An e-commerce supply chain commodity intelligent traceability management method

Abstract

The present application relates to the technical field of electronic commerce supply chain, and discloses an intelligent traceability management method for commodity in an electronic commerce supply chain. The method comprises the following steps: accessing multi-source asynchronous data streams generated in the whole cycle of supply chain circulation, generating a synchronous traceability basic data set after time stamp alignment and cleaning; constructing a space-time trajectory grid of commodity circulation based on the synchronous traceability basic data set, wherein the unit of the space-time trajectory grid comprises a multi-dimensional commodity state vector; performing state evolution deduction on the grid to generate a commodity circulation behavior sequence graph composed of behavior nodes and edges; performing multi-modal deep matching between the graph and a preset path template to output a circulation decision path and a decision credibility; and dynamically optimizing packaging protection parameters to generate a protection instruction set. The method realizes continuous and fine perception of the state of commodity circulation and intelligent path decision, and improves the accuracy of traceability analysis and the initiative of risk response.

Description

Technical Field

[0001] This invention relates to the field of e-commerce supply chain technology, specifically to an intelligent traceability management method for e-commerce supply chain products. Background Technology

[0002] Current e-commerce supply chain product traceability management generally relies on collecting and recording isolated logistics event data at key nodes. This data typically only includes time, location, and simple operational status, forming a flow record composed of discrete points. There is a lack of effective correlation and continuous recording between the actual state of the goods during transportation and the operational status of the transport vehicle. This recording method limits the transparency of the supply chain to the node level, failing to depict the complete state evolution process of goods in a spatiotemporal continuum, and resulting in a lack of refined data support for anomaly detection, risk warning, and responsibility delineation.

[0003] Existing path verification and decision support technologies are mostly based on preset fixed path rules or simple spatiotemporal sequence comparisons. These methods struggle to handle complex, non-standard real-world circulation scenarios. When the circulation process doesn't perfectly match the preset template, the system typically only provides a general judgment of "abnormal" or "mismatch," failing to intelligently infer the most likely path the product would actually follow, and also lacking the ability to quantify the reliability of the inference results. This rigidity in path analysis prevents the system from providing in-depth, confident intelligent evidence for dynamic packaging protection optimization and supply chain decisions. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent traceability management method for goods in the e-commerce supply chain, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent traceability management method for goods in an e-commerce supply chain, the method comprising:

[0006] Access to multi-source asynchronous data streams generated by the target product throughout the entire supply chain cycle, including product location coordinate sequences, carrier operation status records, packaging micro-environment parameters, and multi-angle packaging visual image data;

[0007] The multi-source asynchronous data streams are timestamped and cleaned to generate a synchronized source traceability base data set;

[0008] Based on the synchronized traceability basic data set, a spatiotemporal trajectory grid for commodity circulation is constructed. Each grid cell of the spatiotemporal trajectory grid corresponds to a geographical region and a time window, and contains the commodity status vector within that cell.

[0009] The state evolution of the spatiotemporal trajectory grid is deduced to generate a commodity circulation behavior sequence map. The commodity circulation behavior sequence map consists of a series of behavior nodes and behavior edges. Behavior nodes represent the state of a specific spatiotemporal grid unit, and behavior edges represent the path and conditions of state transition.

[0010] The commodity circulation behavior sequence map is matched with a preset commodity circulation path template using multimodal deep matching. Based on the matching results, the circulation decision path and corresponding decision credibility of the target commodity are generated.

[0011] Based on the aforementioned flow decision path, the product packaging protection parameters are dynamically optimized to generate a protection optimization instruction set that includes packaging reinforcement parameters and environmental isolation parameters.

[0012] Preferably, the construction of the spatiotemporal trajectory grid of commodity circulation based on the synchronized traceability basic data set includes:

[0013] Based on the preset geospatial resolution and time slice length, the entire supply chain physical path and timeline are divided into multiple spatiotemporal grid units;

[0014] For each spatiotemporal grid cell, perform the following processing:

[0015] Extract all data points that fall within the time and geographical range of the spatiotemporal grid unit from the synchronized source tracing base data set;

[0016] Calculate the distribution center and dispersion of the commodity location coordinates within the spatiotemporal grid cell, and use them as the location state vector;

[0017] The frequency distribution characteristics of the carrier's operational status records within this spatiotemporal grid cell are statistically analyzed and used as the carrier's state vector.

[0018] Calculate the average, peak, and fluctuation variance of the packaging microenvironment parameters within the spatiotemporal grid cell, and use them as the environmental state vector;

[0019] Damage features are extracted and texture changes are analyzed from multi-angle packaging visual image data to generate the visual state vector of the spatiotemporal grid unit;

[0020] The location state vector, carrier state vector, environment state vector and visual state vector are concatenated to form a commodity state vector that represents the complete state of the spatiotemporal grid unit.

[0021] The spatiotemporal trajectory grid is generated by combining all spatiotemporal grid cells and their corresponding commodity state vectors.

[0022] Preferably, the step of performing state evolution deduction on the spatiotemporal trajectory grid to generate a commodity circulation behavior sequence map includes:

[0023] Based on the sequential order of adjacent spatiotemporal grid cells in the spatiotemporal trajectory grid on the time axis, the transfer relationship between cells is established;

[0024] Two adjacent spatiotemporal grid cells and their respective commodity state vectors are used as a transition pair;

[0025] Based on the transition pair, extract the change in the commodity state vector from the previous spatiotemporal grid cell to the commodity state vector of the next spatiotemporal grid cell, and calculate the state transition vector.

[0026] Based on preset commodity circulation behavior coding rules, the state transition vector is mapped to specific behavior type codes, including transportation behavior codes, warehousing behavior codes, and loading and unloading behavior codes.

[0027] Each spatiotemporal grid cell is defined as a behavior node, and the attributes of the behavior node include the commodity state vector of that cell;

[0028] A connection between two behavior nodes that have a chronological order is defined as a behavior edge, and the attributes of the behavior edge include the behavior type encoding and state transition vector connecting the two nodes.

[0029] Connect all behavioral nodes and behavioral edges throughout the entire supply chain cycle to form a sequence map of commodity flow behavior that represents the continuous evolution of commodity status.

[0030] Preferably, the step of performing multimodal deep matching between the commodity circulation behavior sequence map and a preset commodity circulation path template, and generating a circulation decision path and corresponding decision credibility for the target commodity based on the matching results, includes:

[0031] Retrieve one or more product flow path templates corresponding to the target product category from the knowledge base. Each product flow path template is stored in the form of a standard behavior sequence graph, which contains a series of standard behavior nodes and standard behavior edges under ideal conditions.

[0032] The sequence map of commodity circulation behavior is compared with the template of each commodity circulation path by comparing the map structure and the node attributes;

[0033] Graph structure alignment includes calculating the edit distance of the behavior node sequence and the topological similarity of the behavior edge connections;

[0034] Node attribute comparison includes calculating a multi-dimensional difference measure between the product state vectors of corresponding behavior nodes;

[0035] By integrating the graph structure comparison results with the node attribute comparison results, an overall matching score is generated between the commodity circulation behavior sequence graph and each commodity circulation path template.

