25064

ارائه مدل پيش بيني تصادفات درون شهري تحت تاثير انواع كاربري زمين با استفاده از روشهاي يادگيري ماشين و شناسايي نقاط حادثه خيز توسط GIS

كارشناسي ارشد

مهندسي عمران - گرايش راه و ترابري

1397

بهمن 1399

دكتر ندا كامبوزيا - دكتر برات مجردي

مهندسي عمران

1400/05/02

Provide Predicting model of urban accidents under the influence of land use type by using machine learning methods and the identification of hotspots by GIS

2/19/2022 12:00:00 AM

Car accidents in both urban and roads, always cause loss of life and finances to governments and the people. In the meantime, urban accidents, despite the low death toll, are significant due to the high number and the imposition of financial and time losses on citizens. Land use is a parameter at the crash site, it will indirectly affect the number and severity of accidents by affecting the absorption of travel. In the recent researches, by using accident information from 1393 to 1395 in three areas of Mashhad (Zones 9, 10, and 12), urban accidents were modeled based on the type of land use and finding the current and future pattern of accidents.Logistic regression model expresses the existing pattern of urban accidents in the regions and to find the future pattern of accidents and predict urban accidents, two neural network algorithms MLP and RBF are used as one of the machine learning methods by SPSS software. The results of this study showed that in residential applications, the human factor (Useless haste and acceleration) according to the existing model and future model, is the main cause of accidents in this type of applications. The results of this study showed that in residential areas, the human factor (Useless haste and acceleration) according to the existing model and future model, is the main cause of accidents in this type of area. While in the office-commercial district, the condition of the carriageway and the weather, and the lighting condition is the most important factor in predicting accidents. After presenting urban accident prediction models based on the type of land use, using ArcMap software and analyzing hotspots to identify Accident-prone areas in each section and investigate the effect of land use type on the accident rate of roads paid. The results of the analysis of Accident-prone areas also showed that land use will affect the accident rate of roads insofar as that in the vicinity of educational, cultural, and sports area, and green space, the accident rate of the road (number of casualties) will be reduced. While approaching official-residential land uses, the accident rate (number of casualties) increases. Also, according to the Ordinary Kriging method, it can be concluded that in general, the accident rate in the central areas of areas 9 and 10 decreases and gradually increases with approaching the border areas of the accident; But in District 12, the lower half is less accident-prone.

Urban accident modeling, land use, logistic regression, neural network, accident hotspots