Cross-border logistics intelligent multimodal transport platform and method

Abstract

The present application relates to the technical field of cross-border logistics, in particular to a cross-border logistics intelligent multimodal transport platform and method, comprising: for the current order, obtaining the basic influence coefficient of each overseas rule in each available candidate combination, then obtaining the theoretical influence coefficient of each cascade attribute of each overseas rule, obtaining the path dependence correction coefficient of each available candidate combination, correcting the theoretical influence coefficient by using the path dependence correction coefficient to obtain the actual influence coefficient, thereby determining the total correction coefficient of each cross-border logistics attribute, correcting the initial weight of each cross-border logistics attribute by using the total correction coefficient, combining the attribute value of each cross-border logistics attribute to determine the comprehensive score of each available candidate combination, and screening the optimal cross-border logistics transportation path of the current order from all available candidate combinations based on the comprehensive score. The present application greatly improves the accuracy of path planning by quantifying the dynamic influence of overseas rules on weights.

Description

Technical Field

[0001] This invention relates to the field of cross-border logistics technology, specifically to a cross-border logistics intelligent multimodal transport platform and method. Background Technology

[0002] With the acceleration of global trade integration, cross-border multimodal transport (a combination of sea, land, air, and rail transport modes) has become a core model for reducing logistics costs and improving transportation efficiency. The global cross-border freight volume continues to grow, and cross-border multimodal transport accounts for an increasingly larger share of cross-border freight, making it the mainstream transportation organization form in cross-border logistics. However, cross-border transport involves the customs, tax, and quarantine rules of multiple countries / regions, and these rules are often dynamically adjusted according to trade policies and geopolitical situations, placing extremely high demands on the accuracy of route planning.

[0003] Existing problems: Existing cross-border multimodal transport route planning schemes generally adopt the logic of "enumeration combination → calculation attribute → fixed weight scoring", but there are three core pain points in actual application: (1) Rigid weights and high deviation rate: The cost, timeliness and compliance weights of route scoring rely entirely on the fixed values ​​preset by the customer, and the impact of dynamic changes in overseas rules is not quantified. (2) Black box model and high threshold for implementation: Some optimization schemes use black box models such as ant colony algorithm, LSTM, graph neural network to calculate dynamic weights, but these models require a large amount of historical transportation data for training. Small and medium-sized freight forwarding companies and cargo owners lack the ability to accumulate data, and the model calculation process is unexplainable. When deviation occurs, the cause cannot be traced, and the implementation cost is high, making it difficult to popularize. (3) Hard constraints of rules and insufficient universality: Most schemes only use overseas rules as hard constraints of "allow / prohibit" to directly filter routes, or only adapt to specific trade scenarios such as RCEP and China-Europe freight train. For example, for cross-border transportation in Southeast Asia, existing schemes cannot adapt to the scattered quarantine rule adjustments of ASEAN countries, resulting in a large number of feasible routes being mistakenly filtered out, and the coverage of applicable scenarios is insufficient. Summary of the Invention

[0004] This invention provides a cross-border logistics intelligent multimodal transport platform and method to solve existing problems.

[0005] The present invention provides a cross-border logistics intelligent multimodal transport platform and method, which adopts the following technical solution: One embodiment of the present invention provides a cross-border logistics intelligent multimodal transport method, the method comprising the following steps: For the current order, obtain the initial weight of each cross-border logistics attribute of the customer's fuzzy demand conversion, each available candidate combination of cross-border logistics transportation routes, and the attribute value of each cross-border logistics attribute corresponding to each available candidate combination; Obtain the enforcement strength and clarity coefficient of each foreign rule in each available candidate combination; determine the basic impact coefficient of each foreign rule based on the magnitude of the enforcement strength and clarity coefficient; Obtain the transmission coefficient of each cascading attribute of each overseas rule; each overseas rule and each cascading attribute correspond to a cross-border logistics attribute; determine the theoretical impact coefficient of each cascading attribute of each overseas rule based on the magnitude of the basic impact coefficient and the transmission coefficient; obtain the path dependency correction coefficient of each available candidate combination; use the path dependency correction coefficient to correct the theoretical impact coefficient to obtain the actual impact coefficient of each cascading attribute of each overseas rule; For each available candidate combination, obtain the basic influence coefficient of all overseas rules corresponding to the same cross-border logistics attribute and the actual influence coefficient of all cascaded attributes, determine the total correction coefficient of each cross-border logistics attribute, use the total correction coefficient to correct the initial weight, and combine the attribute value to determine the comprehensive score of each available candidate combination. Based on the comprehensive score, the optimal cross-border logistics transportation route for the current order is selected from all available candidate combinations.

[0006] Furthermore, the specific steps for determining the basic impact coefficient of each overseas rule are as follows: The product of the normalized value of the enforcement strength of each foreign rule and the clarity coefficient is recorded as the basic impact coefficient of each foreign rule.

[0007] Furthermore, the specific steps for determining the theoretical influence coefficient of each cascading attribute of each overseas rule are as follows: The product of the basic impact coefficient of each foreign rule and the transmission coefficient of each cascade attribute of each foreign rule is denoted as the theoretical impact coefficient of each cascade attribute of each foreign rule.

[0008] Furthermore, the specific steps for obtaining the actual impact coefficient of each cascading attribute of each overseas rule are as follows: The product of the path dependency correction coefficient of each available candidate combination and the theoretical influence coefficient of each cascade attribute of each overseas rule in each available candidate combination is denoted as the actual influence coefficient of each cascade attribute of each overseas rule in each available candidate combination.

[0009] Furthermore, the specific steps for determining the total correction coefficient for each cross-border logistics attribute are as follows: For each available candidate combination, the sum of the basic influence coefficients of all overseas rules corresponding to any cross-border logistics attribute and the actual influence coefficients of all cascaded attributes corresponding to any cross-border logistics attribute is denoted as the attribute influence coefficient of any cross-border logistics attribute. Obtain the path complexity adjustment value and multilateral agreement correction coefficient for each available candidate combination; Based on the attribute influence coefficient, the path complexity adjustment value, and the multilateral agreement correction coefficient, determine the total correction coefficient for each cross-border logistics attribute corresponding to each available candidate combination.

