Intelligent evaluation method for landslide susceptibility based on feature stability guidance and cost adaptive tightening

By employing an intelligent evaluation method guided by feature stability and adaptive cost contraction, the problems of feature selection and model generalization in landslide susceptibility assessment are solved. This method achieves efficient feature subset optimization and asymmetric cost-driven optimization, thereby improving the robustness and prediction accuracy of the model.

CN122451561APending Publication Date: 2026-07-24TIBET UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202610457924.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-09
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing landslide susceptibility assessment methods have technical deficiencies in areas such as global feature selection, feature redundancy removal, asymmetric cost adaptability, and hyperparameter adaptive adjustment, resulting in insufficient generalization ability and interpretability of the models in complex geological environments.

Method used

We adopt an intelligent evaluation method based on feature stability guidance and cost adaptive contraction. Through global feature importance ranking, feature stability self-evaluation and contraction priority construction, asymmetric cost-sensitive contraction and self-evolutionary threshold closure, we achieve global optimal search of high-dimensional feature subsets, stability-aware precise contraction and asymmetric cost-driven evaluation.

Benefits of technology

It significantly improves the model's generalization ability and interpretability, reduces the number of features by 40%, increases the recall rate by 6.8%, and achieves an AUC of 0.897 on the test set, which is superior to traditional methods and has high engineering application value.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a landslide susceptibility intelligent evaluation method based on feature stability guidance and cost adaptive tightening. The method comprises the following steps: acquiring and normalizing landslide sample data; feature importance is calculated through a random forest and sorted; traversing a feature subset scale by using a gradient boosting tree, and determining a global optimal subset through cross validation AUC; calculating a feature stability coefficient, and generating a compressed priority queue; sequentially trying to remove features by using asymmetric cost sensitive AUC measurement, and accepting if the performance reduction does not exceed a threshold value; after confrontation verification, adaptively adjusting a threshold value and tightening again, and recording a historical optimal subset; and retraining the model to output the landslide probability. According to the method, through feature stability guiding, asymmetric cost-sensitive tightening and self-evolution threshold closed loop, redundant features are simplified under the condition that precision is kept, and generalization and interpretability are improved. The actual measurement AUC reaches 0.897, which is improved by 3.2% compared with the traditional GBDT, and the characteristics are reduced by 40%.
Need to check novelty before this filing date? Find Prior Art