Search

CN-121998700-A - Cross-border E-commerce decision method and system based on full link

CN121998700ACN 121998700 ACN121998700 ACN 121998700ACN-121998700-A

Abstract

The invention relates to a cross-border e-commerce decision method and system based on a full link, wherein the method comprises the steps of obtaining multi-source heterogeneous data related to commodities to generate structured input data, analyzing and processing the structured input data by a pre-constructed analysis model to construct and maintain operation state vectors corresponding to the commodities, analyzing the structured input data for cross-border e-commerce business to update the operation state vectors, monitoring changes of the operation state vectors to generate corresponding state events, respectively executing content optimization operation, inventory decision operation and/or analysis operation corresponding to the commodities based on the state events, feeding back execution results to the analysis model to update the operation state vectors, generating operation monitoring results based on the updated operation state vectors, and triggering the state events again when abnormal changes are detected. By adopting the scheme, the comprehensive depiction and continuous monitoring of the commodity operation state can be realized, and the integrity and accuracy of the cross-border e-commerce decision can be improved.

Inventors

  • CHEN LEI
  • CHEN ZHIYI

Assignees

  • 深圳市易仓科技有限公司

Dates

Publication Date
20260508
Application Date
20260126

Claims (10)

  1. 1. The cross-border e-commerce decision method based on the full link is characterized by comprising the following steps of: Acquiring multi-source heterogeneous data related to commodities, and performing entity alignment and time sequence synchronization on the multi-source heterogeneous data based on a commodity unique identifier to generate structured input data; Analyzing and processing the structured input data by a pre-constructed analysis model, and constructing and maintaining an operation state vector corresponding to the commodity, wherein the operation state vector is used for representing the comprehensive operation state of the commodity in the flow, conversion, public praise, inventory and profit dimensions; Performing cross-border e-commerce business oriented analysis on the structured input data, generating operation analysis results by the analysis model, and updating the operation state vector based on the operation analysis results; Monitoring the change of the operation state vector, and generating a corresponding state event through the analysis model when the operation state vector meets a preset trigger condition; Based on the state event, respectively executing content optimization operation, inventory decision operation and/or analysis operation corresponding to the commodity, and feeding back an execution result to the analysis model to update the operation state vector; And generating an operation monitoring result based on the updated operation state vector, and triggering the state event again when detecting abnormal change.
  2. 2. The decision method of claim 1, wherein the generating structured input data comprises: acquiring commodity data from a plurality of different data platforms, wherein the commodity data comprises commodity image data, commodity text description data, user comment data and user behavior data; Respectively carrying out field semantic analysis on commodity data of different data platforms to generate a platform semantic field set; establishing a cross-platform field mapping relation based on the platform semantic field set, and carrying out field reconstruction on the commodity data; And carrying out entity consistency check on the commodity data with the field reconstructed based on the commodity unique identifier, carrying out time sequence consistency check on the commodity data passing through the entity consistency check based on the time stamp, and taking the commodity data simultaneously meeting the entity consistency check and the time sequence consistency check as the structured input data.
  3. 3. The decision method of claim 1, wherein the constructing and maintaining a business state vector corresponding to the commodity comprises: Respectively constructing a flow state feature set, a conversion state feature set, a public praise state feature set and an inventory state feature set based on the structured input data; performing feature compression processing on each state feature set to generate corresponding state sub-vectors; Assigning an initial weight to each of the state subvectors based on the stability performance of the structured input data over different time periods; Based on the prediction error of the operation state vector in the history period, carrying out self-adaptive adjustment on the initial weight; Each of the state sub-vectors of adjusted weights is combined to generate the business state vector.
  4. 4. The decision method of claim 1, wherein the performing cross-border e-commerce business oriented analysis on the structured input data, generating operational analysis results from the analysis model, and updating the operational status vector based on the operational analysis results, comprises: analyzing the commodity image data in the structured input data, and extracting a commodity display characteristic set; Carrying out semantic analysis on the user comment data in the structured input data, and extracting a user demand feature set; Constructing a feature pair set based on the commodity display feature set and the user demand feature set; performing consistency calculation on the feature pair set to generate an initial matching result; And carrying out reverse verification on the initial matching result based on the historical conversion data to obtain a corrected matching result, and generating the operation analysis result based on the corrected matching result.
  5. 5. The decision-making method according to claim 4, wherein one or more of the following is satisfied: The method comprises the steps of carrying out consistency calculation on a feature pair set to generate an initial matching result, wherein the initial matching result comprises the steps of mapping the commodity display feature and the user demand feature to a unified semantic space, calculating a semantic consistency score between the commodity display feature and the user demand feature based on the unified semantic space, carrying out comparison on a consistency calculation result of a competitive commodity based on a detection result of a functional structure in commodity image data and functional parameter information in commodity text description, calculating an evidence consistency score between the commodity display feature and the user demand feature, calculating a conflict penalty factor of a corresponding feature pair based on emotion polarity distribution and divergence degree of the user demand feature in a historical user comment, calculating a scene consistency score based on a matching relation between using scene information corresponding to the user demand feature and scene elements in the commodity display feature, and carrying