A new random forest method based on belief decision trees and its application in intention estimation

Random forest algorithm is a classification and prediction model, which is used in many fields. Random forest is composed of multiple decision trees. In the face of more and more complex uncertain environments, ordinary decision trees can no longer meet the requirements, so belief trees based on belief functions appear. This paper proposes a new random forest method based on belief trees. Compared with ordinary random forest in which voting or average method is used to combine the result of each decision tree, the proposed method fully considers the influence of the weight of each tree, and combine the result of each belief tree through a weighted averaging combination of belief structures. In order to demonstrate the effectiveness of the proposed method, it is used in intention estimation. The results show that the accuracy of intention recognition is improved by using the proposed method compared with original random forest algorithm.