CLUSTER ANALYSIS FOR ROAD ACCIDENTS INVESTIGATION

The paper describes a procedure for the evaluation of road accidents in order to obtain a clustering of similar accidents on the basis of their characteristics. The procedure is formed by different interrelated steps that allow identifying homogeneous groups of accidents. The proposed clustering allows separating the elements that are not similar, but, at the same time, the individual characteristics of each event is maintained in the group. Discussed also is further development that can be expected by defining a different score function and by constructing an analytical function that links the main identified factors for each representative event to the generic accident, in order to verify the impact produced by modifying one ormore of the values of the factors.