Model generation method and device and information processing method and device

Abstract

The invention discloses a model generation method and device and an information processing method and device. The method comprises the following steps: generating two or more than two graph embeddingvectors according to a meta-path node sequence of a sample customer of a heterogeneous information network; inputting each obtained graph embedding vector into a neural network model, and carrying outjoint construction on the neural network model, a customer feature vector of a sample customer and a training label to obtain a scoring model for risk evaluation; wherein each graph embedding vectoris generated by a set of meta-path node sequences generated according to a user identity node and a node except the user identity node; wherein the training label is obtained according to the customerservice state of the sample customer. According to the embodiment of the invention, the scoring model is constructed based on the plurality of graph embedding vectors generated by the meta-path nodesequence, so that the interpretability of the scoring model is improved.

Description

technical field [0001] This article involves but is not limited to technology, especially a method and device for generating a model, and an information processing method and device. Background technique [0002] When traditional financial institutions conduct credit risk quantification, the modeling method used is to construct a scorecard model. The construction of the scorecard model includes: screening out the characteristic variables that describe customer risks through feature engineering; training logistic regression through the screened out characteristic variables Model; convert the logistic regression model obtained through training into a scorecard model. This method has high interpretability, which is beneficial for risk control personnel to control risks by adjusting risk control policies. [0003] With the development of big data technology, some financial institutions began to apply relational network technology to the scoring model for credit risk quantificat...

AI Technical Summary

Problems solved by technology

For heterogeneous information networks with richer information (hereinafter referred to as heterogeneous networks), since there are different types of nodes and edges in the network, and nodes and edges contain a lot of attribute information, it is impossible to classify the types and attributes of nodes and edges When the information is fused into the graph embedding, the generated graph embedding vector will not be able to effectively represent the information of the customer's heterogeneous network, resulting in the lack of interpretability of the constructed scoring model. This is a problem that needs to be solved

Images