Akademik Veri Yönetim Sistemi
14th INTERNATIONAL CONGRESS ON ENGINEERING, ARCHITECTURE AND DESIGN, İstanbul, Türkiye, 28 - 29 Aralık 2024, ss.440-446, (Tam Metin Bildiri)
Every year, millions of people lose their lives, sustain injuries, or suffer
disabilities because of traffic accidents. Humans, vehicles, and road conditions are the
primary factors contributing to traffic accidents. Although modern intelligent
transportation systems have been integrated into the vehicles we use today, the human
factor remains the most critical cause of accidents. It is essential to thoroughly investigate
the human and environmental factors contributing to these accidents and reduce the
resulting deaths and injuries. This study aimed to analyze and identify the contributing
features of traffic accidents, in which only drivers and passengers were injured in
Diyarbakr between 2013 and 2022. Machine learning algorithms were used to predict
the severity of these accidents. Since the dataset was imbalanced, the Synthetic Minority
Oversampling Technique (SMOTE) method was applied to balance the minority classes.
Driver faults, particularly unsafe speed and rear-end collisions, were found to be the most
common causes of traffic accidents. The analyses revealed that traffic accidents peaked
during morning and evening commuting hours. Mondays had the highest daily accidents
due to the start of the workweek. Unsafe speed and rear-end collisions were identified as
the most common driver faults. The Random Forest algorithm achieved the highest
performance with an accuracy of 91.5%.