We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Trafik Kazalarının Sınıflandırılmasında Çok Katmanlı Algılayıcı, Regresyon ve En Yakın Komşuluk Algoritmalarının Performans Analizi.
- Authors
KUŞKAPAN, Emre; ÇODUR, Muhammed Yasin
- Abstract
With the increasing population worldwide, the number of vehicles also increases. The increase in the number of vehicles brings with it many problems. The most important of these problems is traffic accidents. The situation where traffic accidents can cause material and moral losses reveals the necessity of working in this field. Classification is needed for better analysis and easy interpretation of traffic accidents. In this context, various classification methods and computer programs are developed with the development of technology and the introduction of artificial intelligence technologies into human life. In this study; the years are classified according to death and injury situations by using traffic accident data occurring year by year in our country. Then, with the WEKA analysis program, multilayer perceptron, regression and classification performances and error criteria of the nearest neighbor methods were calculated. When the classification values of all three algorithms are compared with each other, it has been found that the nearest neighbor algorithm gives better results in many criteria in terms of both performance analysis and error criteria. Thanks to this study, it has been determined that the rate of death and injury in traffic accidents that have occurred in recent years has reached a high risk level again as it was in the early 2000s. This situation is important for decision makers to increase their measures to reduce traffic accidents. On the other hand, by examining the classification performances, it was revealed which algorithm can be preferred in the classification process of the data set with similar characteristics.
- Publication
Journal of Polytechnic, 2022, Vol 25, Issue 1, p373
- ISSN
1302-0900
- Publication type
Article
- DOI
10.2339/politeknik.697530