Search

CN-121981814-A - Large-scale object matching method, device and storage medium

CN121981814ACN 121981814 ACN121981814 ACN 121981814ACN-121981814-A

Abstract

The embodiment of the application provides a large-scale object matching method, equipment and a storage medium, which relate to the technical field of artificial intelligence, and the method comprises the steps of acquiring first attribute information of target matching objects and second attribute information of objects to be matched aiming at each target matching object in each round of matching process, wherein the objects to be matched are used for matching with the target matching objects; for any object to be matched, when the object to be matched is changed from an original matching object to a target matching object, determining whether the third attribute information of the original matching object after the change and the first attribute information after the change meet respective constraint requirements, if so, moving the object to be matched into the target matching object, after the matching process of each round is finished, determining global deviation of the target matching object in each round based on the change condition of the constraint objects in the constraint requirements of the target matching object, and if the global deviation of at least two rounds tends to be converged, stopping moving the object to be matched into the target matching object.

Inventors

  • LIU NAIHAO

Assignees

  • 深圳前海微众银行股份有限公司

Dates

Publication Date
20260505
Application Date
20260104

Claims (10)

  1. 1. A method of large-scale object matching, comprising: for each target matching object in each round of matching process, acquiring first attribute information of the target matching object and second attribute information of each object to be matched, wherein the object to be matched is used for matching with the target matching object; For any object to be matched, when the object to be matched is changed from an original matching object to the target matching object, determining whether the third attribute information of the original matching object after the change and the first attribute information after the change meet respective constraint requirements; after the matching process of each round is finished, determining the global deviation of the target matching object in each round based on the change condition of the constraint object in the constraint requirement of the target matching object; and if the global deviation of at least two rounds tends to be converged, stopping migrating the object to be matched to the target matching object.
  2. 2. The method as recited in claim 1, further comprising: Determining constraint requirements of each target matching object in each round, wherein the constraint requirements comprise hard constraints and soft constraints, and each constraint object can be set to be any one of the hard constraints or the soft constraints; The constraint requirements support hot plug modification.
  3. 3. The method of claim 1, further comprising, prior to said determining whether the third attribute information of the post-change original matching object and the post-change first attribute information meet respective constraint requirements: for each target matching object, determining state information of the target matching object in the round of matching process according to first attribute information of the target matching object; And determining a matching strategy of the target matching object to be matched in the round of matching process based on the state information of the target matching object, wherein the matching strategy comprises the number of the objects to be matched.
  4. 4. The method of claim 1, further comprising, prior to said determining whether the third attribute information of the post-change original matching object and the post-change first attribute information meet respective constraint requirements: and determining the matching degree of the object to be matched and the target matching object, wherein the matching degree is used for determining the priority of matching the object to be matched and the target matching object.
  5. 5. The method of claim 4, wherein the determining the degree of matching of the object to be matched with the target matching object comprises: Determining weight information of a plurality of matching dimensions, wherein the weight information is determined according to the change condition of a constraint object in the previous round of matching process; And determining the matching degree of the object to be matched and the target matching object according to the matching results of the matching dimensions and the weight information of each matching dimension.
  6. 6. The method according to any one of claims 1 to 5, comprising: If the global deviation of at least two rounds does not tend to converge, determining the stability of the target matching object based on at least one constraint object in the constraint requirements of the target matching object; And adjusting constraint requirements of the target matching object according to the stability of the target matching object and continuing the subsequent round matching process.
  7. 7. The method according to any one of claims 1 to 5, wherein determining whether the third attribute information of the changed original matching object and the changed first attribute information meet respective constraint requirements includes: Determining whether the third attribute information of the original matching object after the change meets the constraint requirement of the original matching object and whether the first attribute information after the change meets the constraint requirement of the target matching object according to the attribute information related in the respective constraint requirements; After each target matching object is matched, carrying out overall statistical calculation on the matching result of each object to be matched.
  8. 8. The method of any one of claims 1 to 5, further comprising: And recording the matching basis for matching the object to be matched with the target matching object in each round of matching process, and forming a traceable log.
  9. 9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1-8 when the program is executed.
  10. 10. A computer readable storage medium, characterized in that it stores a computer program executable by a computer device, which program, when run on the computer device, causes the computer device to perform the steps of the method according to any of claims 1-8.

Description

Large-scale object matching method, device and storage medium Technical Field The present invention relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, and a storage medium for large-scale object matching. Background In the scene of needing to carry out large-scale object matching, at present, the solution complexity is increased sharply by solving a linear or mixed integer programming model, and even the problem of local unbalance occurs. Taking the example that loan users are distributed to a plurality of banks according to an agreement, each loan user performs loan by the bank specified in the agreement according to the bank correspondingly distributed by the agreement. As the number of loan subscribers increases, large-scale loan subscribers need to match the appropriate target banks, even in real-time. And calculating the loan users to be allocated to each bank according to the historical loan conditions of the loan users-banks, so that the loan users are reasonably allocated to the target banks. However, the allocation method is relatively applicable to data sets of small and medium sizes, but when facing massive loan users, the complexity of solving the linear programming model is increased sharply, so that the result of matching between the loan users and banks is unstable, and the problem of local unbalance or poor expansibility occurs. Disclosure of Invention The embodiment of the application provides a large-scale object matching method, equipment and a storage medium, which are used for improving the matching efficiency during large-scale object matching. In a first aspect, an embodiment of the present application provides a method for large-scale object matching, including: for each target matching object in each round of matching process, acquiring first attribute information of the target matching object and second attribute information of each object to be matched, wherein the object to be matched is used for matching with the target matching object; For any object to be matched, when the object to be matched is changed from an original matching object to the target matching object, determining whether the third attribute information of the original matching object after the change and the first attribute information after the change meet respective constraint requirements; after the matching process of each round is finished, determining the global deviation of the target matching object in each round based on the change condition of the constraint object in the constraint requirement of the target matching object; and if the global deviation of at least two rounds tends to be converged, stopping migrating the object to be matched to the target matching object. In the embodiment of the application, the capacity distribution of the group of the objects to be matched is fitted on a macroscopic level, and the capacity potential of the single object to be matched is accurately depicted on a microscopic level, and the mechanism enables the capacity prediction to have self-adaptive characteristics, namely, when the bank scale is enlarged, the model automatically relaxes the upper limit of the capacity and smoothes the client migration into the objects to be matched, thereby reducing the phenomenon of 'high capacity client concentration', and improving the overall stability of the system. The heuristic iterative optimization method is used for greatly reducing the calculation complexity through a local decision-making and multi-round balance correction mechanism, and realizing the rapid distribution of large-scale clients. Optionally, further comprising; Determining constraint requirements of each target matching object in each round, wherein the constraint requirements comprise hard constraints and soft constraints, and each constraint object can be set to be any one of the hard constraints or the soft constraints; The constraint requirements support hot plug modification. In the embodiment of the application, new constraints (such as region matching, client priority and quota adjustment factors) do not need to be added for re-modeling and re-solving, so that the cost is reduced, and the flexibility is high. Optionally, before determining whether the third attribute information of the original matching object after the change and the first attribute information after the change meet respective constraint requirements, the method further includes: For each target matching object, determining state information of the target object in the round of matching process according to first attribute information of the target matching object; And determining a matching strategy of the target matching object to be matched in the round of matching process based on the state information of the target matching object, wherein the matching strategy comprises the number of the objects to be matched. In the embodiment of the application, the number of the mat