公共场所火灾的责任风险不同于一般的责任风险,它的低概率、高损失和厚尾性等特点使得它的风险定价模型不能直接采用被广泛应用的广义线性定价模型,也正是由于此,该问题鲜见人研究。根据公共场所火灾第三者责任风险样本数据特点,传统的参数分布假设无法得到较好的效果,本文用非参数高斯核密度函数替代广义线性模型的指数分布族,对火灾公众责任风险进行定价,取得了较好的效果。 The responsibility risk of public places fire is different from the general liability risk. Its characteristics of low probability, high loss and thick tail make its risk pricing model can not directly adopt the widely used generalized linear pricing model. It is precisely because of this, The problem of rare study. According to the characteristics of risk data of the third party in public places, the traditional assumption of parameter distribution can not get good results. In this paper, the non-parametric Gaussian kernel density function is used to replace the index distribution of generalized linear model, Achieved good results.