Search

CN-122003695-A - Method for evaluating users of a market system by means of a graphical machine learning model

CN122003695ACN 122003695 ACN122003695 ACN 122003695ACN-122003695-A

Abstract

Aspects relate to a method for evaluating users of a marketplace system, including generating training data elements for a graph machine learning model, wherein each training data element includes a graph including a plurality of user nodes, each user node being associated with a user and a user node feature, the user node feature being generated from historical data up to a historical transaction value of a user of a predetermined training date, and for at least one user node, the training data elements including labels generated from the historical transaction value of the user after the predetermined training date in the historical data, training the graph machine learning model to predict labels of the training data elements from the corresponding graphs of the training data elements, generating a graph including nodes of the user for the user of the marketplace system to be evaluated, wherein the user node feature includes a user node feature including the historical transaction value of the user, and predicting the user's value by processing the graph generated for the user using the trained graph machine learning model.

Inventors

  • CHEN MIN
  • HUANG XUEFANG
  • CHEN JIA

Assignees

  • 格步计程车控股私人有限公司

Dates

Publication Date
20260508
Application Date
20240913
Priority Date
20231013

Claims (16)

  1. 1. A method for evaluating users of a marketplace system, comprising: Generating training data elements for a graph machine learning model, wherein each training data element comprises a graph comprising a plurality of user nodes, each user node being associated with a user and a user node feature, the user node feature being generated from historical data up to a historical transaction value of the user for a predetermined training date, and for at least one user node, the training data element comprises a tag, the tag being generated from the historical transaction value of the user in the historical data after the predetermined training date; training the graph machine learning model to predict labels of the training data elements from their respective graphs; Generating a graph comprising nodes of the user for the user of the marketplace system to be evaluated, wherein the nodes of the user comprise user node characteristics including historical transaction values of the user, and The value of the user is predicted by processing the generated graph for the user using a trained graph machine learning model.
  2. 2. The method of claim 1, wherein, for each training data element and for users of the marketplace system to be evaluated, the respective graph includes respective marketplace system nodes, each marketplace system node being associated with a respective additional marketplace system entity.
  3. 3. The method of claim 2, further comprising predicting a value of one or more of the additional market system entities from the predicted value of the user.
  4. 4. A method according to any of claims 1 to 3, wherein at least some of the market system nodes are nodes of various featureless node types, and the graph machine learning model includes a neural network to generate an embedding for each featureless node type, the neural network being trained in the training of the graph machine learning model, the embedding being used as a feature of featureless nodes of featureless node types when processing a graph by the graph machine learning model.
  5. 5. The method of claim 4, wherein the featureless node types include one or more of internet protocol address nodes, geographic address nodes, user equipment nodes, and network nodes.
  6. 6. The method of any of claims 1-5, wherein, for each training data element and for a user of the marketplace system to be evaluated, the respective graph includes respective edges, wherein each of the edges represents interactions between additional marketplace entities associated with nodes connected by edges, interactions between additional marketplace entities associated with one of the nodes connected by edges and a user associated with another one of the nodes connected by edges, or interactions between users associated with the nodes connected by edges.
  7. 7. The method of claim 6, wherein each of at least some of the edges represents a respective transaction by a respective user associated with a node to which the edge is connected.
  8. 8. The method of claim 7, wherein edges connect the respective user with a respective product or service and represent purchase of the product or use of the service by the respective user, respectively.
  9. 9. The method of claim 8, wherein an edge has an edge characteristic determined from a value of the transaction.
  10. 10. The method of any of claims 1 to 9, wherein the graph machine learning model is a graph roll-up neural network.
  11. 11. The method of any of claims 1-10, wherein the user node characteristics include one or more of an age of an account of the user with the marketplace system, a number of transactions completed by the user within the marketplace system, a total revenue of the marketplace system by the user, and a total profit of the marketplace system by the user.
  12. 12. The method of any one of claims 1 to 11, wherein the marketplace system is an on-line to off-line marketplace system.
  13. 13. The method of any of claims 1-12, wherein each value indicates a value of a profit of a corresponding market system entity to an operator of the market system.
  14. 14. A server computer comprising a wireless interface, a memory interface and a processing unit configured to perform the method of any of claims 1 to 13.
  15. 15. A computer program element comprising program instructions which, when executed by one or more processors, cause the one or more processors to perform the method of any of claims 1 to 13.
  16. 16. A computer readable medium comprising program instructions which, when executed by one or more processors, cause the one or more processors to perform the method of any of claims 1 to 13.

