Model construction method and device based on rejection inference method and electronic equipment

Abstract

The invention provides a model construction method and device based on a rejection inference method and electronic equipment. The method comprises the following steps: acquiring full-scale sample data, and defining positive and negative samples for the received samples to establish a training data set, wherein the the training data set comprises user feature data and financial performance data; constructing an initial risk assessment model, and training the initial risk assessment model by using the training data set; scoring the rejected samples by using the trained initial risk assessment model to obtain the deterioration probability of each rejected sample; adopting a rejection inference method to carry out weighted expansion on rejected samples, and carrying out weighted processing onall received samples; integrating the receiving sample and the rejecting sample after weighting processing, and establishing a new training data set; and retraining the initial risk assessment model by using the new training data set to obtain a final risk assessment model. According to the method, the problems of sample deviation or sample data missing and the like are effectively solved, and themodel prediction precision is improved.

Description

technical field [0001] The invention relates to the field of computer information processing, in particular to a model building method, device and electronic equipment based on a rejection inference method. Background technique [0002] Risk forecasting is the quantification of risk and a key technology of risk management. At present, risk prediction is generally carried out through modeling. In the process of model establishment, there are mainly steps such as data extraction, feature generation, feature selection, algorithm model generation, and rationality assessment. [0003] In the prior art, the main purpose of financial risk prediction is how to distinguish good customers from bad customers, evaluate the risk situation of users, so as to reduce credit risk and realize profit maximization. In addition, as the sources of data become more and more abundant, more and more data can be used as risk characteristic variables. However, when building a model, usually only use...

AI Technical Summary

Problems solved by technology

However, when building a model, usually only users who have passed (or user data with labels) are used as modeling samples, without considering the good or bad status of those rejected users, resulting in the built model always "using some sample data to estimate the population", so there is a bias in parameter estimation

Furthermore, there are still problems such as not using unlabeled user data and other related data, which will lead to inaccurate model calculations and low accuracy of risk prediction

In addition, there is still much room for improvement in terms of model calculation accuracy and data update

Images