基于Huber损失函数的稳健随机森林模型及应用OACHSSCD

With the development of artificial intelligence technology,random forest model has been rapidly developed and widely used in many fields.However,in practical problems,the traditional random forest model is vulnerable to heavy-tailed data-sets or outliers,resulting in estimation bias.In view of this,this paper proposes a robust random forest model based on Huber loss function(HuberRF),and gives the estimation algorithm,variable importance measurement method,and partial dependence rela-tionship measurement method.With the advantage of Huber loss function,this model has better robustness when dealing with data with skewed distribution or outliers,and can better weaken the adverse impact of extreme outliers on model estimation.The nu-merical simulation results show that when dealing with data with skewed distribution or outliers,the robust random forest model based on Huber loss function significantly outperforms the mean regression forest model and the median regression forest model in terms of prediction performance.When the robust random forest model based on Huber loss function is applied to the data set of China's county digital finance and farmers'income,the results indicate that the proposed method has better robustness and great-er prediction ability than the traditional random forest models.