[0036] Select the product flow path template with the highest overall matching score as the reference template;

[0037] Identify the sequence of behavioral nodes in the commodity circulation behavior sequence map that deviate significantly from the reference template. The determination of significant deviation is based on whether the multi-dimensional difference measurement exceeds a preset deviation threshold.

[0038] The identified behavioral node sequences with significant deviations are marked as decision points in the commodity circulation behavior sequence map;

[0039] Starting from the initial behavior node, passing through all decision points, and ending at the final behavior node, a flow decision path consisting of key decision points is formed.

[0040] A decision credibility is associated with each decision point in the flow decision path, which is calculated based on the matching score of all upstream behavioral nodes that reach that decision point.

[0041] Preferably, the step of dynamically optimizing the product packaging protection parameters based on the flow decision path to generate a protection optimization instruction set including packaging reinforcement parameters and environmental isolation parameters includes:

[0042] Analyze the decision-making process and extract all decision points in the path whose decision credibility is lower than the warning value;

[0043] Obtain the product state vector from the behavior node corresponding to each decision point below the warning value, and separate the carrier state vector, environment state vector and visual state vector from it;

[0044] For the components in the carrier state vector that indicate abnormal vibration or impact, calculate the required density and distribution location of the additional buffer material, and generate the first type of packaging reinforcement parameters;

[0045] For the components in the environmental state vector that indicate excessive temperature, humidity, or other harmful factors, calculate the required thickness and material properties of the enhanced isolation layer to generate environmental isolation parameters;

[0046] For the components in the visual state vector that indicate packaging wear or deformation, calculate the required strength of the wear-resistant coating or structural reinforcement scheme at the corresponding location, and generate the second type of packaging reinforcement parameters;

[0047] The first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points are combined and conflict-resolved to form a protection optimization sub-instruction applicable to the flow path segment corresponding to the decision point or multiple consecutive related decision points.

[0048] The protection optimization sub-instructions corresponding to all decision points below the warning value are summarized and sorted according to the order in which the decision points appear in the decision flow path to generate the protection optimization instruction set.

[0049] Preferably, the step of combining and resolving conflicts among the first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points to form a protection optimization sub-instruction applicable to the flow path segment corresponding to the decision point or the multiple consecutive related decision points includes:

[0050] List all packaging reinforcement parameters and environmental isolation parameters generated for the same decision point;

[0051] Check for physical conflicts or performance offsets between different parameters. Physical conflicts include space occupancy conflicts, and performance offsets include the reinforcement of one material causing a decrease in the isolation performance of another material.

[0052] If a physical conflict exists, the parameters will be adjusted according to the preset conflict resolution rules, which prioritize the protection dimensions that have the greatest impact on the safety of the goods.

[0053] If performance offset exists, calculate the overall protection effectiveness after offset, and iteratively adjust the relevant parameters until the overall protection effectiveness meets the protection requirements of the decision point;

[0054] The final parameter combination after conflict resolution and performance offsetting is bound to the corresponding decision point information and flow path segment information to form the protection optimization sub-instruction.

[0055] Preferably, the method further includes:

[0056] After completing the traceability management of a batch of target products, collect the real-time flow feedback dataset of the batch of products;

[0057] The real-time flow feedback dataset is transformed into a feedback behavior sequence map;

[0058] Calculate the overall matching score between the feedback behavior sequence map and the reference template, and use it as the actual matching score;

[0059] The decision confidence deviation is obtained by comparing the actual matching score with the previously predicted matching score.

[0060] Based on the decision credibility deviation, the standard behavior node attributes, standard behavior edge connections, and deviation threshold in the commodity circulation path template are adjusted to achieve adaptive optimization of the commodity circulation path template and deviation judgment rules.

[0061] The optimized product circulation path template and deviation judgment rules have been updated to the knowledge base.

[0062] Preferably, adjusting the standard behavior node attributes, standard behavior edge connections, and deviation threshold in the commodity circulation path template based on the decision credibility deviation includes:

[0063] If the decision credibility deviation indicates that the actual flow is better than expected, then analyze the state vector features of the nodes with attributes that are better than the standard behavior in the feedback behavior sequence graph;

[0064] By utilizing state vector features that are superior to the standard, the attribute vectors of the corresponding standard behavior nodes in the commodity circulation path template are positively fine-tuned;

[0065] At the same time, the deviation threshold for the dimensions related to the state vector features that are superior to the standard is relaxed;

[0066] If the decision credibility deviation indicates that the actual flow is worse than expected, then analyze the key behavioral nodes and their state vectors in the feedback behavior sequence graph that lead to a decrease in matching degree.

[0067] Strengthen the standard attribute constraints of the key behavior nodes in the commodity circulation path template;

[0068] At the same time, the deviation thresholds for dimensions related to the attributes of the key behavioral nodes are tightened;

[0069] In the commodity flow path template, the state transition conditions represented by the standard behavior edges located between key behavior nodes are synchronously adjusted to adapt them to the adjusted standard behavior node attributes.

[0070] Preferably, the method further includes:

[0071] Establish a product traceability management knowledge base, which is used to store product circulation path templates, behavior type coding rules, state vector comparison rules, deviation thresholds, and historical protection optimization instruction sets for different product categories;

[0072] The knowledge base receives new behavioral pattern data generated from the commodity circulation behavior sequence graph matching process;

[0073] The knowledge base receives feedback data on the effects of the protection optimization instruction set after execution.

[0074] The new behavioral pattern data is used to incrementally update the product circulation path templates for the corresponding product categories;

[0075] The effect feedback data is used to verify and calibrate the behavior type encoding rules, state vector comparison rules, and deviation thresholds.

[0076] The knowledge base provides updated product flow path templates, behavior type coding rules, state vector comparison rules, and deviation thresholds for the next traceability management of the same or similar product categories.

[0077] Preferably, the method includes:

[0078] Access the multi-source asynchronous data streams of the target product in the supply chain and process them to obtain a synchronized set of traceability basic data;

[0079] Construct a spatiotemporal trajectory grid to depict the spatiotemporal evolution of goods, and deduce and generate a sequence map of goods circulation behavior;

[0080] By deeply matching the sequence graph of commodity circulation behavior with the commodity circulation path template in the knowledge base, circulation decision paths and decision credibility are generated;

[0081] Based on the flow decision path, dynamically generate a set of instructions for optimizing commodity packaging protection;

[0082] By utilizing actual circulation feedback data, the product circulation path template and matching rules are adaptively optimized and updated, and the experience is stored in the knowledge base.

[0083] Compared with the prior art, the beneficial effects of the present invention are:

[0084] By aligning and fusing timestamps of multi-source asynchronous data, a discretized spatiotemporal trajectory grid was constructed. Each grid cell integrates geographical location, time interval, and a product state vector that incorporates carrier state, environmental parameters, and visual features. This expands the recording dimension of product circulation from discrete event points to continuous spatiotemporal blocks with multi-dimensional states. It enables the accurate recording and traceability of the comprehensive physical state and environmental exposure of products at any time segment throughout the entire circulation chain, achieving a panoramic, high-fidelity digital mapping of the circulation process and providing a structured data foundation for analyzing state evolution patterns.