[0010] Furthermore, the specific steps for determining the total correction coefficient for each cross-border logistics attribute corresponding to each available candidate combination based on the attribute influence coefficient, the path complexity adjustment value, and the multilateral agreement correction coefficient are as follows: Obtain the sum of the attribute influence coefficient of each cross-border logistics attribute corresponding to each available candidate combination and the path complexity adjustment value of each available candidate combination. Multiply the sum by the multilateral agreement correction coefficient of each available candidate combination and record it as the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination.

[0011] Furthermore, the specific steps for determining the comprehensive score of each available candidate combination are as follows: The product of the initial weight of each cross-border logistics attribute and the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination is denoted as the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination. The overall score for each available candidate combination is determined based on the dynamic weight and attribute value of each cross-border logistics attribute corresponding to each available candidate combination.

[0012] Furthermore, the specific steps for determining the comprehensive score of each available candidate combination based on the dynamic weight and attribute value of each cross-border logistics attribute corresponding to each available candidate combination are as follows: Based on the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination, the attribute values ​​of all cross-border logistics attributes corresponding to each available candidate combination are weighted and averaged to obtain the comprehensive score of each available candidate combination.

[0013] Furthermore, the specific steps for selecting the optimal cross-border logistics transportation route for the current order from all available candidate combinations based on the comprehensive score are as follows: The candidate combination with the highest overall score is designated as the optimal cross-border logistics transportation route for the current order.

[0014] The present invention also proposes a cross-border logistics intelligent multimodal transport platform, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned cross-border logistics intelligent multimodal transport method.

[0015] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, for the current order, the enforcement strength and clarity coefficients of each overseas rule in each available candidate combination are obtained, the basic influence coefficient of each overseas rule is determined, the transmission coefficient of each cascade attribute of each overseas rule is obtained, the theoretical influence coefficient of each cascade attribute of each overseas rule is determined, the path dependency correction coefficient of each available candidate combination is obtained, the theoretical influence coefficient is corrected using the path dependency correction coefficient to obtain the actual influence coefficient, thereby determining the total correction coefficient of each cross-border logistics attribute, the initial weight of each cross-border logistics attribute is corrected using the total correction coefficient, and the comprehensive score of each available candidate combination is determined by combining the attribute value of each cross-border logistics attribute. Based on the comprehensive score, the optimal cross-border logistics transportation route for the current order is selected from all available candidate combinations. Thus, an intelligent multimodal transport route planning scheme that relies solely on publicly available data, is fully explainable throughout the process, and is adaptable to all cross-border scenarios is proposed. By quantifying the dynamic influence of overseas rules on weights, the accuracy of route planning is significantly improved. It can reduce the deviation rate between route planning and actual transportation, prevent cost overruns and significant time delays, and is highly interpretable with transparent calculation logic throughout the process. Customers can clearly understand the reasons for weight adjustments, increasing trust. It has low implementation costs, relying only on publicly available data and requiring no pre-trained models. It has broad applicability: adaptable to all cross-border transportation scenarios, regardless of the region involved, such as RCEP, the EU, or North America, dynamic weight adjustments can be achieved through rule-structured extraction. Thus, this invention achieves intelligent and precise cross-border multimodal transport route planning by quantifying the dynamic impact of overseas rules. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of an intelligent multimodal transport method for cross-border logistics according to the present invention. Figure 2 This is a flowchart of the modules of a cross-border logistics intelligent multimodal transport platform. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a cross-border logistics intelligent multimodal transport platform and method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific solution of the intelligent multimodal transport platform and method for cross-border logistics provided by this invention.

[0021] Please see Figure 1 The diagram illustrates a flowchart of a cross-border logistics intelligent multimodal transport method according to an embodiment of the present invention, which includes the following steps: Step S001: For the current order, obtain the initial weight of each cross-border logistics attribute of the customer's fuzzy demand transformation, each available candidate combination of cross-border logistics transportation routes, and the attribute value of each cross-border logistics attribute corresponding to each available candidate combination.

[0022] This embodiment collects only two types of low-cost, readily available data: basic order information submitted by customers and publicly available rule announcements from various countries. First, natural language processing technology is used to transform fragmented, unstructured rule announcements into standardized, computable structured data. Then, information such as the current order's transportation scope and cargo category is matched to objectively calculate the correction coefficients of each rule for cost, timeliness, and compliance attributes. A dynamic weight matrix is ​​obtained by combining the customer's initial weights, and this matrix is ​​substituted into the objective function to calculate the suitability of each intermodal transport combination, outputting the optimal route. Simultaneously, the rule impact coefficients are automatically calibrated based on the actual results of each transport. No pre-trained model is required, the entire process logic is interpretable, and the implementation cost is extremely low.

[0023] The core operations of this embodiment include: (1) Order basic information collection: obtaining the origin, destination, goods category, expected departure time, initial weight requirements and loss tolerance threshold of customer orders; (2) Candidate intermodal transport combination enumeration and filtering: enumerating all intermodal transport combinations that meet the capacity conditions, calculating the initial attribute values ​​and filtering combinations that do not meet the customer's minimum requirements; (3) Overseas rule structure extraction: capturing official rule announcements from various countries and storing them as structured data according to standardized dimensions; (4) Rule-weight adaptation calculation: filtering effective rules that are suitable for the current order, calculating the influence coefficient of the rules on the attributes, and generating a dynamic weight matrix in combination with the order characteristics; (5) Candidate group suitability calculation: calculating the suitability score of each combination through the dynamic weight matrix and filtering the optimal path; (6) Rule influence coefficient calibration: updating the rule influence coefficient with actual transportation results to continuously improve the accuracy of planning.

[0024] The main purpose of this embodiment is to solve the problem of rigid weights caused by dynamic changes in overseas rules in cross-border multimodal transport route planning, to achieve accuracy, interpretability and universality of route planning, to reduce the landing costs for small and medium-sized freight forwarders, and to improve customer satisfaction.