Description

Method for evaluating users of a market system by means of a graphical machine learning model Technical Field Aspects of the present disclosure relate to methods for evaluating users of a marketplace system. Background Understanding the value of the user is critical to the successful operation of the marketplace system. High value users typically generate most of the revenue for the system and are thus highly important to the operation of the marketplace system. By knowing the value of the user, the marketplace system is also able to recommend the correct product or service. Unlike e-commerce or social media platforms that focus primarily on user value, online-to-offline (O2O) services such as taxi taking services and food (or other product) delivery involve multiple entities. They acquire users and perform transactions using online and offline channels. For example, in a food delivery service, users ordering food online require offline riders to deliver food to their locations. Thus, the value of the user, rider, food merchant, delivery address are all key factors in the success of the food delivery service. Accordingly, efficient methods suitable for online to offline (O2O) traffic for evaluating users of a marketplace system are desired. Disclosure of Invention Various embodiments relate to a method for evaluating users of a marketplace system, including generating training data elements for a graph machine learning model, wherein each training data element includes a graph including a plurality of user nodes, each user node being associated with a user and a user node feature, the user node feature being generated from historical data up to a historical transaction value of a user of a predetermined training date, and for at least one user node, the training data elements including labels generated from the historical transaction value of the user after the predetermined training date in the historical data, training the graph machine learning model to predict labels of the training data elements from corresponding graphs of the training data elements, generating a graph including nodes of the user for the user of the marketplace system to be evaluated, wherein the user node feature includes a user node feature, the user node feature including the historical transaction value of the user, and predicting the value of the user by processing the graph generated for the user using the trained graph machine learning model. According to one embodiment, for each training data element and for users of the market system to be evaluated, the respective graph includes respective market system nodes, each market system node being associated with a respective further market system entity. According to one embodiment, the method further comprises predicting a value of one or more of the further market system entities from the predicted value of the user. According to one embodiment, at least some of the market system nodes are nodes of various featureless node types, and the graph machine learning model includes a neural network to generate an embedding for each featureless node type, the neural network being trained in the training of the graph machine learning model, the embedding being used as a feature of the featureless node type when the graph is processed by the graph machine learning model. According to one embodiment, the featureless node types include one or more of internet protocol address nodes, geographical address nodes, user equipment nodes, and network nodes. According to one embodiment, for each training data element and for the users of the market system to be evaluated, the respective graph comprises respective edges, wherein each edge represents interactions between further market entities associated with nodes connected by an edge, interactions between further market entities associated with one of the nodes connected by an edge and users associated with another of the nodes connected by an edge, or interactions between users associated with the nodes connected by an edge. According to one embodiment, each of at least some of the edges represents a respective transaction by a respective user associated with the node to which the edge is connected. According to one embodiment, the edges connect the respective user with the respective product or service and represent the purchase of the product or use of the service by the respective user, respectively. According to one embodiment, the edges have edge characteristics that are determined based on the value of the transaction. According to one embodiment, the graph machine learning model is a graph neural network (e.g., a graph convolution neural network). According to one embodiment, the user node characteristics include one or more of an age of an account of the user in the marketplace system, a number of transactions completed by the user within the marketplace system, a total revenue of the marketplace system by the user, and a total profit of t