[0085] Based on a spatiotemporal trajectory grid, state evolution is extrapolated to generate an abstract behavioral sequence graph, which is then subjected to multimodal deep matching with a preset template. This scheme does not process raw, heterogeneous data streams, but rather a graph-based model representing states and their transition logic. The matching process simultaneously analyzes the spatiotemporal relationships, state features, and evolution patterns contained within the graph. Through multimodal feature fusion and pattern recognition using a deep learning model, the most suitable flow decision path for the observed state can be inferred from potentially noisy, missing, or biased data, and the confidence level of this path inference is calculated simultaneously. This enables intelligent and flexible reconstruction and evaluation of commodity flow paths under complex real-world conditions, with output results possessing a clear confidence level. Attached Figure Description

[0086] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent traceability management method for e-commerce supply chain goods described in this invention.

[0087] Figure 2 A flowchart for generating a sequence map of commodity circulation behavior;

[0088] Figure 3 To generate a flowchart of the decision-making process and decision credibility;

[0089] Figure 4 A multi-dimensional effect analysis chart for template adaptive optimization;

[0090] Figure 5 This is a graph showing the distribution analysis of the weights and correlation coefficients of the state vector dimensions. Detailed Implementation

[0091] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0092] Please see Figure 1 This invention provides an intelligent traceability management method for e-commerce supply chain goods. The method includes: accessing multi-source asynchronous data streams generated by the target goods throughout the entire supply chain cycle, including product location coordinate sequences, carrier operation status records, packaging micro-environment parameters, and multi-angle packaging visual image data; performing timestamp alignment and data cleaning on the multi-source asynchronous data streams to remove invalid and abnormal data, generating a time-synchronized traceability basic data set; constructing a spatiotemporal trajectory grid for product circulation based on this synchronized traceability basic data set, where each cell of the grid is associated with a geographical region and a time window, and contains a product state vector within that cell; performing state evolution deduction on the spatiotemporal trajectory grid to generate a product circulation behavior sequence map composed of behavior nodes and behavior edges, where behavior nodes represent the state of a specific spatiotemporal grid cell, and behavior edges represent the path and conditions for state transition; and performing multimodal deep matching between the product circulation behavior sequence map and a preset product circulation path template, generating the circulation decision path and corresponding decision credibility of the target product based on the matching results. Based on the flow decision path, the protection parameters of commodity packaging are dynamically optimized to generate a protection optimization instruction set that includes packaging reinforcement parameters and environmental isolation parameters.

[0093] Example 1: See Figure 2In practical implementation, the system accesses multi-source asynchronous data streams generated throughout the entire supply chain cycle of the target product. These multi-source asynchronous data streams include product location coordinate sequences, carrier operation status records, packaging micro-environment parameters, and multi-angle packaging visual image data. The system performs timestamp alignment and data cleaning on these multi-source asynchronous data streams to generate a synchronized traceability foundation data set. Based on this synchronized traceability foundation data set, a spatiotemporal trajectory grid for product circulation is constructed. The entire supply chain physical path and timeline are divided into multiple spatiotemporal grid units according to a preset geospatial resolution and time slice length. For each spatiotemporal grid cell, all data points falling within the time and geographical range of that spatiotemporal grid cell are extracted from the synchronized traceability base data set. The distribution center and dispersion of the commodity location coordinates within the spatiotemporal grid cell are calculated as the location state vector. The frequency distribution characteristics of the carrier operation status records within the spatiotemporal grid cell are statistically analyzed as the carrier state vector. The average value, peak value, and fluctuation variance of the packaging microenvironment parameters within the spatiotemporal grid cell are calculated as the environmental state vector. Damage features are extracted and texture changes are analyzed from multi-angle packaging visual image data to generate the visual state vector of the spatiotemporal grid cell. The location state vector, carrier state vector, environmental state vector, and visual state vector are concatenated to form the commodity state vector representing the complete state of the spatiotemporal grid cell. All spatiotemporal grid cells and their corresponding commodity state vectors are combined to generate a spatiotemporal trajectory grid.

[0094] In some embodiments, state evolution is deduced for the spatiotemporal trajectory grid. Transition relationships between units are established based on the sequential order of adjacent spatiotemporal grid units on the time axis. Two consecutively adjacent spatiotemporal grid units and their respective commodity state vectors are considered as a transition pair. A state transition vector is calculated by extracting the change in commodity state vector from the previous spatiotemporal grid unit to the next spatiotemporal grid unit based on the transition pair. This state transition vector is mapped to a specific behavior type code based on a preset commodity flow behavior coding rule. Behavior type codes include transportation behavior codes, warehousing behavior codes, and loading / unloading behavior codes. Each spatiotemporal grid unit is defined as a behavior node, and the attributes of a behavior node include the commodity state vector of that unit. The connection between two behavior nodes with a temporal sequence is defined as a behavior edge, and the attributes of the behavior edge include the behavior type code connecting the two nodes and the state transition vector. All behavior nodes and behavior edges throughout the entire supply chain cycle are connected to form a commodity flow behavior sequence map representing the continuous evolution of commodity states.

[0095] It is understandable that when calculating the position and state vector of a spatiotemporal grid cell, a weighting factor can be introduced in the discreteness calculation to distinguish the importance of different coordinate points. For example, position coordinate discreteness. It can be calculated in the following ways:

[0096]

[0097] in: It is the total number of location coordinates within this spatiotemporal grid cell. It is the latitude and longitude value of the i-th coordinate point. It is the distribution center of all coordinate points within this unit. These are the normalized weighting coefficients of the carrier velocity or vibration data associated with the i-th coordinate point. These coefficients reflect the reliability or stability of the state represented by that coordinate point; higher velocity or vibration values ​​typically correspond to more unstable states. The corresponding reduction.

[0098] Example 2: See Figure 3 After generating the product flow behavior sequence graph, multimodal deep matching is performed between this graph and preset product flow path templates. One or more product flow path templates corresponding to the target product category are retrieved from the knowledge base. Each product flow path template is stored in the form of a standard behavior sequence graph, containing a series of standard behavior nodes and standard behavior edges under ideal conditions. The product flow behavior sequence graph is compared with each product flow path template in terms of graph structure and node attributes. Graph structure comparison includes calculating the edit distance of the behavior node sequence and the topological similarity of the behavior edge connections. Node attribute comparison includes calculating the multi-dimensional difference measure between the product state vectors of corresponding behavior nodes. The graph structure comparison results and node attribute comparison results are fused to generate an overall matching score between the product flow behavior sequence graph and each product flow path template. The product flow path template with the highest overall matching score is selected as the reference template. Behavior node sequences in the product flow behavior sequence graph that deviate significantly from the reference template are identified. The determination of significant deviation is based on whether the multi-dimensional difference measure exceeds a preset deviation threshold. The identified sequences of behavioral nodes exhibiting significant deviations are marked as decision points in the commodity flow behavior sequence map. Starting from the initial behavioral node, a flow decision path is formed by traversing all decision points until the final behavioral node. Each decision point in the flow decision path is associated with a decision credibility score, calculated based on the matching scores of all upstream behavioral nodes reaching that decision point.

