Abstract:
We offer the credibility model for claim numbers (evolutionary model), where risk parameter is an intensity of a claim amounts, which is allowed to vary by dependent way in successive period. The special type of dependency structure is generated by convolution procedure with «overlapping» components. We investigate some properties of offered model, in particular, the example of weakly stationary sequence is considered, the expression for parameters of the credible mean optimal linear forecast is obtained, the recursive formula for linear forecast parameters is introduced.