out comparison on the competitive product matching result and the consistency calculation result of the feature pair to generate a relative consistency score; The method comprises the steps of carrying out reverse verification on an initial matching result based on historical conversion data to obtain a corrected matching result, constructing a verification data set containing the initial matching result, the historical conversion data and page change information based on a time dimension, carrying out time alignment on the change of the initial matching result and the change of the historical conversion data, constructing a conversion prediction model based on the verification data set, carrying out deviation calculation on the predicted conversion result and actual historical conversion data to generate a matching effectiveness evaluation parameter, carrying out weighted update on the initial matching result based on the matching effectiveness evaluation parameter, carrying out weight reduction processing on the corresponding initial matching result when the matching effectiveness evaluation parameter represents negative influence on conversion caused by the corresponding initial matching result, and taking the updated matching result as the corrected matching result.
  6. 6. The decision method of claim 1, wherein the monitoring the change of the business state vector, when the business state vector meets a preset trigger condition, generating a corresponding state event through the analysis model, comprises: periodically sampling the operation state vector in a preset time window to generate a historical state sequence; carrying out trend modeling on the historical state sequence to obtain state change trend parameters; Calculating a trend deviation degree based on the state change trend parameter; And comparing the trend deviation with a historical fluctuation threshold interval, and generating the state event when the trend deviation exceeds the historical fluctuation threshold interval.
  7. 7. The decision method of claim 6, wherein one or more of the following is satisfied: The trend modeling is carried out on the historical state sequence to obtain state change trend parameters, which comprises the steps of collecting operation state vectors at a plurality of time points to form corresponding historical state sequences, and executing time alignment, missing value restoration and abnormal value cleaning processing on the historical state sequences; The trend deviation degree is calculated based on the state change trend parameters, the trend deviation degree comprises the steps of carrying out root cause inference on the candidate abnormal set based on a root cause map describing causal relations among all state dimensions in an operation state vector, determining potential root causes corresponding to all candidate abnormal sets and confidence degrees of the potential root causes, generating gating parameters based on stock saleable states, page content change states and advertisement delivery states, judging whether the potential root causes have executable performance or not according to the potential root causes and the confidence degrees of the potential root causes, generating corresponding state events when the potential root causes meet preset gating conditions, grading the state events based on the influence degrees of the potential root causes and the confidence degrees, determining corresponding execution strategy types according to the grades of the state events, arranging the state events and other state events according to preset dependent relations after the state events are generated, forming event execution sequences, combining the state events of the same type within a preset cooling time window, inhibiting triggering of the state events when the correction results of the operation state vector indicate that the abnormal states are relieved, generating corresponding state events after the state strategies are executed, grading the corresponding state events based on the corresponding state vectors, and carrying out a decision-making a decision again when the operation state vector is improved, and the operation state vector is estimated again based on the expected state vectors.
  8. 8. Decision method according to claim 1, characterized in that said executing content optimization operations, inventory decision operations and/or re-analysis operations corresponding to said commodity, respectively, based on said status event, and feeding back execution results to said analysis model to update said business status vector, comprises: Generating an execution constraint parameter set based on the current business state vector, wherein the execution constraint parameter set is used for constraining the execution range and the adjustment amplitude of the content optimizing operation, the inventory decision operation and/or the analysis operation; Determining an adjustable interval of a corresponding operation based on the execution constraint parameter set, wherein the operation adjustable interval is used for limiting an adjustment range of content parameters, adjustment amplitude of inventory decision and/or triggering conditions of a re-analysis operation; In the operation adjustable interval, performing content optimization operation, inventory decision operation and/or analysis operation corresponding to the commodity so as to realize targeted adjustment of commodity operation state; Recording an execution result of the operation and corresponding operation state change information, and inputting the execution result and the state change information as feedback data into the analysis model for updating the operation state vector to provide basis for identification and operation decision of subsequent state events.
  9. 9. The decision method of claim 8, wherein generating a business monitoring result based on the updated business state vector and re-triggering the state event upon detection of an abnormal change comprises: generating an operation monitoring result based on the current operation state vector, and detecting whether the operation state is abnormally changed or not according to the operation monitoring result; Generating a state event and entering a corresponding decision flow in response to the detection of the abnormal change, so as to obtain an execution result for the state event; Mapping the execution result into a state correction quantity for a corresponding state dimension in the management state vector; Updating the operation state vector based on the state correction amount to obtain an updated operation state vector; Recalculating state change trend parameters reflecting the operation state change trend based on the updated operation state vector; And responding to the state change trend parameter meeting a re-triggering condition, judging that the abnormal change is not eliminated or further evolved, and re-generating a state event to enter the next round of decision flow.
  10. 10. A full link based cross-border e-commerce decision system comprising an analysis model adapted to perform an operation decision using the full link based cross-border e-commerce decision method of any one of claims 1 to 9.