[0025] Order basic information collection is the data entry point and constraint definition layer for the entire multimodal transport route planning process. Its core objective is to transform the customer's vague transportation needs into a standardized, structured order feature vector. This vector will serve as the sole data source for subsequent capacity enumeration, rule matching, dynamic weight generation, and risk filtering. The comprehensiveness and accuracy of the collected information directly determine the final optimal route fit. Information collection method: The platform offers diverse information collection channels to cater to the usage habits of different customers: Front-end visual data entry: Provides a form interface through a web page or mini-program, supporting interactive methods such as drop-down selection, slider adjustment, and date picking, reducing the difficulty for customers to fill in the form; Standardized file batch import: Supports batch order uploads in formats such as Excel and CSV, making it convenient for large and medium-sized cargo owners to submit multiple orders at once; API Interface Integration: Provides a RESTful API interface to support real-time push of order data from customers' own business systems (such as ERP, WMS) to achieve automated workflow; Automatic data validation: During data collection, fields are validated for legality (e.g., date format, weight normalization, threshold range). If the data does not conform to the specifications, corrections are prompted immediately to ensure the validity of the input data.

[0026] The function of each data item is as follows: Departure point / Destination: Data format: The internationally recognized UN / LOCODE (United Nations Trade and Transport Location Code) standard mapping is adopted, such as "CNSZX" representing Shenzhen Yantian Port and "DEHAM" representing Hamburg Port; Function: To determine the spatial origin and destination of transportation, for subsequent capacity database retrieval and intermodal combination enumeration; and to serve as the basis for the country / region scope of subsequent rule capture.

[0027] Product categories: Data format: Use the platform's built-in standardized category thesaurus (such as "consumer electronics", "fresh cold chain", "hazardous chemicals"), and recommend that customers provide at least the first 4 digits of the HS international customs code; Purpose: To ensure accurate semantic matching of the "Scope of Applicable Goods" with overseas rules, thereby avoiding rule omissions or mismatches due to ambiguous product categories.

[0028] Expected departure time: Data format: Standard date format (YYYY-MM-DD, year-month-day), the system automatically checks whether it is later than the collection time; Purpose: To serve as a time benchmark for subsequent filtering of valid rules. The system will use this benchmark to extract rules that are effective within the transportation cycle (usually within 30 days after the departure date), and also to calculate the interval between the rule publication time and the departure time.

[0029] Initial weight requirements: Cost attribute, namely total logistics cost (also known as total freight or door-to-door cost): includes the sum of all costs such as sea, land and air freight, THC (terminal handling charge), customs declaration fee, destination port clearance fee, estimated tariff, fuel surcharge, etc.

[0030] Time-of-use attribute, also known as door-to-door transit time or ETA commitment days: the total number of calendar days from when the container is loaded at the factory or the port of origin until it arrives at the end customer's warehouse or port of destination.

[0031] Compliance attributes, namely customs clearance stability (also known as customs inspection risk or cargo damage / loss rate): mainly refer to the probability of customs inspection, the difficulty of document compliance, the risk of sensitive goods being detained, and whether it involves sanctioned / embargoed countries.

[0032] In real-world cross-border logistics Request for Quotation (RFQ) scenarios, clients rarely provide precise percentage weights. Small and medium-sized freight forwarders and factories typically only make qualitative business requests to logistics providers. This platform transforms clients' vague requests into an initial weight vector that the system can recognize. The specific data collection method is as follows: Scenario-based selection (choose one of four, the most commonly used and best suited to business habits): The platform provides four scenario-based options on the order form page. Customers only need to select one based on the actual urgency of the goods and their budget. "Economy First" (also known in the industry as "price war" or "taking the cheap route"): The system gives the highest initial weight to the overall logistics cost (e.g., 0.65). "Timeliness priority" (i.e., "rushing to meet delivery deadlines" or "grabbing cargo space"): The system gives the highest initial weight to the timeliness of the entire transportation process (e.g., 0.65). "Safety First" (i.e., "fear of inspection" or "fear of seizure", commonly seen in high-value or sensitive goods): The system gives the highest initial weight to the safety of customs clearance (e.g., 0.65). "Standard equilibrium" (i.e. "cost-effective route"): The initial weights of the three dimensions are evenly distributed (approximately 0.33 for each).

[0033] The platform pre-sets these mapping relationships, so customers don't need to understand the weight values; they only need to select the tags that meet their business requirements.

[0034] Extracting implicit hard constraints (usually hidden in customer notes or order records): In addition to the qualitative choices mentioned above, the system also needs to identify the customer's implicit non-negotiable conditions, which are often more critical than "weights" in reality. Cut-off time / latest arrival date (corresponding to the hard ceiling of delivery time): For example, if the customer notes "must catch the ship on June 25th, otherwise the factory will run out of materials." After the system extracts this date, it automatically forces the initial weight of the entire transportation time to be increased to above 0.5 (if the selected scenario is lower than this value, it will be automatically corrected), and uses it as a hard filter condition to directly eliminate all overdue combinations.

[0035] Single-container cost budget ceiling (corresponding to a hard ceiling on costs): For example, if the customer states, "Total cost exceeding RMB 12,000 / TEU will not be considered." The system stores this value in the filter criteria, and any subsequent combinations with initial total costs exceeding this value will be directly eliminated.

[0036] Maximum tolerable loss threshold: Data format: expressed as a percentage (e.g., 15%), representing the customer's overall tolerance limit for initial cost overruns and time delays; Function: As a hard rejection condition in the subsequent filtering stage, any candidate combination whose initial attribute value may exceed this threshold will be directly excluded from the subsequent scoring process, reducing invalid calculations while ensuring that the risk level of the recommended path is within the acceptable range for the customer.

[0037] Following the above method, for the current order (e.g., cross-border multimodal transport route planning for electronics products from Shenzhen, China to Hamburg, Germany), the initial weight of each cross-border logistics attribute is obtained to convert the customer's vague demands. These cross-border logistics attributes include: cost, timeliness, and compliance.

[0038] Example data: The customer submits the following order information: origin: Shenzhen Yantian Port, destination: Hamburg Port, cargo category: consumer electronics products, expected departure date: June 1, 2024, initial weights (cost 0.5, timeliness 0.3, compliance 0.2), and maximum loss threshold: 15%.

