CN-121658960-B - Multi-platform e-commerce order aggregation method and device based on dynamic clustering rules

CN121658960BCN 121658960 BCN121658960 BCN 121658960BCN-121658960-B

Abstract

The invention relates to the technical field of e-commerce data processing, in particular to a multi-platform e-commerce order aggregation method and device based on dynamic clustering rules. The technical scheme includes that the method comprises the steps of collecting electronic commerce platform order data, and performing standardized processing to obtain the platform order data. And extracting each cluster characteristic value in the cluster characteristic system. Matching with rules in the dynamic clustering rule set, and determining candidate clustering clusters to which the orders belong. The overall similarity between the orders is calculated. Judging whether the comprehensive similarity is greater than or equal to a preset similarity threshold, classifying the orders into the clustering cluster, and detecting missing orders of the orders which are not clustered. And outputting the order result of the cluster. The clustering precision is obviously improved, the single leakage phenomenon is effectively reduced, and the operation and transportation cost is maximally reduced. The order processing efficiency is remarkably improved, and the requirement of large-scale order processing is met. The system stably operates and simultaneously supports the high-efficiency processing of mass data.

Inventors

JIANG YAPING
WANG KAI
XIE PENGFEI

Assignees

杭州蓝川科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260209

Claims (7)

1. The multi-platform e-commerce order aggregation method based on the dynamic clustering rule is characterized by comprising the following steps of: s1, acquiring order data of an electronic commerce platform to obtain original order data; S2, carrying out standardized processing on the original order data to obtain platform order data; s3, extracting each clustering characteristic value in a clustering characteristic system from platform order data; S4, matching the clustering characteristic value with rules in a dynamic clustering rule set to determine candidate clustering clusters to which orders belong, wherein the dynamic clustering rule set obtaining method comprises the steps of constructing an initial clustering rule set and a dynamic rule updating mechanism, constructing the initial clustering rule set, taking standardized order data as samples, calculating comprehensive similarity among the samples based on the clustering characteristic system, classifying the order with the comprehensive similarity being greater than or equal to a preset similarity threshold as a clustering cluster and determining a core order to generate the initial clustering rule, acquiring order aggregation effect feedback data and business scene change data in real time, triggering rule updating when the order aggregation effect feedback data exceeds the preset threshold or the business scene has great change, adopting a reinforcement learning algorithm, taking the minimized missing rate and the maximized clustering efficiency as optimization targets, adjusting the similarity weight and the similarity threshold of each clustering characteristic, updating the clustering rule set, and simultaneously establishing a rule verification mechanism to perform off-line verification and on-line test points on the updated rules, and enabling formal verification after passing; S5, calculating the comprehensive similarity between the clustering characteristic value and the core order of the candidate cluster; The comprehensive similarity Wherein I is the number of the cluster features in the cluster feature system, I is the total number of the cluster features in the cluster feature system, i=1, 2,.. I, & ω i is the similarity weight of the cluster feature numbered I, c i is the cluster feature value of the cluster feature numbered I, and a i is the influence scaling value of the cluster feature numbered I; s6, judging whether the comprehensive similarity is greater than or equal to a preset similarity threshold, if so, classifying the order into the cluster, otherwise, executing S7; S7, carrying out missing order detection on the orders which are not clustered, judging whether the missing orders are missing, if so, executing S3, carrying out supplementary aggregation treatment on the missing orders, and if not, creating a clustering cluster; S8, outputting order results of the clustering clusters.
2. The method for aggregating multiple platform e-commerce orders based on dynamic clustering rules of claim 1, wherein the normalization process comprises field mapping, data cleansing, correcting abnormal data and/or format unification.
3. The method for multi-platform e-commerce order aggregation based on dynamic clustering rules of claim 1, wherein the clustering features in the clustering feature system comprise a receiving address dimension, a commodity dimension, a user dimension, an order time dimension and a distribution demand dimension.
4. The multi-platform e-commerce order aggregation method based on the dynamic clustering rule of claim 1, wherein the order aggregation effect feedback data comprises a missing order rate, a clustering accuracy rate and a repeated clustering rate.
5. The method for aggregating multiple platform e-commerce orders based on dynamic clustering rules of claim 1, wherein the business scenario change data comprises new platform access, promotional program development, distribution area adjustment.
6. The multi-platform e-commerce order aggregation method based on the dynamic clustering rule of claim 1, wherein the clustering result comprises an order list, the number of orders in a cluster and distribution information summary of each clustering cluster.
7. The utility model provides a multi-platform electronic commerce order polymerization device based on dynamic clustering rule which characterized in that, multi-platform electronic commerce order polymerization device based on dynamic clustering rule includes: the data acquisition and standardization module is used for acquiring original order data of a plurality of e-commerce platforms through a distributed acquisition architecture, and carrying out standardization processing on the original order data to obtain platform order data; the cluster feature construction module is used for constructing a multi-dimensional cluster feature system; the dynamic rule management module is used for constructing an initial clustering rule set, optimizing the clustering rule in real time through a dynamic rule updating mechanism and generating a dynamic clustering rule set; The order clustering execution module is used for extracting a clustering characteristic value from platform order data, matching the clustering characteristic value with a rule in a dynamic clustering rule set, and determining a candidate clustering cluster to which an order belongs; The dynamic clustering rule set obtaining method comprises the steps of constructing an initial clustering rule set and a dynamic rule updating mechanism, constructing the initial clustering rule set, adopting a hierarchical clustering algorithm, taking standardized order data as a sample, calculating comprehensive similarity among samples based on a clustering feature system, classifying orders with the comprehensive similarity being greater than or equal to a preset similarity threshold as a clustering cluster, determining a core order, and generating an initial clustering rule, wherein the dynamic rule updating mechanism comprises the steps of collecting order aggregation effect feedback data and business scene change data in real time, triggering rule updating when the order aggregation effect feedback data exceeds the preset threshold or the business scene has great change, adopting a reinforcement learning algorithm, adjusting the similarity weight and the similarity threshold of each clustering feature by taking 'minimizing single leakage rate' and 'maximizing clustering efficiency' as optimization targets, and updating the clustering rule set; The comprehensive similarity Wherein I is the number of the cluster features in the cluster feature system, I is the total number of the cluster features in the cluster feature system, i=1, 2,.. I, & ω i is the similarity weight of the cluster feature numbered I, c i is the cluster feature value of the cluster feature numbered I, and a i is the influence scaling value of the cluster feature numbered I; The result output and application module is used for outputting the order clustering result and applying the order clustering result to the links of order sorting, inventory allocation and distribution path planning.