[0099] In practice, a multimodal deep matching process is performed between the commodity circulation behavior sequence graph and a pre-defined commodity circulation path template. Based on the matching results, a circulation decision path and corresponding decision credibility for the target commodity are generated. One or more commodity circulation path templates corresponding to the target commodity category are retrieved from the knowledge base. Each commodity circulation path template is stored in the form of a standard behavior sequence graph, containing a series of standard behavior nodes and standard behavior edges under ideal conditions. The commodity circulation behavior sequence graph contains behavior nodes and behavior edges derived from a spatiotemporal trajectory grid. Behavior node attributes include commodity state vectors, and behavior edge attributes include behavior type encoding and state transition vectors. The commodity circulation behavior sequence graph is compared with each commodity circulation path template in terms of graph structure and node attributes. Graph structure comparison includes calculating the edit distance of the behavior node sequence and the topological similarity of the behavior edge connections. Node attribute comparison includes calculating a multi-dimensional difference measure between the commodity state vectors of corresponding behavior nodes.

[0100] In some embodiments, the overall matching score between the product flow behavior sequence map and each product flow path template is generated by fusing the graph structure comparison results and the node attribute comparison results. The product flow path template with the highest overall matching score is selected as the reference template. Behavioral node sequences in the product flow behavior sequence map that deviate significantly from the reference template are identified. The determination of significant deviation is based on whether the multi-dimensional difference measure exceeds a preset deviation threshold. The identified behavioral node sequences with significant deviations are marked as decision points in the product flow behavior sequence map. Starting from the initial behavioral node, a flow decision path is formed by passing through all decision points until the terminating behavioral node. A decision credibility is associated with each decision point in the flow decision path. The decision credibility is calculated based on the matching scores of all upstream behavioral nodes that reach the decision point.

[0101] It's understandable that the overall match score... The calculation combines structural similarity and attribute similarity, and one calculation method is as follows:

[0102]

[0103] in: This represents the overall matching score between the sequence map of commodity circulation behavior and a commodity circulation path template. This represents the edit distance of the sequence of behavior nodes. Indicates the standard length of the reference sequence. This represents the topological similarity of the edges connected by the behavior, and its value is between 0 and 1. This indicates the number of action node pairs participating in the comparison. This represents the multi-dimensional difference metric between the product state vectors of the k-th action node, after normalization. , , These are the weighting coefficients assigned to edit distance, topological similarity, and node attribute differences, respectively, and satisfying the following conditions: .

[0104] Optionally, the edit distance calculation of the behavior node sequence is based on a standard dynamic programming algorithm, measuring the minimum operational cost required to convert the behavior node sequence of the commodity flow behavior sequence graph into a standard behavior node sequence of the commodity flow path template through insertion, deletion, or replacement operations. The operational cost can be defined based on the differences in behavior type encoding. The topological similarity calculation of behavior edge connections is comprehensively evaluated by comparing the existence, directionality, and consistency of behavior type encoding attached to the edges between corresponding behavior nodes in two graphs. In some embodiments, when calculating the multi-dimensional difference measure between the commodity state vectors of corresponding behavior nodes, Euclidean distance or cosine similarity is calculated for the position state vector, carrier state vector, environmental state vector, and visual state vector in the commodity state vector, respectively. Then, weights are assigned according to the importance of each state dimension to the commodity quality, and a weighted sum is obtained to finally obtain a scalarized multi-dimensional difference measure value. The preset deviation threshold can be a fixed value or a variable dynamically calculated based on the fluctuation range of the standard behavior node attributes in the commodity flow path template.

[0105] Example 3: Dynamically optimize product packaging protection parameters based on the flow decision path. Analyze the flow decision path and extract all decision points in the path whose decision credibility is below the warning value. Obtain the product state vector in the behavior node corresponding to each decision point below the warning value, and separate the carrier state vector, environmental state vector, and visual state vector from it. For the component in the carrier state vector indicating abnormal vibration or impact, calculate the required density and distribution location of the added buffer material to generate the first type of packaging reinforcement parameters. For the component in the environmental state vector indicating excessive temperature, humidity, or other harmful factors, calculate the required thickness and material properties of the enhanced isolation layer to generate environmental isolation parameters. For the component in the visual state vector indicating packaging wear or deformation, calculate the required strength of the wear-resistant coating or structural reinforcement scheme at the corresponding location to generate the second type of packaging reinforcement parameters. Combine and resolve conflicts of the first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points to form a protection optimization sub-instruction applicable to the flow path segment corresponding to the decision point or multiple consecutive related decision points. Summarize all protection optimization sub-instructions corresponding to decision points below the warning value, sort them according to the order in which the decision points appear in the decision flow path, and generate a protection optimization instruction set.

[0106] This process combines and resolves conflicts among the first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple related decision points. All packaging reinforcement and environmental isolation parameters generated for the same decision point are listed. The process checks for physical conflicts or performance offsets between different parameters. Physical conflicts include space occupancy conflicts, and performance offsets include a decrease in the isolation performance of one material due to reinforcement of another. If physical conflicts exist, parameters are adjusted according to preset conflict resolution rules, prioritizing the protection dimensions with the greatest impact on product safety. If performance offsets exist, the overall protection effectiveness after offsetting is calculated, and relevant parameters are iteratively adjusted until the overall protection effectiveness meets the protection requirements of that decision point. The final parameters after conflict resolution and performance offsetting are combined and bound to the corresponding decision point information and flow path segment information to form a protection optimization sub-instruction.

[0107] In practical implementation, the protective parameters of product packaging are dynamically optimized based on the flow decision path, generating a protective optimization instruction set that includes packaging reinforcement parameters and environmental isolation parameters. The flow decision path is analyzed to extract all decision points whose decision credibility is below a warning value, which is a preset numerical threshold for decision credibility. The product state vector in the behavioral node corresponding to each decision point below the warning value is obtained, and the carrier state vector, environmental state vector, and visual state vector are separated from the product state vector. For the component in the carrier state vector indicating abnormal vibration or impact, the required density and distribution location of the added buffer material are calculated, generating the first type of packaging reinforcement parameters. For the component in the environmental state vector indicating excessive temperature, humidity, or other harmful factors, the required thickness and material properties of the enhanced isolation layer are calculated, generating environmental isolation parameters. For the component in the visual state vector indicating packaging wear or deformation, the required strength of the anti-wear coating or structural reinforcement scheme at the corresponding location is calculated, generating the second type of packaging reinforcement parameters. The first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points are combined and conflict-resolved to form protection optimization sub-instructions applicable to the flow path segment corresponding to the decision point or the multiple consecutive related decision points. All protection optimization sub-instructions corresponding to decision points below the warning value are summarized and sorted according to the order in which the decision points appear in the flow decision path to generate the final protection optimization instruction set.

[0108] In some embodiments, the specific process of combining and resolving conflicts among the first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points involves listing all packaging reinforcement parameters and environmental isolation parameters generated for the same decision point and checking whether there are physical conflicts or performance offsets between different parameters. Physical conflicts include space occupation conflicts, such as the increased material volume for enhanced cushioning potentially encroaching on the air layer space designed for enhanced thermal insulation. Performance offsets include the reinforcement of one material causing a decrease in the isolation performance of another material, such as the added structural reinforcement material potentially weakening the original thermal insulation effect due to its thermal conductivity. If physical conflicts exist, the parameters are adjusted according to preset conflict resolution rules, which prioritize the protection dimension that has the greatest impact on product safety. If performance offsets exist, the comprehensive protection effectiveness after offsetting is calculated, and the relevant parameters are iteratively adjusted until the comprehensive protection effectiveness meets the protection requirements of the decision point. The final parameters after conflict resolution and performance offset are combined and bound to the corresponding decision point information and flow path segment information to form a protection optimization sub-instruction.