Description

Cross-border E-commerce decision method and system based on full link Technical Field The invention relates to the technical field of electronic commerce data analysis, in particular to a cross-border electronic commerce decision method and system based on a full link. Background With the rapid development of cross-border e-commerce business, it has become normal for goods to be operated simultaneously in multiple e-commerce platforms and multiple countries and regions. The display form, the user feedback content, the user behavior characteristics and the inventory and logistics state of the commodity show the characteristics of various data sources, complex data structures, high change frequency and the like. In the existing scheme, aiming at the operation decision of cross-border electronic commerce commodities, commodity flow, conversion rate, user evaluation, inventory level and other indexes are generally monitored in a scattered mode, and commodity page content, inventory replenishment or marketing strategies are adjusted based on preset rules or manual experience. However, most of the technical schemes depend on fixed rules or manually set threshold conditions, and it is difficult to uniformly model complex association relations among multidimensional data. Therefore, how to uniformly model commodity operation states in a multi-source heterogeneous data environment and drive content optimization, inventory decision and continuous monitoring of cross-border e-commerce commodities based on analysis results of complex data relationships becomes a technical problem to be solved urgently by those skilled in the art. Disclosure of Invention Based on the method and the system, the invention provides the cross-border e-commerce decision method and the system based on the full link, which can realize the comprehensive depiction and continuous monitoring of commodity operation states and improve the integrity and the accuracy of the cross-border e-commerce decision. A cross-border e-commerce decision-making method based on full links comprises the steps of obtaining multi-source heterogeneous data related to commodities, carrying out entity alignment and time sequence synchronization on the multi-source heterogeneous data based on commodity unique identification to generate structured input data, carrying out analysis processing on the structured input data by a pre-built analysis model, constructing and maintaining operation state vectors corresponding to the commodities, wherein the operation state vectors are used for representing comprehensive operation states of the commodities in flow, conversion, public praise, inventory and profit dimensions, carrying out cross-border e-commerce business oriented analysis on the structured input data, generating operation analysis results by the analysis model, updating the operation state vectors based on the operation analysis results, monitoring changes of the operation state vectors, generating corresponding state events by the analysis model when the operation state vectors meet preset trigger conditions, respectively executing content optimization operation, inventory decision operation and/or re-analysis operation corresponding to the commodities based on the state events, feeding back execution results to the analysis model to update the operation state vectors, and generating operation state vector abnormal state triggering again based on the operation state vector triggering detection results after the operation state vector updating. The method comprises the steps of obtaining commodity data from a plurality of different data platforms, wherein the commodity data comprise commodity image data, commodity text description data, user comment data and user behavior data, respectively carrying out field semantic analysis on the commodity data of the different data platforms to generate a platform semantic field set, establishing a cross-platform field mapping relation based on the platform semantic field set, carrying out field reconstruction on the commodity data, carrying out entity consistency check on the commodity data after field reconstruction based on a commodity unique identifier, carrying out time sequence consistency check on the commodity data which passes through the entity consistency check based on a time stamp, and carrying out the commodity data which simultaneously meets the entity consistency check and the time sequence consistency check as the structured input data. The method comprises the steps of constructing a flow state feature set, a conversion state feature set, a public praise state feature set and an inventory state feature set respectively based on structured input data, carrying out feature compression processing on each state feature set to generate corresponding state sub-vectors, distributing initial weights to each state sub-vector based on stability performance of the structured input data in different time periods, carrying ou