[0039] Candidate intermodal transport combination enumeration and filtering, operation process: Pre-enumeration screening: Available capacity filtering. Before enumerating combinations, the system first connects to the capacity database to perform a round of availability screening, eliminating transport segments that cannot be actually booked or executed at the current time. Specific filtering conditions include: Current booking status of shipping companies / railway companies / airlines: Check the actual remaining booking status of each carrier within 3 days before and after the expected departure time, and exclude sailings that are sold out or temporarily suspended. Port / terminal congestion status: Based on publicly available port congestion indices (such as port waiting time data released by the Shanghai Shipping Exchange), if the average berthing time at a certain transshipment port currently exceeds 7 days, routes involving that port will be temporarily suspended or have their priority reduced. Geopolitical / Sanctions Restrictions: Match the sanctions lists issued by the customs of China and the destination country, and directly exclude combinations involving sanctioned countries / regions (such as rail routes passing through a sanctioned country). Carrier qualifications and historical performance rate: Only routes of carriers with no major container / container rollovers in the past 3 months and a schedule on time rate higher than the industry average will be retained.

[0040] Enumerating intermodal transport combinations: Based on the origin and destination, the system enumerates all feasible multimodal transport combinations that have passed the pre-screening criteria from the capacity database. The enumeration rules adopt a "segmented connection" strategy (e.g., origin (Shenzhen Yantian Port) and destination (Hamburg Port)): First leg of transport: Options for departure from Yantian Port in Shenzhen include sea freight (to major European ports) and air freight (to European hub airports). Transit points: After arriving at the Port of Rotterdam by sea, connect to the European inland rail or road to the Port of Hamburg; after arriving at Frankfurt Airport by air, connect to land transport to the Port of Hamburg. Combined generation: The optional carriers, sailing schedules / times, and transfer connection times for each segment are combined into a complete door-to-door link.

[0041] The enumeration result in this example (pre-screened): Combination A: Shenzhen Yantian Port → (Maersk Line, voyage MAERSK EUX) → Rotterdam Port → (DB Deutsche Bahn, China-Europe Railway Express) → Hamburg Port; Combination B: Shenzhen Bao'an Airport → (Lufthansa, flight LH783) → Frankfurt Airport → (German land transport) → Port of Hamburg; Combination C (pre-screened and suspended, for reference only): Shenzhen Yantian Port → (Pacific International Lines) → Ho Chi Minh City Port, Vietnam (9 days waiting for berthing) → (Maersk transshipment) → Hamburg Port.

[0042] Calculating initial attribute values: For each candidate combination that passes the pre-screening, the system calculates three key initial attribute values ​​using publicly available data sources. All calculations are based on objective, publicly available, and verifiable data, and do not involve subjective judgment.

[0043] The total logistics cost (initial) is an estimated value of the door-to-door full-link cost, which includes the following quantifiable sub-items: (1) Basic freight: the publicly available freight rates of each segment carrier (US dollar ocean freight published on the shipping company's official website, the train fare published on the railway platform, and the freight rate published by the airline), converted to RMB at the exchange rate of the day; (2) Port / airport handling fees (THC / ground handling fees): calculated according to the publicly available charging standards of the port of origin and the port of transit; (3) Customs declaration and clearance agency fees: estimated based on the industry average customs declaration service fees of Shenzhen Yantian Port and Hamburg Port; (4) Estimated value of customs duties and taxes: estimated based on the product category (consumer electronics products) and HS code prefix (8517), according to the current import tariff rate and VAT rate implemented in the destination country (Germany / EU), and does not include the impact of sudden policy adjustments (this impact will be corrected by subsequent dynamic weighting).

[0044] The total transit time (initial) is the total number of natural days from the port of origin to the port of destination for customs clearance and unloading, calculated by adding up the publicly announced planned operating times for each segment: (1) Sea voyage / air flight time: according to the schedule published by the carrier (e.g., Maersk Shenzhen → Rotterdam standard voyage 25 days, Lufthansa Shenzhen → Frankfurt standard flight + stopover time 7 days); (2) Transit connection time: including waiting time and transshipment time between the two ports (e.g., Rotterdam Port sea-rail transshipment standard operation time 1 day); (3) Estimated customs clearance time: based on the average customs clearance processing time published by the customs at each port (Shenzhen Customs export clearance about 1 day, EU Customs import clearance about 2-3 days).

[0045] Customs clearance reliability (initial risk index) reflects the probability that goods will be inspected, detained or encounter document problems by customs under this route. It is a risk index between 0 and 1. The lower the value, the more reliable the goods are. The calculation basis includes: (1) Historical customs clearance success rate: the proportion of similar goods released by customs in the past 6 months under this port of origin-port of destination combination (data source: platform historical order anonymization statistics and inspection rates published by customs at various ports); (2) Transit node complexity: for each additional transit country, an additional customs clearance link is introduced, and the risk index is added by 0.05 for each transit node; (3) Sensitivity level of goods category: consumer electronics products are of medium sensitivity, and the basic risk value is 0.08; air freight has a slightly lower basic risk value than sea freight because of stricter document requirements and faster inspection.

[0046] Filter candidate combinations: First layer: Actual availability of transport capacity (already completed in the pre-screening).

[0047] Conditions: Carriers in each segment of the combination are currently available for booking, and the transit port is not in a state of severe congestion.

[0048] In this example, combination C is excluded due to congestion at Ho Chi Minh Port.

[0049] The second layer: hard constraints on timeliness.

[0050] If the customer sets a "latest arrival date" or "cut-off time", the system calculates the total time for each combination and directly eliminates combinations that exceed that date.

[0051] In this example, the customer requires that the goods must arrive in Hong Kong before June 28th. Combination A has a delivery time of 28 days (expected arrival in Hong Kong on June 28th), which just meets the requirement. Combination B (7 days) meets the requirement. Combination C (35 days, expected arrival in Hong Kong on July 5th) is overdue and is therefore excluded.

[0052] The third layer: constraints on costs and compliance bottom lines.

[0053] Cost floor: If the customer has set a "total cost cap for the entire logistics process," combinations exceeding this value will be automatically eliminated. In this example, the customer has not set a cost cap, therefore this will not be triggered. Compliance baseline: If the customs clearance stability risk index of a portfolio exceeds the client's acceptable threshold (e.g., risk index > 0.3), it will be directly eliminated. In this example, none of the three portfolios exceeded the threshold, and all were retained.

[0054] It should be noted that: when enumerating combinations, mainstream transportation routes should be selected first to ensure the feasibility of the routes; when calculating the initial attribute values, publicly available logistics quotations and historical customs clearance data should be used to ensure the objectivity of the data; filtering operations can reduce the amount of subsequent calculations and improve planning efficiency.