Description

Multi-platform e-commerce order aggregation method and device based on dynamic clustering rules Technical Field The invention relates to the technical field of e-commerce data processing, in particular to a multi-platform e-commerce order aggregation method and device based on dynamic clustering rules. Background With the rapid development of the electronic commerce industry, multi-platform operation has become a mainstream mode for expanding market coverage and improving sales volume of electronic commerce enterprises. More and more electronic commerce enterprises select to set up shops on a plurality of platforms for multi-channel operation so as to expand market coverage and promote product sales. The existing multi-platform order aggregation technology is mainly divided into two types, namely a manual summarizing mode, relying on operators to manually collect and classify each platform order, the mode is low in efficiency and easy to cause missing and misplacement due to human errors, when the amount of the order is increased rapidly, the operation cost is increased exponentially, and the other type is a fixed rule aggregation mode, order clustering is realized through a matching rule of a preset single dimension (such as a receiving address and a placing time), the mode cannot adapt to dynamic changes of order data, such as scene of order amount fluctuation during sales promotion activities, field adaptation after new platform access, distribution area adjustment and the like, clustering precision is reduced, a large number of effective orders cannot be aggregated accurately, and accordingly missing of the order occurs. The problem of bill leakage not only can influence customer experience, but also can cause chain reactions such as repeated distribution, stock backlog and the like, and the operation and transportation cost of enterprises is obviously increased. Disclosure of Invention In order to solve the technical problems, the invention provides a multi-platform e-commerce order aggregation method based on dynamic clustering rules, which realizes accurate and real-time aggregation of multi-platform orders by constructing a dynamic clustering rule system, effectively reduces missing orders and reduces operation and transportation costs. The multi-platform e-commerce order aggregation method based on the dynamic clustering rule comprises the following steps: s1, acquiring order data of an electronic commerce platform to obtain original order data. S2, carrying out standardization processing on the original order data to obtain platform order data. S3, extracting each clustering characteristic value in the clustering characteristic system from the platform order data. And S4, matching the clustering characteristic value with a rule in the dynamic clustering rule set, and determining a candidate cluster to which the order belongs. S5, calculating the comprehensive similarity between the clustering characteristic value and the core order of the candidate clustering cluster. S6, judging whether the comprehensive similarity is larger than or equal to a preset similarity threshold, if so, classifying the order into the cluster, and if not, executing S7. And S7, carrying out missing order detection on the orders which are not clustered, judging whether the missing orders exist, and if so, executing S3 and carrying out the aggregation supplementing processing on the missing orders. If not, a cluster is newly built. S8, outputting order results of the clustering clusters. Preferably, the original order data comprises an order number, a user ID, a receiving address, commodity information, order placing time and distribution requirements. Preferably, the standardization process comprises field mapping, data cleaning, abnormal data correction and unified format, so that standardized platform order data can be obtained. The field mapping is to map heterogeneous order fields of different platforms to a preset unified field set. Data cleansing is to reject invalid order data. And correcting the abnormal data. The format unification is to unify the time format, the amount format, the address hierarchy format and the like of the order data as preset standards. Preferably, the clustering features in the clustering feature system comprise static features and dynamic features. Preferably, the static features include a shipping address dimension, a merchandise dimension, a user dimension, and the like. Preferably, the dynamic characteristics include a time dimension, a delivery demand dimension, and the like. Preferably, the dynamic clustering rule set obtaining method comprises the steps of initial clustering rule set construction and dynamic rule updating mechanism. The method for constructing the initial clustering rule set comprises the steps of adopting a hierarchical clustering algorithm, taking standardized order data as samples, calculating the comprehensive similarity among the samples based on a clustering feat