[0109] It is understandable that one way to quantify the overall protection effectiveness E involves modeling and aggregating the contributions of each protection parameter, and its expression is:

[0110]

[0111] in: Indicates the overall protective effectiveness index. This indicates the number of decision points or related decision point sequences that need to be considered. , , These represent the quantified performance values ​​of the first type of packaging reinforcement parameter (buffering), environmental isolation parameter, and second type of packaging reinforcement parameter (strengthening) calculated for the j-th decision point or decision point sequence, respectively. , , These are the effectiveness weighting coefficients corresponding to the three types of parameters mentioned above. These weighting coefficients are jointly determined by the characteristics of the product category, historical damage data, and the specific circulation stage at which the decision point is located. During the iterative adjustment of parameters to optimize the comprehensive protection effectiveness index E, it is necessary to ensure that the E value is not lower than the minimum protection requirement threshold preset for this circulation path. .

[0112] Optionally, the preset conflict resolution rules can be based on a predefined priority list of protection dimensions. For example, for fresh produce, temperature isolation has a higher priority than shock resistance; for precision electronic products, shock resistance may have a higher priority than humidity isolation. When a physical space occupancy conflict is detected, the system will prioritize the implementation of high-priority dimension parameters according to this priority list, and proportionally reduce or restructure the parameters of low-priority dimensions.

[0113] In some embodiments, when calculating the required increase in the density and distribution location of the buffer material, the input is the component in the carrier state vector representing the vibration spectrum energy or impact acceleration peak, and the output is the basis weight density distribution map of the buffer material and its attachment coordinates on the inner wall of the packaging box. When calculating the required enhanced insulation layer thickness and material properties, the input is the component in the environmental state vector representing the degree and duration of temperature and humidity deviations from safe ranges, and the output is the target thickness of the insulation material, thermal conductivity, or the number of added moisture-proof film layers. When calculating the required increase in the strength of the abrasion-resistant coating or structural reinforcement scheme, the input is the component in the visual state vector representing the surface scratch density or box deformation, and the output is the basis weight of the local coating spray or the size and layout of the reinforcing ribs. Optionally, the determination of multiple consecutive related decision points is based on the spatiotemporal proximity of these decision points on the flow decision path and the similarity of their state anomaly patterns. For example, if multiple decision points appearing consecutively on a highway transport segment all indicate abnormal high-frequency vibrations, these decision points are considered related, and the protection optimization parameters generated for them will be combined to generate a coherent reinforced buffer scheme for that transport segment, rather than multiple isolated instructions.

[0114] Example 4: After completing the traceability management of a batch of target products, collect the real-time flow feedback dataset of these products. Transform the real-time flow feedback dataset into a feedback behavior sequence graph. Calculate the overall matching score between the feedback behavior sequence graph and the reference template, using this as the actual matching score. Compare the actual matching score with the previously predicted matching score to obtain the decision credibility deviation. Based on the decision credibility deviation, adjust the standard behavior node attributes, standard behavior edge connections, and deviation thresholds in the product flow path template to achieve adaptive optimization of the product flow path template and deviation judgment rules. Update the optimized product flow path template and deviation judgment rules to the knowledge base.

[0115] The specific process for adjusting the standard behavioral node attributes, standard behavioral edge connections, and deviation thresholds in the commodity circulation path template based on decision credibility deviation is as follows: If the decision credibility deviation indicates that the actual circulation is better than expected, the state vector features of the superior-than-standard behavioral node attributes in the feedback behavioral sequence graph are analyzed. Using these superior-than-standard state vector features, the attribute vectors of the corresponding standard behavioral nodes in the commodity circulation path template are positively fine-tuned. Simultaneously, the deviation thresholds for dimensions related to the superior-than-standard state vector features are relaxed. If the decision credibility deviation indicates that the actual circulation is worse than expected, the key behavioral nodes and their state vectors that lead to a decrease in matching degree in the feedback behavioral sequence graph are analyzed. The standard attribute constraints of the corresponding key behavioral nodes in the commodity circulation path template are strengthened. Simultaneously, the deviation thresholds for dimensions related to the key behavioral node attributes are tightened. The state transition conditions represented by the standard behavioral edges located between key behavioral nodes in the commodity circulation path template are adjusted synchronously to adapt to the adjusted standard behavioral node attributes.

[0116] In practical implementation, after completing the traceability management of a batch of target goods, a real-time circulation feedback dataset of these goods is collected. This dataset includes new location coordinates, carrier operating status, packaging microenvironment parameters, and visual image data generated during the subsequent actual circulation of the goods. The real-time circulation feedback dataset is transformed into a feedback behavior sequence map. The transformation process follows the same steps as generating a product circulation behavior sequence map from a multi-source asynchronous data stream, including constructing a spatiotemporal trajectory grid and performing state evolution inference. The overall matching score between the feedback behavior sequence map and the reference template is calculated, and the result is used as the actual matching score. The actual matching score is compared with the previously predicted matching score to obtain the decision confidence deviation, which is the difference between the actual and predicted matching scores. Based on the decision confidence deviation, the standard behavior node attributes, standard behavior edge connections, and deviation thresholds in the product circulation path template are adjusted to achieve adaptive optimization of the product circulation path template and deviation judgment rules. The optimized product circulation path template and deviation judgment rules are then updated to the knowledge base.

[0117] In some embodiments, the specific operation of adjusting the standard behavior node attributes, standard behavior edge connections, and deviation thresholds in the commodity circulation path template based on decision credibility deviation is as follows: If the decision credibility deviation indicates that the actual circulation is better than expected, then the state vector features of the superior-than-standard behavior node attributes in the feedback behavior sequence graph are analyzed, and the attribute vectors of the corresponding standard behavior nodes in the commodity circulation path template are positively fine-tuned using the superior-than-standard state vector features, while simultaneously relaxing the deviation thresholds of the dimensions related to the superior-than-standard state vector features. If the decision credibility deviation indicates that the actual circulation is worse than expected, then the key behavior nodes and their state vectors that lead to a decrease in matching degree in the feedback behavior sequence graph are analyzed, the standard attribute constraints of the corresponding key behavior nodes in the commodity circulation path template are strengthened, and the deviation thresholds of the dimensions related to the key behavior node attributes are tightened. The state transition conditions represented by the standard behavior edges located between key behavior nodes in the commodity circulation path template are adjusted synchronously to adapt the state transition conditions to the adjusted standard behavior node attributes.