[0055] Following the above method, for the current order, the enumerated candidate intermodal transport combinations are filtered to obtain several available candidate combinations of cross-border logistics transportation routes. Then, the attribute values ​​for each cross-border logistics attribute corresponding to each available candidate combination are obtained. These attribute values ​​include: standardized end-to-end comprehensive logistics cost, standardized end-to-end transportation timeliness, and customs clearance reliability, all of which are between 0 and 1.

[0056] It should be noted that: for the current order, the maximum cost and maximum delivery time set by the customer are obtained (if the customer has not set them, the platform will set them automatically). The ratio of the total logistics cost to the maximum cost for each available candidate combination is used as the standardized total logistics cost; the ratio of the total delivery time to the maximum delivery time for each available candidate combination is used as the standardized total delivery time.

[0057] Step S002: Obtain the enforcement strength and clarity coefficient of each overseas rule in each available candidate combination; determine the basic influence coefficient of each overseas rule based on the magnitude of the enforcement strength and clarity coefficient.

[0058] The structured extraction of overseas rules serves as a crucial bridge connecting "static order information" and "dynamic weight generation." Its goal is to transform unstructured and fragmented rule change texts scattered across government websites and international organization bulletin boards into standardized, dimensionally unified, and directly usable structured rule datasets for subsequent numerical calculations. Through this process, previously obscure legal provisions and administrative announcements will be transformed into structured objects with computable fields such as "impact attributes," "enforcement strength," "effective window period," and "trigger conditions." Operation process: Rule information sources and collection strategies.

[0059] The target information source (official and authoritative channels) is defined by the system's preset rules, which cover the following types of sources to ensure the authority and accuracy of the information: Customs and Taxation: China General Administration of Customs website (dynamic announcement board), EU Tax and Customs Union (TAXUD) website, U.S. Customs and Border Protection (CBP) website, Vietnam General Department of Customs website, ASEAN Customs Coordination Committee announcement board; Quarantine and Access: World Organisation for Animal Health (WOAH) epidemic reports, and lists of prohibited and restricted items issued by the agricultural ministries / animal and plant quarantine bureaus of various countries; Trade agreements / policy matters: RCEP Secretariat's announcements of agreement updates, WTO Trade Policy Review reports, and tariff adjustment documents issued by the EU's Trade Defence Instruments; Logistics infrastructure: official congestion reports from major ports / airports, and temporary train schedule adjustments announced by railway operators.

[0060] Data collection method: Web crawler (main channel): For the announcement list pages of official websites of various countries, crawl the title, publication time and full text of new announcements at regular intervals (UTC 00:00 every day) and save them as raw HTML / PDF files; RSS Subscription (Secondary Channel): Connects to the EU's EUR-Lex and the General Administration of Customs of China's RSS feed to enable real-time push of announcements with a delay of no more than 1 hour; PDF / Word document parsing: Some official announcements are released as PDF attachments. The system has a built-in OCR + text extraction tool (such as Apache PDFBox) to extract the main text; Incremental data collection strategy: Each collection only captures announcements newly published since the last collection (by comparing announcement IDs or publication timestamps) to avoid duplicate processing.

[0061] The data collection scope window defaults to capturing rule announcements published within the last 60 days. The reason is as follows: The planning cycle for cross-border multimodal transport is usually 7-45 days, but the "announcement period" for some rules is longer (such as EU tariff adjustments, which are usually announced 30-60 days in advance). Capturing 60 days can prevent missing important rules that are about to take effect. At the same time, the amount of rule data within 60 days is still within a controllable range and does not affect the calculation efficiency. Rules that are more than 60 days in advance are regarded as "historical rules" and are only used as a reference for calibrating the rule impact coefficient. They are not included in the selection of valid rules for current orders.

[0062] The structured extraction dimension definition: The system utilizes natural language processing techniques (word segmentation, named entity recognition, dependency parsing, and regular expression matching) to map key information in the original announcement text into the following six standardized dimensions: Country / Region of Publication: Data format is standardized country code (ISO 3166-1 alpha-2), for example: CN (China), EU (European Union), VN (Vietnam), DE (Germany).

[0063] Rule change types: The data format is an enumeration of values: TAX (taxes and fees), TIME (time-limited), ACCESS (access rights), OTHER (others). For example, taxes and fees include tariff adjustments, VAT changes, and anti-dumping duties; time-limited includes customs clearance process optimization / extension and inspection rate adjustments; access rights include modifications to the prohibited and restricted transport list and changes to quarantine requirements.

[0064] Effective start and end time: The data format is a date range (YYYY-MM-DD to YYYY-MM-DD), for example: "2024-05-20 to 2024-12-31". If it is valid indefinitely, the end time is recorded as "2099-12-31".

[0065] Applicable Goods Scope: The data format is an array of product category labels + HS code prefix range. For example: ["Consumer Electronics", "Lithium Batteries"] + HS prefix ["8507", "8517"]. The system determines whether an order falls within the applicable scope by matching the product category thesaurus and the HS code range.

[0066] Triggering conditions: The data format is a structured logical expression (JSON format), describing the additional conditions under which the rule is triggered. For example, EU tariff adjustment rules may include additional conditions such as "applies only to goods with a declared value of less than 150 euros" or "applies only to goods originating from non-EU member states." These conditions must be evaluated during subsequent matching; otherwise, the rule may be mistakenly applied to irrelevant orders.

[0067] Enforcement strength: The data format is an enumeration of values ​​from 1 to 5 (based on the rule-based mapping of penalty measures), see “Enforcement strength level rules” below for details.

[0068] The enforcement severity rating rules (non-subjective scoring) do not rely on manual subjective assignment. Instead, the system uses keyword-based rule mapping based on the descriptions of "punishment measures" and "consequences of violations" in the original announcement. Level 1: The announcement contains words such as "reminder," "suggestion," or "notification," but lacks explicit penalties. For example: A country issues a friendly reminder: "Enterprises are advised to pay attention to the new regulations and to apply as soon as possible."

[0069] Level 2: The announcement contains words such as "warning," "temporary suspension of release," or "tax payment." For example: Failure to declare as required will result in a warning or the need to pay the difference in taxes.

