Scorecard modeling method and apparatus, computer device and storage medium

Abstract

The embodiment of the application discloses a scorecard modeling method and device, a computer device and a storage medium. The method comprises the following steps: obtaining customer data; Countingmissing values corresponding to each of the variable data, and processing the variable data according to preset processing rules corresponding to the missing values to obtain sample data; Extracting afeature vector matrix corresponding to a preset number of principal components according to the sample data based on a preset principal component extraction rule; Based on a logistic regression model, performing modeling training according to the feature vector matrix to complete modeling of the scorecard. By extracting principal components instead of manual grouping which is the most time-consuming in the traditional modeling process, the method optimizes the modeling flow, greatly shortens the modeling time and improves the modeling efficiency.

Description

technical field [0001] The present application relates to the technical field of the Internet, and in particular to a scoring card modeling method, device, computer equipment and storage medium. Background technique [0002] Internet finance is booming in recent years. Internet finance has the characteristics of both the Internet and finance. Compared with traditional finance, Internet finance relies more on big data risk control systems, and judges credit risks based on users' multi-dimensional information. The scorecard is a common model in the risk control model. Among them, the logistic model is widely used in the scorecard. The main modeling steps include data preparation, variable exploration, model building and model verification. [0003] At present, the data preparation process of logical model modeling mainly has the following problems: it takes a lot of time in the process of variable exploration, and the part of the work involving data grouping is cumbersome; usu...

AI Technical Summary

Problems solved by technology

[0003] At present, the data preparation process of logical model modeling mainly has the following problems: it takes a lot of time in the process of variable exploration, and the part of the work involving data grouping is cumbersome; usually there are thousands of observation dimensions in the risk control model. It takes a long time to observe and group each variable, and the model cannot be updated and iterated quickly; because the distribution of variables is not absolutely regular, manual grouping is highly subjective, and the difference in grouping has a greater impact on the modeling effect. Large; the effectiveness of model grouping is based on the fact that the relationship between the variable and the default state does not change. When the corresponding relationship between a variable and the default rate changes, the prediction effect of the model may be poor

Images