[0118] It is understandable that there is a bias in the credibility of decisions. The calculation formula is ,in This represents the actual matching score obtained by comparing the feedback behavior sequence map with the reference template. This represents the predicted matching score obtained by comparing the sequence map of commodity circulation behavior in previous traceability management with the same reference template. Based on the deviation... The sign and magnitude of the value, deviating from the threshold. Adjustment amount It can be determined in the following ways:

[0119]

[0120] in: This indicates the adjustment amount that deviates from the threshold. It is a preset learning rate coefficient used to control the magnitude of adjustment. It is a decision credibility bias. This is the original deviation threshold baseline value for the dimension to be adjusted. When When (actual performance is better than expected), A positive value means relaxing the deviation threshold; when When (actual performance is worse than expected), A negative value means tightening the deviation from the threshold.

[0121] Optionally, see Table 1, which shows a simplified example of a comparison of decision credibility biases, involving three different key behavioral node dimensions (e.g., vibration intensity, temperature stability, and packaging wear rate).

[0122] Table 1: Comparison of Actual Matching Score and Predicted Matching Score, and Deviation Analysis

[0123]

[0124] In practical implementation, when analyzing the state vector features of attributes superior to standard behavior nodes in the feedback behavior sequence map, it is necessary to locate specific behavior nodes whose actual matching scores are higher than the predicted matching scores. The components exhibiting better performance in the environmental or carrier state vectors of these nodes are extracted, such as lower average vibration acceleration or a more stable temperature fluctuation range. These values ​​are then used to update the attribute vectors of the corresponding standard behavior nodes in the commodity circulation path template. The update method can be a weighted moving average calculation of the corresponding components of the standard vector in the direction of better actual performance. Relaxing the deviation threshold involves multiplying the deviation threshold related to that dimension by a coefficient greater than 1, or adding a positive offset.

[0125] In some embodiments, strengthening the standard attribute constraints of key behavioral nodes in the commodity circulation path template can be achieved by narrowing the allowable fluctuation range of each component in the attribute vector of the standard behavioral node, for example, modifying the temperature standard value range of ±3℃ to ±2℃. Tightening the deviation threshold is manifested by multiplying the deviation threshold related to this dimension by a coefficient less than 1, or reducing it by a negative offset. The state transition conditions represented by the standard behavioral edge are adjusted synchronously, including updating the allowable range of change in the state transition vector to ensure consistency with the trend of attribute vector changes after adjustment of the two related standard behavioral nodes. Optionally, a preset learning rate coefficient... The learning rate coefficient can be dynamically set based on the value of the product category, the amount of historical data accumulated, and the stability of model convergence. For high-value product categories with little data accumulation, a smaller learning rate coefficient can be used to maintain the stability of the template, while for product categories with sufficient data accumulation, a larger learning rate coefficient can be used to accelerate the adaptive optimization process.

[0126] See Figure 4In the effect analysis of template adaptive optimization (knowledge base update stage), this graph uses the optimization round as the horizontal axis and simultaneously presents the changes in three dimensions: average matching degree improvement, threshold adjustment range (left vertical axis), and protection optimization effect score (right vertical axis). Specifically: the average matching degree improvement (blue line) reaches its peak in the second round and then gradually decreases, reflecting that with optimization iterations, the space for improving the matching degree between the actual flow and the template gradually narrows; the threshold adjustment range (red line) is positively correlated with the average matching degree improvement, with the peak matching degree in the second round corresponding to the highest point of the threshold adjustment range, and subsequently decreasing as the matching degree falls back, reflecting the linkage between threshold adjustment and matching degree optimization; the protection optimization effect score (green bar) remains at a high level in each round and shows a slight upward trend, indicating that even if the matching degree improvement range decreases, the actual effectiveness of protection optimization is still steadily increasing. The results verify that in the knowledge base update stage, template adaptive optimization, by dynamically adjusting the threshold and node attributes, can continuously improve the effectiveness of protection while the matching degree converges.

[0127] Example 5: Establish a product traceability management knowledge base. This knowledge base stores product flow path templates, behavior type coding rules, state vector comparison rules, deviation thresholds, and historical protection optimization instruction sets for different product categories. The knowledge base receives new behavior pattern data generated from the product flow behavior sequence graph matching process and effect feedback data from the execution of protection optimization instruction sets. The new behavior pattern data is used to incrementally update the product flow path templates for the corresponding product categories, and the effect feedback data is used to verify and calibrate the behavior type coding rules, state vector comparison rules, and deviation thresholds. The knowledge base provides updated product flow path templates, behavior type coding rules, state vector comparison rules, and deviation thresholds for the next traceability management of the same or similar product categories.

[0128] The overall methodology includes: accessing multi-source asynchronous data streams of the target product in the supply chain; processing the data to obtain a synchronous traceability base data set; constructing a spatiotemporal trajectory grid depicting the spatiotemporal evolution of the product's state; generating a sequence graph of product circulation behavior; performing deep matching between the sequence graph and product circulation path templates in the knowledge base to generate circulation decision paths and decision credibility; dynamically generating a set of product packaging protection optimization instructions based on the circulation decision paths; adaptively optimizing and updating the product circulation path templates and matching rules using actual circulation feedback data; and storing the experience in the knowledge base.

[0129] In practical implementation, a product traceability management knowledge base is established. This knowledge base stores product circulation path templates, behavior type coding rules, state vector comparison rules, deviation thresholds, and historical protection optimization instruction sets for different product categories. The knowledge base receives new behavior pattern data generated from the product circulation behavior sequence map matching process and effect feedback data from the execution of protection optimization instruction sets. The new behavior pattern data is used to incrementally update the product circulation path templates for the corresponding product categories, and the effect feedback data is used to verify and calibrate the behavior type coding rules, state vector comparison rules, and deviation thresholds. The knowledge base provides updated product circulation path templates, behavior type coding rules, state vector comparison rules, and deviation thresholds for the next traceability management of the same or similar product categories.

[0130] In some embodiments, the new behavioral pattern data received by the product traceability management knowledge base refers to behavioral node sequence patterns or state transition patterns that are not fully covered or explained by existing templates, discovered during multimodal deep matching of the product circulation behavior sequence map with existing product circulation path templates in the knowledge base. The effect feedback data received by the product traceability management knowledge base includes the state data of the product in subsequent actual circulation after packaging adjustments according to the protection optimization instruction set, as well as the integrity assessment data upon final arrival. Incremental updates using new behavioral pattern data are performed by fusing new pattern fragments with existing product circulation path templates, such as adding an optional behavioral branch to the template or modifying the attribute vector value range of a standard behavioral node to cover newly observed states.

[0131] It's understandable that the process of incrementally updating the product distribution path template can be represented by a template evolution function. Assume that for a certain product category, its product distribution path template has an iteration cycle... The state is In the cycle The collection of new behavioral pattern data is as follows The effect feedback data set is The updated template It can be represented as:

[0132]

[0133] in: The function represents the specific update operation. This represents the product circulation path template for this category stored in the knowledge base at the beginning of the t-th iteration cycle. This represents the set of new behavioral pattern data identified and extracted during the matching process of commodity circulation behavior sequence map in the t-th period. This represents the set of effect feedback data collected after executing the protection optimization instruction set in the t-th cycle. This indicates the new version of the product flow path template obtained after the update operation, which will be used for the next cycle. The specific implementation of the function includes based on Expanding template structures or node attributes, and based on Fine-tune the standard attribute values ​​or state transition conditions of associated nodes in the template.