[0070] Level 3: The announcement includes specific amounts or clear actions such as "fine of X yuan", "late payment fee", or "detention of goods". For example: Violation will be subject to a fine of 5% of the value of the goods, or the goods will be temporarily detained for 3 working days.

[0071] Level 4: The announcement includes severe measures such as "heavy fines," "suspension of import qualifications," and "revocation of licenses." For example, importers may be fined more than 5,000 euros or have their import qualifications suspended for three months.

[0072] Level 5: The announcement includes extreme measures such as "confiscation of goods," "import ban," and "criminal prosecution." For example, the goods in violation will be confiscated and destroyed, and those responsible will face legal consequences.

[0073] Example of rating: The original text of the EU tariff adjustment announcement states that "importers who fail to declare as required will be fined 15% of the value of the goods". The keyword "fine X yuan" is matched, and the system automatically rates it as Level 3.

[0074] NLP extraction process: Text preprocessing: Convert PDF / Word documents to plain text, handling line breaks and special characters; Multi-language support: For announcements from different countries (Chinese, English, French, German, Vietnamese, etc.), call the corresponding language word segmentation model (e.g., jieba for Chinese, spaCy for English) to ensure entity recognition accuracy.

[0075] Entity and Relationship Extraction: Time Entity Extraction: Using regular expressions to match phrases such as "from," "effective at," and "valid until," the effective date is extracted; Category Entity Extraction: Identifying noun phrases following keywords such as "scope of application," "applicable goods," and "the following categories" in the announcement, and performing fuzzy matching with the platform's built-in category thesaurus; Trigger Condition Extraction: Identifying clauses following logical qualifiers such as "but only," "provided that," "excluding," and "only applicable to," and extracting them into structured condition expressions.

[0076] Manual review as a fallback: For structured results with a confidence level below the threshold (<80%), the system marks them as "awaiting manual review" and pushes them to the platform management backend for confirmation and correction by operations personnel. This mechanism ensures a high accuracy rate in extracting key rules.

[0077] Rule conflict and deduplication handling: In real-world business scenarios, multiple related rules may be issued simultaneously in the same country / region, sometimes even resulting in rule conflicts (e.g., Announcement A raises tariffs, while Announcement B on the same day announces tariff exemptions for similar goods). This operation incorporates conflict resolution and deduplication logic: Same publisher, same category, same effective time: by default, the later published rule will override the earlier published rule (judged by the announcement publication time), and the system will only retain the latest rule; For different publishers, the same product category, and the same attribute dimension (such as when Chinese customs and EU customs issue new regulations on electronic products at the same time): both rules are retained as independent entries and are used together to adjust the weight by summing, reflecting the superimposed effect of multiple rules. Rules for mutually exclusive triggering conditions: If the triggering conditions of two rules are logically mutually exclusive (e.g., one only applies to FOB terms and the other only applies to CIF terms), the system will automatically determine which rule to follow based on the trade terms of the current order, and the other rule will not be included in this calculation.

[0078] To improve system efficiency and avoid frequent invalid requests to official websites of various countries, the system establishes a rule cache database using a rule caching and incremental update mechanism. Newly captured rules: Stored in the main table with (publishing country + rule ID + publication time) as the unique key, and marked with the status "pending effect / effective / expired"; Subsequent requests: For the same order route (e.g., Shenzhen → Hamburg), the system will prioritize retrieving rules from the cache within the last 60 days that have an effective time covering the current order period, without needing to fetch them repeatedly; Incremental Update: The daily scheduled task only captures incremental announcements (by comparing the latest release time in the cache). New rules are automatically appended, and expired rules (effective / end date < current date) are automatically marked as "expired" and no longer participate in the valid rule filtering.

[0079] Example structured rule data: In this example, the system captured 12 rule changes related to the transportation of electronic products between May 20 and May 30, 2024. After NLP extraction and conflict resolution, two valid rules were ultimately implemented (the remaining 10 rules were filtered out due to category mismatch, effective time not covering the order cycle, or being covered by subsequent rules): Rule 1 (EU Taxes and Fees): Issued by the EU; Rule change type: TAX (Taxes and Fees); Effective from May 20, 2024 to December 31, 2024; Applicable goods: Consumer electronics (HS prefix 8517 / 8507); Triggering condition: When the country of origin is China; Enforcement level: Level 4 (heavy fines); Impact attribute mapping: Total logistics costs.

[0080] Rule 2 (Vietnam Time-Limited): Issuing country / region is VN; rule change type is TIME (time-limited); effective period is from 2024-05-25 to 2024-07-01; applicable goods are all categories (no category restrictions); trigger condition is that the transit port is Ho Chi Minh Port; enforcement level is 2 (warning + suspension); the affected attribute is mapped to the entire transportation time-limited period.

[0081] Global effective rule screening (based on order dimension): The NLP extraction mentioned above is not 100% certain. Some announcements use vague wording such as "maybe," "at discretion," or "depending on the situation." Directly treating these as certainties (levels 1-5) will introduce bias. Therefore, it is necessary to attach a rule clarity coefficient to each rule and calculate the expected value and fluctuation range of its influence coefficient.

[0082] The system determines the clarity coefficient of rules by classifying them into three confidence levels based on the clarity of the penalty measures described in the original announcement: High confidence level: The judgment criteria include specific numerical values ​​(such as "a fine of 5,000 euros" or "an additional 15% tariff"); the clarity coefficient is 1.0; the business meaning is that the impact is certain and can be directly substituted into the calculation.

[0083] The criteria for judgment include a range (e.g., "fine of 2000-5000 euros" or "tax rate may be increased"); the clarity factor is 0.7; and the business implication is that there is some room for flexibility, and the expected value should be taken.

[0084] Low confidence: The criteria for judgment are that it contains only qualitative descriptions (such as "punishment measures will be taken as appropriate" or "handled at the discretion of the authorities"); the clarity coefficient is 0.4; the business meaning is highly uncertain, and the impact coefficient needs to be discounted and included.

[0085] This example application: The original EU tax rules state: "Importers who fail to declare as required will be fined 15% of the value of the goods" → including the specific percentage, deemed highly certain, with a clarity factor of 1.0.

