EN
The problem in the analysis of insurance data is modeling the number of claims occurring in a given portfolio policy using regression assuming a Poisson distribution which is not always justified, since sometimes the data contains a large number of zeros. This paper presents a generalized Poisson regression for the counter variable and a modified version of Poisson regression taking into account the situation of the presence of a large number of zeros in the data (called zero-inflated Poisson regression). Various types were analyzed in order to determine which variables influence the occurrence tarification zeros in the portfolio policy using the procedure 10 times the patch validation. The result is ranking for classification policies because of the number of generated damage.