[0134] Optionally, validating and calibrating behavior type coding rules using performance feedback data involves checking whether the behavior type codes obtained by mapping from state transition vectors according to existing rules are highly consistent with the actual physical operations reflected in the performance feedback data. If a systematic deviation in the mapping relationship is found, the mapping threshold or classification boundary in the behavior type coding rules is adjusted. Validating and calibrating state vector comparison rules involves analyzing the correlation between the integrity of goods and multi-dimensional difference measures in the performance feedback data, and adjusting the weight coefficients of each dimension in the state vector comparison rules so that the matching score can more accurately predict the risk of goods damage. Validating and calibrating deviation thresholds involves dynamically adjusting the strictness of the deviation threshold based on the overlap between the actual problem nodes marked in the performance feedback data and the decision points previously determined by the system, in order to improve the precision and recall rate of anomaly identification.

[0135] In some embodiments, the product traceability management knowledge base provides updated resources for the next traceability management cycle. Specifically, when the system processes multi-source asynchronous data streams of the same product category again, it retrieves an incrementally updated product flow path template from the knowledge base. The behavior type encoding rules, state vector comparison rules, and deviation thresholds used are also the latest versions after verification and calibration based on the previous round of effect feedback data. This allows the system's matching accuracy, decision reliability, and the specificity of the generated protection optimization instruction set to gradually evolve with increased usage. Optionally, the overall method integrates a closed loop from data access to knowledge updates. The process includes accessing multi-source asynchronous data streams of the target product in the supply chain, processing them to obtain a synchronous traceability basic data set, constructing a spatiotemporal trajectory grid depicting the spatiotemporal state evolution of the product and generating a product circulation behavior sequence map, performing deep matching of the product circulation behavior sequence map with the product circulation path template in the knowledge base to generate circulation decision paths and decision credibility, dynamically generating a product packaging protection optimization instruction set based on the circulation decision path, adaptively optimizing and updating the product circulation path template and matching rules using actual circulation feedback data, and storing the optimized and updated experience in the product traceability management knowledge base. The product traceability management knowledge base plays the role of core experience storage and continuous learning engine in this process.

[0136] See Figure 5 In the construction of the state vector of the spatiotemporal trajectory grid of commodity circulation, the weight distribution and correlation coefficients of four dimensions—location state, carrier state, environmental state, and visual state—are presented. Specifically, the figure uses a bar chart to depict the weight ratio of each state dimension, and a line graph (red node) to show the correlation coefficient between each dimension and the target traceability indicator: the correlation coefficient of the carrier state dimension reaches its peak (approximately 0.875), corresponding to its relatively prominent weight among the four dimensions; the correlation coefficient of the visual state dimension drops to its lowest (0.700), and its weight also shows a corresponding adjustment characteristic; the correlation coefficients of the location state and environmental state are in the middle range (approximately 0.83 and 0.775, respectively), matching the gradient change of their weights. This multi-dimensional weight-correlation linkage distribution provides a quantitative basis for the dimensional importance ranking of the state vector and the subsequent optimization of node attributes in the commodity circulation behavior sequence map. Among them, the high correlation feature of the carrier state dimension can serve as a core reference dimension for the behavior edge transition conditions in the state evolution deduction.

[0137] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0138] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent traceability management of goods in an e-commerce supply chain, characterized in that, The method includes: Access to multi-source asynchronous data streams generated by the target product throughout the entire supply chain cycle, including product location coordinate sequences, carrier operation status records, packaging micro-environment parameters, and multi-angle packaging visual image data; The multi-source asynchronous data streams are timestamped and cleaned to generate a synchronized source traceability base data set; Based on the synchronized traceability basic data set, a spatiotemporal trajectory grid for commodity circulation is constructed. Each grid cell of the spatiotemporal trajectory grid corresponds to a geographical region and a time window, and contains the commodity status vector within that cell. The state evolution of the spatiotemporal trajectory grid is deduced to generate a commodity circulation behavior sequence map. The commodity circulation behavior sequence map consists of a series of behavior nodes and behavior edges. Behavior nodes represent the state of a specific spatiotemporal grid unit, and behavior edges represent the path and conditions of state transition. The commodity circulation behavior sequence map is matched with a preset commodity circulation path template using multimodal deep matching. Based on the matching results, the circulation decision path and corresponding decision credibility of the target commodity are generated. Based on the aforementioned flow decision path, the protective parameters of the product packaging are dynamically optimized to generate a set of protective optimization instructions that includes packaging reinforcement parameters and environmental isolation parameters; The step of performing multimodal deep matching between the commodity circulation behavior sequence map and a preset commodity circulation path template, and generating the circulation decision path and corresponding decision credibility of the target commodity based on the matching results, includes: Retrieve one or more product flow path templates corresponding to the target product category from the knowledge base. Each product flow path template is stored in the form of a standard behavior sequence graph, which contains a series of standard behavior nodes and standard behavior edges under ideal conditions. The sequence map of commodity circulation behavior is compared with the template of each commodity circulation path by comparing the map structure and the node attributes; Graph structure alignment includes calculating the edit distance of the behavior node sequence and the topological similarity of the behavior edge connections; Node attribute comparison includes calculating a multi-dimensional difference measure between the product state vectors of corresponding behavior nodes; By integrating the graph structure comparison results with the node attribute comparison results, an overall matching score is generated between the commodity circulation behavior sequence graph and each commodity circulation path template. Select the product flow path template with the highest overall matching score as the reference template; Identify the sequence of behavioral nodes in the commodity circulation behavior sequence map that deviate significantly from the reference template. The determination of significant deviation is based on whether the multi-dimensional difference measurement exceeds a preset deviation threshold. The identified behavioral node sequences with significant deviations are marked as decision points in the commodity circulation behavior sequence map; Starting from the initial behavior node, passing through all decision points, and ending at the final behavior node, a flow decision path consisting of key decision points is formed. A decision credibility is associated with each decision point in the flow decision path, which is calculated based on the matching score of all upstream behavioral nodes that reach that decision point.

2. The intelligent traceability management method for e-commerce supply chain goods according to claim 1, characterized in that, The construction of the spatiotemporal trajectory grid for commodity circulation based on the synchronized traceability basic data set includes: Based on the preset geospatial resolution and time slice length, the entire supply chain physical path and timeline are divided into multiple spatiotemporal grid units; For each spatiotemporal grid cell, perform the following processing: Extract all data points that fall within the time and geographical range of the spatiotemporal grid unit from the synchronized source tracing base data set; Calculate the distribution center and dispersion of the commodity location coordinates within the spatiotemporal grid cell, and use them as the location state vector; The frequency distribution characteristics of the carrier's operational status records within this spatiotemporal grid cell are statistically analyzed and used as the carrier's state vector. Calculate the average, peak, and fluctuation variance of the packaging microenvironment parameters within the spatiotemporal grid cell, and use them as the environmental state vector; Damage features are extracted and texture changes are analyzed from multi-angle packaging visual image data to generate the visual state vector of the spatiotemporal grid unit; The location state vector, carrier state vector, environment state vector and visual state vector are concatenated to form a commodity state vector that represents the complete state of the spatiotemporal grid unit. The spatiotemporal trajectory grid is generated by combining all spatiotemporal grid cells and their corresponding commodity state vectors.