[0086] The original text of Vietnam's time limit rule states: "For goods with delayed customs clearance, customs may extend the inspection time as appropriate." → The inclusion of "may" and "as appropriate" indicates a moderate degree of certainty, with a clarity factor of 0.7.

[0087] Using the above method, for the current order, obtain the enforcement strength (level 1-5) and clarity coefficient (0-1) of each overseas rule in each available candidate combination.

[0088] It should be noted that the technical pain point to be addressed in this embodiment is the planning distortion caused by the dynamic changes in overseas rules. Therefore, it is named "overseas rules" to highlight the core improvement target in this embodiment, but this does not mean filtering or excluding domestic rules. That is, the structured extraction of overseas rules includes both domestic and overseas rules. For example, the rules extracted from the official website of the General Administration of Customs of China (domestic rules) mentioned above are also included in the subsequent analysis.

[0089] For the current order, in each available candidate combination, the enforcement strength of each overseas rule will be determined. normalized value The product of the resolution coefficient and the resolution coefficient is denoted as the basic impact coefficient for each overseas rule.

[0090] The basic impact coefficient reflects the degree of enforceability of the rule. Example: The basic impact coefficient of EU rules: (4 / 5) × 1.0 = 0.8; The basic impact coefficient of the Vietnam rule is (2 / 5)×0.7=0.28 (due to the vague wording, the effective impact is reduced from 0.4 to 0.28).

[0091] Step S003: Obtain the transmission coefficient of each cascading attribute of each overseas rule; each overseas rule and each cascading attribute correspond to a cross-border logistics attribute; determine the theoretical impact coefficient of each cascading attribute of each overseas rule based on the magnitude of the basic impact coefficient and the transmission coefficient; obtain the path dependency correction coefficient of each available candidate combination; correct the theoretical impact coefficient using the path dependency correction coefficient to obtain the actual impact coefficient of each cascading attribute of each overseas rule.

[0092] Rule-weight adaptation calculation is the core of achieving dynamic weights. It generates a weight matrix that conforms to the actual scenario by quantifying the impact of rules on attributes.

[0093] The calculation of the cascading effect between rules involves selecting a set of candidate rules that may be related to the current order from all the structured rules output above, filtering out obviously irrelevant rules, and reducing the amount of subsequent calculations.

[0094] In real-world business, the impact of a rule often extends beyond a single attribute, cascading along the logistics chain to other attributes. For example: Vietnam's time limit rules have led to longer customs clearance and inspection times at Ho Chi Minh Port, which in turn has delayed the departure time of Group C from Ho Chi Minh Port, causing them to miss the scheduled connecting rail train from Rotterdam Port, and incurring additional storage and rebooking fees during the extra 3 days of waiting in Rotterdam, thus increasing the associated costs.

[0095] When Vietnam's timeliness rules only modify the "timeliness weight" without affecting the "cost weight," it will deviate significantly from actual business practices.

[0096] This operation adds a rule cascading propagation matrix to capture the transmission effect of "one cause, multiple effects".

[0097] Define the "direct attributes" and "cascading attributes" affected by the rules. Each rule first has a directly affected attribute (determined by the "Affected Attribute Mapping" field mentioned above). At the same time, the system pre-defines a cascading transmission mapping table to define the transmission coefficients of different rule types to indirect attributes. The rule type is tax (TAX), the direct attribute is fee (cost attribute), and it is cascaded to timeliness (timeliness attribute). The transmission coefficient is 0.1, and the transmission logic is that the imposition of tariffs may lead to an increase in customs clearance inspection rate and a longer customs clearance time.

[0098] The rule type is time-sensitive (TIME), the direct attribute is time-sensitive (time-sensitive attribute), and it is cascaded to cost (cost attribute). The transmission coefficient is 0.25, and the transmission logic is that delays lead to additional costs such as demurrage fees, storage fees, and rescheduling fees.

[0099] The rule type is ACCESS, the direct attribute is stability (compliance attribute), and it is cascaded to costs (cost attribute). The transmission coefficient is 0.3, and the transmission logic is that being inspected or having goods seized directly generates operating costs and fines.

[0100] The rule type is time-sensitive (TIME), the direct attribute is time-sensitive (time-sensitive attribute), it is cascaded to stability (compliance attribute), the transmission coefficient is 0.15, and the transmission logic is that delays may trigger additional checks in subsequent nodes.

[0101] Following the above method, in each available candidate combination, obtain one cross-border logistics attribute (directly affecting attribute) and several cascading attributes (indirectly affecting attributes) corresponding to each overseas rule, as well as the transmission coefficient (0-1) of each cascading attribute of each overseas rule. Among them, the cascading attributes include cost attributes, timeliness attributes, and compliance attributes, that is, each cascading attribute corresponds to one cross-border logistics attribute.

[0102] For the current order, in each available candidate combination, the product of the basic influence coefficient of each overseas rule and the transmission coefficient of each cascade attribute of each overseas rule is recorded as the theoretical influence coefficient of each cascade attribute of each overseas rule.

[0103] For each rule, the basic influence coefficient and the theoretical influence coefficients of all cascaded attributes are used to form the complete influence vector, which consists of two parts: direct contribution and cascaded contribution.

[0104] This example application: For EU tax rules (TAX type), the basic impact coefficient is 0.8, which directly affects the cost attribute. If the impact is cascaded to timeliness, the theoretical impact coefficient is 0.8 × 0.1 = 0.08 (customs will strengthen its review after the tax is added, and the clearance time will increase slightly). If the impact is cascaded to security, and tax rules do not affect security, the theoretical impact coefficient is 0.

[0105] For Vietnam's timeliness rules (TIME category), the basic impact coefficient is 0.28, which directly affects the timeliness attribute. When cascading to costs, the theoretical impact coefficient is 0.28 × 0.25 = 0.07 (delays result in additional costs for detention and rebooking). When cascading to security, the theoretical impact coefficient is 0.28 × 0.15 = 0.042 (delays may trigger subsequent inspections).

