An innovative methodology to analyse urban accidents

摘要

Urban accidents have the highest percentage impact (approx 75percent) over the entirety of accidents, therefore they represent a crucial event which potentially may lead to disastrous consequences. The understanding of this phenomenon is not easy because of the numerous variables which must be taken into account. Traditional statistical approaches may be not sufficient to tackle this problem and hence new methods should be carried out. Artificial Intelligence may be helpful for providing more powerful techniques. In a previous paper some of the authors used neural networks to analyse accidents occurred in a small city (about 100.000 inhabitants). In this paper the same AI methodology is applied to study accidents occurred in the metropolitan area of the city of Milan over a period of four years. The high number of data gives to the analysis a good statistical significance. The paper describes the methodology used to prepare data, it is rather important to discharge incomplete or wrong data; to fill in accidents records with geometrical and flow characteristics of road intersections; to calculate an accidents index to weigh dangerousness of the place where accidents occurred.