3. The intelligent traceability management method for e-commerce supply chain goods according to claim 2, characterized in that, The step of performing state evolution deduction on the spatiotemporal trajectory grid to generate a commodity circulation behavior sequence map includes: Based on the sequential order of adjacent spatiotemporal grid cells in the spatiotemporal trajectory grid on the time axis, the transfer relationship between cells is established; Two adjacent spatiotemporal grid cells and their respective commodity state vectors are used as a transition pair; Based on the transition pair, extract the change in the commodity state vector from the previous spatiotemporal grid cell to the commodity state vector of the next spatiotemporal grid cell, and calculate the state transition vector. Based on preset commodity circulation behavior coding rules, the state transition vector is mapped to specific behavior type codes, including transportation behavior codes, warehousing behavior codes, and loading and unloading behavior codes. Each spatiotemporal grid cell is defined as a behavior node, and the attributes of the behavior node include the commodity state vector of that cell; A connection between two behavior nodes that have a chronological order is defined as a behavior edge, and the attributes of the behavior edge include the behavior type encoding and state transition vector connecting the two nodes. Connect all behavioral nodes and behavioral edges throughout the entire supply chain cycle to form a sequence map of commodity flow behavior that represents the continuous evolution of commodity status.

4. The intelligent traceability management method for e-commerce supply chain goods according to claim 3, characterized in that, The process of dynamically optimizing product packaging protection parameters based on the flow decision path generates a protection optimization instruction set that includes packaging reinforcement parameters and environmental isolation parameters, including: Analyze the decision-making process and extract all decision points in the path whose decision credibility is lower than the warning value; Obtain the product state vector from the behavior node corresponding to each decision point below the warning value, and separate the carrier state vector, environment state vector and visual state vector from it; For the components in the carrier state vector that indicate abnormal vibration or impact, calculate the required density and distribution location of the additional buffer material, and generate the first type of packaging reinforcement parameters; For the components in the environmental state vector that indicate excessive temperature, humidity, or other harmful factors, calculate the required thickness and material properties of the enhanced isolation layer to generate environmental isolation parameters; For the components in the visual state vector that indicate packaging wear or deformation, calculate the required strength of the wear-resistant coating or structural reinforcement scheme at the corresponding location, and generate the second type of packaging reinforcement parameters; The first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points are combined and conflict-resolved to form a protection optimization sub-instruction applicable to the flow path segment corresponding to the decision point or multiple consecutive related decision points. The protection optimization sub-instructions corresponding to all decision points below the warning value are summarized and sorted according to the order in which the decision points appear in the decision flow path to generate the protection optimization instruction set.

5. The intelligent traceability management method for e-commerce supply chain goods according to claim 4, characterized in that, The first type of packaging reinforcement parameters, environmental isolation parameters, and second type of packaging reinforcement parameters generated for the same decision point or multiple consecutive related decision points are combined and conflict-resolved to form a protection optimization sub-instruction applicable to the flow path segment corresponding to the decision point or the multiple consecutive related decision points, including: List all packaging reinforcement parameters and environmental isolation parameters generated for the same decision point; Check for physical conflicts or performance offsets between different parameters. Physical conflicts include space occupancy conflicts, and performance offsets include the reinforcement of one material causing a decrease in the isolation performance of another material. If a physical conflict exists, the parameters will be adjusted according to the preset conflict resolution rules, which prioritize the protection dimensions that have the greatest impact on the safety of the goods. If performance offset exists, calculate the overall protection effectiveness after offset, and iteratively adjust the relevant parameters until the overall protection effectiveness meets the protection requirements of the decision point; The final parameter combination after conflict resolution and performance offsetting is bound to the corresponding decision point information and flow path segment information to form the protection optimization sub-instruction.

6. The intelligent traceability management method for e-commerce supply chain goods according to claim 5, characterized in that, The method further includes: After completing the traceability management of a batch of target products, collect the real-time flow feedback dataset of the batch of products; The real-time flow feedback dataset is transformed into a feedback behavior sequence map; Calculate the overall matching score between the feedback behavior sequence map and the reference template, and use it as the actual matching score; The decision confidence deviation is obtained by comparing the actual matching score with the previously predicted matching score. Based on the decision credibility deviation, the standard behavior node attributes, standard behavior edge connections, and deviation threshold in the commodity circulation path template are adjusted to achieve adaptive optimization of the commodity circulation path template and deviation judgment rules. The optimized product circulation path template and deviation judgment rules have been updated to the knowledge base.

7. The intelligent traceability management method for e-commerce supply chain goods according to claim 6, characterized in that, The step of adjusting the standard behavior node attributes, standard behavior edge connections, and deviation threshold in the commodity circulation path template based on the decision credibility deviation includes: If the decision credibility deviation indicates that the actual flow is better than expected, then analyze the state vector features of the nodes with attributes that are better than the standard behavior in the feedback behavior sequence graph; By utilizing state vector features that are superior to the standard, the attribute vectors of the corresponding standard behavior nodes in the commodity circulation path template are positively fine-tuned; At the same time, the deviation threshold for the dimensions related to the state vector features that are superior to the standard is relaxed; If the decision credibility deviation indicates that the actual flow is worse than expected, then analyze the key behavioral nodes and their state vectors in the feedback behavior sequence graph that lead to a decrease in matching degree. Strengthen the standard attribute constraints of the key behavior nodes in the commodity circulation path template; At the same time, the deviation thresholds for dimensions related to the attributes of the key behavioral nodes are tightened; In the commodity flow path template, the state transition conditions represented by the standard behavior edges located between key behavior nodes are synchronously adjusted to adapt them to the adjusted standard behavior node attributes.

8. The intelligent traceability management method for e-commerce supply chain goods according to claim 1, characterized in that, The method further includes: Establish a product traceability management knowledge base, which is used to store product circulation path templates, behavior type coding rules, state vector comparison rules, deviation thresholds, and historical protection optimization instruction sets for different product categories; The knowledge base receives new behavioral pattern data generated from the commodity circulation behavior sequence graph matching process; The knowledge base receives feedback data on the effects of the protection optimization instruction set after execution. The new behavioral pattern data is used to incrementally update the product circulation path templates for the corresponding product categories; The effect feedback data is used to verify and calibrate the behavior type encoding rules, state vector comparison rules, and deviation thresholds. The knowledge base provides updated product flow path templates, behavior type coding rules, state vector comparison rules, and deviation thresholds for the next traceability management of the same or similar product categories.

9. The intelligent traceability management method for e-commerce supply chain goods according to claim 1, characterized in that, The method includes: Access the multi-source asynchronous data streams of the target product in the supply chain and process them to obtain a synchronized set of traceability basic data; Construct a spatiotemporal trajectory grid to depict the spatiotemporal evolution of goods, and deduce and generate a sequence map of goods circulation behavior; By deeply matching the sequence graph of commodity circulation behavior with the commodity circulation path template in the knowledge base, circulation decision paths and decision credibility are generated; Based on the flow decision path, dynamically generate a set of instructions for optimizing commodity packaging protection; By utilizing actual circulation feedback data, the product circulation path template and matching rules are adaptively optimized and updated, and the experience is stored in the knowledge base.

Images