Automated decision platform

Patent metadata

Field	Value
Publication number	US-2021350262-A1
Application number	US-202117314810-A
Country	US
Kind code	A1
Filing date	May 7, 2021
Priority date	May 8, 2020
Publication date	Nov 11, 2021
Grant date	—

Abstract

A computer system receives a set of user-defined rules are that useable by a computer service to automate a decision flow. The computer system generates a graph model from the user-defined rules. From the graph model, the computer system determines an input dependency model that is indicative of a set of inputs referred to in the graph model. The input dependency model is useable by an orchestrator to coordinate accesses to the one or more data stores in which the set of inputs is stored. The computer system receives the set of inputs and determines one or more automated decisions by applying the set of inputs to the graph model.

First claim

What is claimed is:

1 . A method comprising:
receiving, at a computer system, a set of user-defined rules, wherein the user-defined rules are useable by a computer service to automate a decision flow; generating, with the computer system, a graph model from the set of user-defined rules; determining, by the computer system from the graph model, an input dependency model that is indicative of a set of inputs referred to in the graph model, wherein the input dependency model is useable by an orchestrator to coordinate accesses to one or more data stores in which the set of inputs is stored; receiving, by the computer system from the one or more data stores, the set of inputs; and determining, with the computer system, one or more automated decisions by applying the set of inputs to the graph model.

2 . The method of claim 1 , wherein the generating and the determining do not include compiling the user-defined rules.

3 . The method of claim 1 , further comprising:
receiving, at the computer system, a set of user-defined functions useable by the computer service to supplement the user-defined rules; wherein the graph model is generated from the set of user-defined rules and user-defined functions.

4 . The method of claim 1 , further comprising validating the set of user-defined rules, wherein the validating includes validating dependencies referred to by the set of user-defined rules.

5 . The method of claim 1 ,
wherein the input dependency model indicates, for each individual input of the set of inputs:
a name of the individual input,
a namespace of the individual input, and
a datatype of the individual input.

6 . The method of claim 1 , wherein the input dependency model is useable by the orchestrator to determine one or more of: an order in which to perform the accesses, which of the accesses can be performed in parallel, or how long the accesses will take to perform.

7 . The method of claim 1 , wherein receiving the set of inputs and determining one or more automated decisions is performed in a production computing environment, the method further comprising, replacing, at the production computing environment, the graph model with a second graph model for subsequent automated decision determination, wherein operations related to the replacing include:
receiving, at the computer system, a second set of user-defined rules; validating, with the computer system, the second set of user-defined rules; and generating, with the computer system, the second graph model from the second set of user-defined rules.

8 . The method of claim 7 , wherein the replacing does not include compiling the second set of user-defined rules.

9 . The method of claim 1 , wherein the set of user-defined rules are defined in a hierarchy including one or more rule packages at a first level of the hierarchy, and wherein each rule package includes one or more individual rules at a second level of the hierarchy, wherein there is no dependency between individual rules in the same rule package.

10 . The method of claim 1 , wherein the graph model is a directed acyclic graph model.

11 . The method of claim 1 , wherein the one or more automated decisions include a prediction of whether a particular transaction request is fraudulent.

12 . The method of claim 1 , wherein the one or more automated decisions include determining a recommendation for a product or service.

13 . A non-transitory, computer-readable medium storing instructions that when executed by a computer system cause the computer system to perform operations comprising:
receiving, at a computer system, a set of user-defined rules, wherein the user-defined rules are useable by a computer service to automate a decision flow; receiving, at the computer system, a set of user-defined functions useable by the computer service to supplement the user-defined rules; generating, with the computer system, a graph model from the set of user-defined rules and user-defined functions; and determining, by the computer system from the graph model, an input dependency model that is indicative of a set of inputs referred to in the graph model, wherein the input dependency model is useable to an orchestrator to coordinate accesses to one or more data stores in which the set of inputs is stored; wherein one or more automated decisions are determinable by applying the set of inputs to the graph model.

14 . The non-transitory, computer-readable medium of 13 ,
wherein the set of user-defined rules are input by a user as structured text; and wherein generating the graph model from the set of user-defined rules includes parsing the structured text to identify indications of one or more inputs, one or more conditions, and one or more actions.

15 . The non-transitory, computer-readable medium of 13 , wherein the generating and the determining include neither compiling the user-defined rules nor compiling the user-defined functions.

16 . A method comprising:
receiving, at a multi-tenant computer system, a set of user-defined rules, wherein the user-defined rules are useable by a computer service to automate a decision flow; and performing, by an instance of an automated decision platform running on the multi-tenant computer system, an automated decision flow using the set of user-defined rules, wherein the performing includes:
generating, with the automated decision platform, a graph model from the set of user-defined rules;
receiving, with the automated decision platform, a set of inputs referred to in the graph model; and
determining, with the automated decision platform, one or more automated decisions by applying the set of inputs to the graph model.

17 . The method of claim 16 , wherein a plurality of nodes of the graph model take as input one or more respective ones of the set of inputs, and wherein applying the set of inputs to the graph model includes inputting the respective ones of the set of inputs to the corresponding nodes.

18 . The method of claim 16 , further comprising:
performing, by a second instance of the automated decision platform running on the multi-tenant computer system, a second automated decision flow using a second set of user-defined rules.

19 . The method of claim 16 , wherein the set of user-defined rules was input to a third instance of the automated decision platform running on a user device, and wherein the set of user-defined rules was validated by the third instance of the automated decision platform prior to the receiving.

20 . The method of claim 19 , wherein the multi-tenant computer system and the user device run different operating systems.

Assignees

• Paypal Inc

Inventors

• Song Lin
• Li Xin
• Ho Chun Kiat
• Wu Yuehao
• Xie Kai
• Zang Meng
• Rathi Chinmayee
• Yang Jun
• Zheng Ke
• Yang Zhenyin
• Lu Wen

Classifications

• G06Q20/4016

  involving fraud or risk level assessment in transaction processing · CPC title

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title