Research on an Innovative Feature Importance Recognition Algorithm Based on GINI-OOB Index

Based on the decision tree algorithm (CART), this paper constructs an integrated model and trains the random forest algorithm (RF). On this basis, the GINI coefficient and the OOB coefficient are derived respectively and the coupling of the two is studied. Based on 10 importance indicators for training and learning, a more efficient, accurate and robust importance recognition algorithm is proposed. The results show that: (1) There are significant differences in the results of coupling Gini index and OOB index in different aspects, and the improved algorithm has stronger importance recognition accuracy. (2) The recognition algorithm proposed in this paper can identify the importance of financial risks. The research in this paper expands the identification methodology of importance.