Automatic feature engineering credit risk evaluation system and method

Abstract

The invention provides an automatic feature engineering credit risk evaluation system and method, and belongs to the field of computers. The system comprises an original feature extraction module, a new feature derivation module, a new feature inspection module, an automatic feature generation module and a customer credit risk classification model. The method comprises the steps of collecting customer data to extract original features, constructing a user-defined conversion function to derive new features, constructing a verification function to verify the new features, automatically generating features, constructing a customer credit classification model, generating a sample training classification model according to a feature set, and classifying the customer data by using the trained model. The credit score card model constructed by the method has an interpretable automation characteristic with a causal relationship, the modeling time is improved by more than 50%, and the model classification effect is more remarkable.

Description

technical field [0001] The invention relates to technologies such as computer, big data mining, and credit score card models, and relates to the application of automatically generated causality features in the credit score card model, and in particular to an automated feature engineering credit risk evaluation system and method. Background technique [0002] The credit score card model is a delineation model of personal financial authority established in recent years to ensure the financial security of banks and other financial sectors. It is a decision-making tool for risk management and control. This model refers to the analysis and mining of customer credit behavior data based on the customer's credit history data, using a certain credit scoring model to obtain credit scores of different levels, and according to the customer's credit score, to determine the amount of money that the customer can hold. Thereby ensuring the security of repayment and other businesses. With t...

AI Technical Summary

Problems solved by technology

[0004] In recent years, with the rise of machine learning and deep learning, many financial institutions have begun to try to use black-box models such as nonlinear models or neural networks to make predictions. However, in traditional financial scenarios, the output results of the models are required to be interpretable and traceable. Non-linear models have no way to establish a causal relationship between input features and prediction results, and naturally have no way to explain the prediction results

Secondly, to complete a credit score card model, it is generally necessary to construct hundreds of model features. The main expert industry experience in constructing model features is limited by past experience. It is difficult to find some non-linear but causal indicators. Re-doing feature work is not only time-consuming and labor-intensive, but may also cause omissions

Images