Abstract: This paper proposes a forest fire risk assessment method based on ensemble feature selection. Considering the diversity and independence of algorithms, 15 kinds of feature select algorithms are selected to form the heterogeneous selectors based on their difference. By using the feature select algorithms, a feature subset set is obtained. And then a forest fire risk assessment model is constructed based on BP neural network by using each feature subset. The important factors of forest fires are selected based on the accuracy of neural network to construct the optimal forest fire risk assessment model. The results show that the accuracy of the algorithm proposed is 85.96%. The proposed model has good generalization ability and can assess forest fire risk effectively.

Key words: ensemble feature selection, forest fire, BP neural network, risk assessment