Credit risk control model generation method and device, score card generation method, machine readable medium and equipment

Abstract

The invention provides a credit risk control model generation method. The method comprises the following steps: determining to carry out feature engineering processing on original attribute data of acredit business object by utilizing a pre-trained GBDT model with the maximum depth of a base classifier node being 1; and training and generating an LR score card model based on the data processed bythe feature engineering, and taking the LR score card model as a credit risk control model. According to the characteristic that a gradient boosting tree of which the maximum tree depth is limited tobe 1 can be degraded into a linear model, automatic and end-to-end training of the LR score card model is realized, and the prediction performance of feature screening and continuous variable binningsignificantly superior to that of heuristic rules is achieved, so that automatic and end-to-end training of a linear, interpretable and high-performance machine learning classification model is realized.

Description

technical field [0001] The present invention relates to the field of credit risk control, in particular to a method and device for generating a credit risk control model, a method for generating a scorecard, a machine-readable medium and equipment. Background technique [0002] The classification problem is one of the most important problems that a supervised machine learning model can solve. In reality, problems such as credit risk control, fraud identification, and recommendation recall can be abstracted into classification problems. Statistical machine learning models (that is, narrowly defined machine learning model) or deep learning model (neural network) to solve. Among them, statistical machine learning models are more used for tabular and structured data mining tasks, while deep learning models are more suitable for unstructured data sets such as image recognition, speech recognition, and natural language processing. [0003] At present, the most mainstream statisti...

AI Technical Summary

Problems solved by technology

[0004] At present, there is a lack of a method to realize the automatic and end-to-end training of the LR scorecard under the premise of ensuring the prediction performance of the model; there is also a lack of a method to realize the automatic and end-to-end training like GBDT for the nonlinear machine learning classification model. End-to-end training of high-performance linear machine learning classification models has become the standard general solution for linear machine learning classification models

Images