Risk identification model optimization method and device

Abstract

The invention discloses a risk identification model optimization method and device. The method comprises the following steps: acquiring a false identification sample and N alternative samples; inputting the error identification sample and the feature vector of each alternative sample into a deep learning model for processing, and extracting the processed feature vector from the middle layer of thedeep learning model to serve as a standardized feature vector of the corresponding sample; according to the standardized feature vector of the false identification sample and the standardized featurevectors of the M alternative samples, determining K alternative samples similar to the false identification sample from the M alternative samples by adopting a K nearest neighbor algorithm, and taking the K alternative samples as similar samples; and determining the mark of each similar sample as a first risk type, and optimizing the risk identification model based on the error identification sample and each similar sample.

Description

technical field [0001] The embodiments of this specification relate to the field of machine learning, and in particular to a method and device for optimizing a risk identification model. Background technique [0002] When the risk identification model is actually used, it is possible to find objects that are misidentified by the risk identification model. For example, the risk identification model identifies an actual risky object to be identified as risk-free, or identifies an actual risk-free object to be identified as risky. In this case, it is generally necessary to optimize the risk identification model by taking the found misidentified objects as samples, so as to improve the identification accuracy of the risk identification model. [0003] However, the number of misidentified objects found is often limited, and risk identification models cannot be well optimized based only on these small number of objects. Contents of the invention [0004] In order to further im...

AI Technical Summary

Problems solved by technology

[0003] However, the number of misidentified objects found is often limited, and risk identification models cannot be well optimized based on these small number of objects only.

Images