1results about How to "Diversity" patented technology

This invention discloses a method for identifying multi-feature interaction patterns in medical insurance fraud, comprising: constructing a RuleFit model based on gradient boosting decision trees (GWO) according to medical insurance data, wherein the GWO generates composite fraud rules, linear regression fitting generates linear fraud rules and rule quantification indicators, and an initial fraud rule set is obtained; optimizing the fraud rule set based on the Grey Wolf Optimization Algorithm (GWO), mining the optimal synergistic relationship between rules through multi-path parallel search, and iteratively filtering to obtain the optimal fraud rule set; constructing a 0-1 rule matrix and binary classification target labels based on the optimal fraud rule set, and classifying and identifying medical insurance fraud behavior through logistic regression. This invention, based on RuleFit, can generate a fraud rule set reflecting feature correlation; based on GWO, it achieves multi-path parallel search within the solution space composed of the full set of candidate rules, fully exploring the interaction and combination potential between rules.