve Tedarik Zincirleri (Diğer)

Journal:

Journal of Transportation and Logistics Vol.9 Issue 1, pp.20-47

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001313792

pISSN:

2459-1718

Keywords:

Abstract:

 Traffic accident duration is defined as the time difference between the occurrence and the return of the accident scenes initial state. The aim of this paper is to predict the traffic accident duration based on traffic accident data in Istanbul with Ensemble Tree and Neural Networks methods and to compare the performance of these methods. The secondary aim of the paper is to identify the main factors affecting the accident duration. The accident data sets obtained from Istanbul Metropolitan Municipality and General Directorate of Security are used in this paper. The dataset includes 1.905 traffic accident records in Istanbul from 2013 to 2021. The data were analyzed within the scope of data mining. Statistical tests and machine learning algorithms were applied to the extracted data set and prediction of traffic accident duration was performed. R², MSE, RMSE and MAE metrics were used for the performance measures of the algorithms applied in this paper. It was found that the Ensemble Tree algorithm achieved an R-Square of 0.85 in training, while the Neural Networks algorithm performed better with 0.91 in testing.

 Traffic accident duration is defined as the time difference between the occurrence and the return of the accident scenes initial state. The aim of this paper is to predict the traffic accident duration based on traffic accident data in Istanbul with Ensemble Tree and Neural Networks methods and to compare the performance of these methods. The secondary aim of the paper is to identify the main factors affecting the accident duration. The accident data sets obtained from Istanbul Metropolitan Municipality and General Directorate of Security are used in this paper. The dataset includes 1.905 traffic accident records in Istanbul from 2013 to 2021. The data were analyzed within the scope of data mining. Statistical tests and machine learning algorithms were applied to the extracted data set and prediction of traffic accident duration was performed. R², MSE, RMSE and MAE metrics were used for the performance measures of the algorithms applied in this paper. It was found that the Ensemble Tree algorithm achieved an R-Square of 0.85 in training, while the Neural Networks algorithm performed better with 0.91 in testing.

 Traffic accident duration is defined as the time difference between the occurrence and the return of the accident scenes initial state. The aim of this paper is to predict the traffic accident duration based on traffic accident data in Istanbul with Ensemble Tree and Neural Networks methods and to compare the performance of these methods. The secondary aim of the paper is to identify the main factors affecting the accident duration. The accident data sets obtained from Istanbul Metropolitan Municipality and General Directorate of Security are used in this paper. The dataset includes 1.905 traffic accident records in Istanbul from 2013 to 2021. The data were analyzed within the scope of data mining. Statistical tests and machine learning algorithms were applied to the extracted data set and prediction of traffic accident duration was performed. R², MSE, RMSE and MAE metrics were used for the performance measures of the algorithms applied in this paper. It was found that the Ensemble Tree algorithm achieved an R-Square of 0.85 in training, while the Neural Networks algorithm performed better with 0.91 in testing.

 Trafik kaza süresi, bir kazanın meydana gelmesi ile kaza yerinin başlangıç durumuna dönmesi arasındaki zaman farkı olarak ifade edilmektedir. Bu araştırmanın birincil amacı İstanbuldaki trafik kaza verilerine dayalı olarak trafik kaza süresini Topluluk Ağacı ve Sinir Ağları yöntemleri ile tahmin etmek ve bu yöntemlerin performanslarını karşılaştırmaktır. Araştırmanın ikincil amacı ise trafik kaza süresini etkileyen temel faktörleri belirlemektir. Bu araştırmada İstanbul Büyükşehir Belediyesi ve Emniyet Genel Müdürlüğü kurumlarından elde edilen İstanbula ait kaza bilgisi veri setleri kullanılmıştır. Veri seti, 2013-2021 yılları arasındaki İstanbulda gerçekleşen 1.905 trafik kaza kaydını içermektedir. Veriler, veri madenciliği kapsamında incelenmiştir. Ayıklanan veri setine istatistik testleri ve makine öğrenmesi algoritmalarından Topluluk Ağacı ve Sinir Ağları uygulanarak trafik kaza süresi tahmini gerçekleştirilmiştir. Bu araştırmada uygulanan algoritmaların performans ölçümleri için R², MSE, RMSE ve MAE metrikleri kullanılmıştır. Topluluk Ağacı algoritmasının eğitimde R-Kare: 0.85 ile başarılı bir performans elde ettiği, testte ise R-Kare: 0.91 ile Sinir Ağları algoritmasının daha iyi performans gösterdiği sonucuna ulaşılmıştır.

Article Details

A Comparison of the Performance of Ensemble Tree and Neural Networks for The Prediction of Traffic Accident Duration

Authors:

Hüseyin Korkmaz | Mehmet Ali Ertürk | Mehmet Adak

Publisher:

Istanbul University

Main Topics:

Ulaşım | Lojistik ve Tedarik Zincirleri (Diğer)

Journal:

Journal of Transportation and Logistics Vol.9 Issue 1, pp.20-47

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001313792

pISSN:

2459-1718

Accident Duration

Intelligent Transportation Systems

Machine Learning

Neural Networks

Ensemble TreeKaza Süresi

Akıllı Ulaşım Sistemleri

Makine Öğrenmesi

Sinir Ağları

Topluluk Ağacı

Issues

Article Details

Authors: Hüseyin Korkmaz | Mehmet Ali Ertürk | Mehmet Adak

Publisher: Istanbul University

Main Topics: Ulaşım | Lojistik ve Tedarik Zincirleri (Diğer)

Journal: Journal of Transportation and Logistics Vol.9 Issue 1, pp.20-47

Document Type: Article

Publication Date: 5/2024

Document Language: Turkish

Document ID (Org No.): 2001313792

pISSN: 2459-1718

Issues

Authors:

Hüseyin Korkmaz | Mehmet Ali Ertürk | Mehmet Adak

Publisher:

Istanbul University

Main Topics:

Ulaşım | Lojistik ve Tedarik Zincirleri (Diğer)

Journal:

Journal of Transportation and Logistics Vol.9 Issue 1, pp.20-47

Document Type:

Article

Publication Date:

5/2024

Document Language:

Turkish

Document ID (Org No.):

2001313792

pISSN:

2459-1718