Multi-sampling model training method and device

Abstract

The invention discloses a multi-sampling model training method and device. The training method comprises the steps of firstly conducting multi-sampling on all samples to obtain a training set and a verification set in each time of sampling; then using the training set and the verification set obtained in each time of sampling as a group, and adopting the training set in each time of sampling to conduct model training; adopting the model obtained by training to evaluate the training set and the verification set separately, and according to an evaluation result of the training set and the verification set and a set elimination criterion, eliminating a model obtained by training; finally adopting the kept model to predict all the samples, and adopting a result obtained by prediction to conduct model training on the kept model to obtain a final model. The multi-sampling model training model comprises a sampling module, a first training module, an evaluation module and a second training module. The model obtained through the multi-sampling-model training method and device has higher robustness and stability, the prediction precision is higher, and the modeling efficiency is drastically improved.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to a multi-sampling model training method and device. Background technique [0002] "Huabei" is an online shopping service provided by Ant Financial under Ant Financial Services Group to consumers on Taobao and Tmall to "buy this month and pay back next month". The minimum amount is 1,000 yuan and the maximum is 50,000 yuan. Since "Huabei" has the same function as a credit card - pay first after consumption, this creates room for cash out. [0003] In the business scenario of the "Huabei" anti-cash model, it is necessary to use machine learning algorithms to classify or perform regression calculations on data. Among them, the quality and quantity of training samples will have a direct impact on the prediction effect of the model. On the other hand, the anti-cash model is a model developed for "Huabei" sellers, that is, one seller corresponds to one sample, so ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a multi-sampling model training method and device to solve the problem of poor model training effect in the prior art due to limited data sample size

Images