Abstract: In this paper,the distribution of random recovery rate is studied,and double Beta distribution density model is established.This model has a characteristic of two peaks,which meets the new findings of Moody company,and improves the result that the present models all have only one peak.By using the sequential optimization algorithm based the NT-net to estimate parameters and the tools of kernel density estimation,the demonstration study is made.The result shows that the fitting error of double Beta model is quite little and far less than that of Beta model,so double Beta model is an ideal model to denote the distribution of recovery rate.The method is given to draw the random numbers from double Beta model at the end of the paper.

Key words: recovery rate, double Beta distribution, kernel density estimation, two peak distribution, se- quential optimization algorithm