Consumption credit scene risk assessment method based on random forest algorithm

Abstract

The invention discloses a risk assessment method for a consumption credit scene based on a random forest algorithm. The risk assessment method comprises an information acquisition module, a data preprocessing module, a feature engineering module, a model training and parameter adjustment module, a feature importance assessment module, a model evaluation and selection module and a model deploymentmonitoring module. The method has the advantages that data preprocessing is simple, feature engineering and model training efficiency is high, model accuracy is high, a random forest model is combinedwith a consumption credit scene of Internet finance, a random forest can better solve the problems of high-dimensional sparsity, much noise and variable redundancy of Internet data according to algorithm superiority of the random forest, and the method is suitable for popularization and application. Compared with other traditional score card models, the random forest model has higher risk prediction accuracy and stability, improves the recognition of credit risks, and has a certain reference value for the actual application of consumption credit of Internet finance.

Description

technical field [0001] The present invention relates to the technical field of risk control in the internet financial consumer credit industry, in particular to a risk assessment method for consumer credit scenarios based on a random forest algorithm. Background technique [0002] With the rise of the Internet + concept, Internet financial consumer credit companies represented by P2P lending, consumer finance, and car leasing have sprung up. However, after the savage growth, the company's development speed and vitality are all focused on risk control. Traditional risk control audits are scorecard models based on machine learning algorithms, including logistic regression, decision trees, support vector machines, and neural networks. Weights can also easily absorb new data to update the model, so when integrated algorithms such as GBDT, random forest, and lightGBM appear one after another, the traditional scorecard model is still a common method for risk assessment in the cons...

AI Technical Summary

Problems solved by technology

[0006] Insufficient model accuracy: The algorithm of the traditional scorecard model is a weak classifier. Compared with the integrated algorithm model based on the combination of multiple weak classifiers, the single model of the traditional scorecard lacks exploration and verification between models, and may have insufficient model stability. and the problem of weak generalization ability

Images