Credit factory order scheduling method and device based on multi-agent reinforcement learning

Abstract

The invention relates to a credit factory order scheduling method and device based on multi-agent reinforcement learning. The method comprises the following steps that multiple agents send actions toan environment; the environment calculates the sharing state of the multiple agents and the rewards of the agents; the environment sends the sharing state of the multiple agents and the rewards of theagents to the corresponding agents; each intelligent agent selects an action according to the received state and the award; wherein the multi-agent sequentially processes the credit factory order. According to the invention, the order approval time can be shortened, and intelligent scheduling management of a credit factory is realized; a credit factory can formulate a scientific and reasonable scheduling scheme in a dynamic environment, and the anti-interference capability of order scheduling in the credit factory is greatly enhanced.

Description

technical field [0001] The invention relates to the field of big data processing, in particular to a credit factory order scheduling method and device based on multi-agent reinforcement learning. Background technique [0002] In recent years, consumer credit has achieved great success in China. Consumer finance companies need to approve the loan order submitted by the customer, and then determine whether to grant the loan. Credit Factory is an important means to speed up loan approval. The characteristics of consumer credit are small amount and high frequency. Considering these characteristics of consumer credit, many companies have introduced the credit factory model. Credit Factory is a new loan approval model invented by Singapore Temasek Group. The credit factory model has been adopted by several financial institutions including Bank of China, China Construction Bank and China Merchants Bank. Similar to a factory assembly line, the credit factory divides credit appr...

AI Technical Summary

Problems solved by technology

In the prior art, multi-agent reinforcement learning (MARL) is used to solve the shop-floor scheduling problem. This algorithm realizes decentralized scheduling and can be used in unknown situations without retraining. However, the arrival time and processing time of each job in this work are known, compared to the more complex scheduling of credit factory orders where arrival and processing times are unknown

[0005] Although there have been many works on the application of MARL algorithm in order scheduling problems, there is no mature technical solution for large-scale real-time order scheduling problems with the characteristics of multi-machine, multi-process, random arrival time and processing time.

Images