[0106] For the current order, obtain the path dependency correction coefficient for each available candidate combination. Specifically: The path dependence correction coefficient (the actual trigger probability of a cascade effect) indicates that cascade transmission is not 100% guaranteed to occur; rather, it depends on the tightness of the connection between specific paths. For example: Combination A: The Rotterdam Port sea-rail transshipment window has a 2-day flexible buffer → even if there is a slight delay in timeliness, it may still catch the original train → the cascading transmission coefficient is reduced to 0.3 times; Combination C: The ocean shipping time from Ho Chi Minh Port to Rotterdam Port is fixed. If there is a delay in Ho Chi Minh Port, the train to Rotterdam will likely be missed (the window is only 1 day) → the cascade transmission coefficient is not discounted (1.0 times). The route connectivity flexibility factor is defined to be between 0.3 and 1.0, and is automatically calculated by the system based on the historical on-time rate and transshipment window time of the transit port to obtain the actual connectivity buffer days for each available candidate combination of the current order. and industry average buffer days Then the path dependency correction coefficient for each available candidate combination of the current order. The calculation formula is: In the formula, This is a function to minimize the value. In this embodiment, the industry average buffer days are always set to be greater than 0.

[0107] This example: Combination A: The average buffer period for Rotterdam sea-rail transshipment is 2 days, with an actual buffer period of 1.5 days. The path dependency correction factor is [missing value]. .

[0108] Combination C: The average buffer period for transit through Ho Chi Minh City is 2 days, with an actual buffer period of 0.5 days. The path dependence correction factor is [missing value]. .

[0109] For the current order, the product of the path dependency correction coefficient of each available candidate combination and the theoretical influence coefficient of each cascade attribute of each overseas rule in each available candidate combination is recorded as the actual influence coefficient of each cascade attribute of each overseas rule in each available candidate combination.

[0110] The actual impact coefficient is the actual effective value of the cascaded transmission. Example: In combination A, the EU rules are cascaded to the effective date: 0.08 × 0.8 = 0.064; In combination C, the cost of cascading the Vietnam rule is: 0.07 × 0.4 = 0.028 (due to insufficient buffering, the cascading effect is weakened because the delay time is insufficient to trigger the additional cost). In combination C, the Vietnamese rule cascades to a stability of: 0.042 × 0.4 = 0.0168.

[0111] It should be noted that the screening criteria in this embodiment (three-level matching) are as follows: Country / Region Matching: The country / region where the rule is issued must fall within the set of countries involved in all candidate combinations of the current order. In this example, the countries involved in all candidate combinations (A, B, C) are: China (country of origin), EU (destination region), Germany (country of destination), and Vietnam (transit country for combination C). Therefore, rules with EU and VN as the issuing countries are included in the initial screening; Effective Time Matching: The effective start and end time interval of the rule must intersect with the order's transportation period (expected departure time to expected arrival time). In this example, the order's transportation period is from June 1, 2024 to June 28, 2024. Both rules cover this period and are approved; Goods Category Matching: The applicable goods scope of the rule (category label or HS prefix range) must cover the goods category of the current order (consumer electronics, HS prefix 8517). The EU rule explicitly covers this category, and the Vietnam rule covers all categories, both of which are approved.

[0112] Step S004: For each available candidate combination, obtain the basic influence coefficients of all overseas rules corresponding to the same cross-border logistics attribute and the actual influence coefficients of all cascaded attributes, determine the total correction coefficient of each cross-border logistics attribute, use the total correction coefficient to correct the initial weight, and combine the attribute value to determine the comprehensive score of each available candidate combination.

[0113] The total influence coefficient (including direct and cascading effects) is calculated by summing the combined effects. This operation builds upon the basic "Matching Rules by Combination" by overlaying cascading effects, calculating the total total influence of each available candidate combination across the three attributes. Specifically: For the current order, in each available candidate combination, for any cross-border logistics attribute, such as the cost attribute, the sum of the basic influence coefficients of all overseas rules corresponding to the cost attribute and the actual influence coefficients of all cascaded attributes corresponding to the cost attribute is recorded as the attribute influence coefficient of the cost attribute.

[0114] This example: The effective direct rules for combination C are the EU rules and the Vietnam rules. Specifically, the EU rules have a direct cost attribute with a base impact coefficient of 0.8, cascading to the timeliness attribute with an actual impact coefficient of 0.064; the Vietnam rules have a direct timeliness attribute with a base impact coefficient of 0.28, cascading to the cost attribute with an actual impact coefficient of 0.028, and cascading to the compliance attribute with an actual impact coefficient of 0.0168.

[0115] The attribute impact coefficient of the cost attribute in combination C is the sum of the basic impact coefficient of the EU rule corresponding to the cost attribute (0.8) and the actual impact coefficient of the cascaded attribute corresponding to the cost attribute (0.028), which equals 0.828. Therefore, due to the cascading transmission of the Vietnam timeliness rule, the cost impact rises to 0.828. Although the increase is small, it reflects the true business logic of rule transmission.

[0116] The attribute influence coefficient of the timeliness attribute of combination C is: the sum of the basic influence coefficient of the Vietnam rule corresponding to the timeliness attribute (0.28) and the actual influence coefficient of the cascade attribute corresponding to the timeliness attribute (0.064), which is 0.344.

[0117] The attribute influence coefficient of the compliance attribute in combination C is: the actual influence coefficient of the cascade attribute corresponding to the compliance attribute is 0.0168.

[0118] For the current order, obtain the path complexity adjustment value and multilateral agreement correction coefficient for each available candidate combination. Specifically: The path complexity adjustment value is increased by 0.1 for each additional foreign country involved in the path, reflecting the increasing complexity of the rules as more countries are involved. In the example, combination A involves two foreign countries, the EU and Germany, so the path complexity adjustment value is 0.2. Multilateral trade agreement amendment: If the countries involved in the route are all contracting parties to multilateral economic and trade cooperation mechanisms such as RCEP, the multilateral agreement amendment coefficient is 0.7 (the rules within the agreement are more unified and the impact is smaller), otherwise it is 1. When the route of the available candidate combination does not involve the agreement area, the multilateral agreement amendment coefficient is 1. When the route of the available candidate combination involves the agreement area, the multilateral agreement amendment coefficient is 0.7.

[0119] For the current order, obtain the sum of the attribute influence coefficient of each cross-border logistics attribute corresponding to each available candidate combination and the path complexity adjustment value of each available candidate combination. Multiply this sum by the multilateral agreement correction coefficient of each available candidate combination and record it as the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination.

[0120] For the current order, the product of the initial weight of each cross-border logistics attribute and the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination is recorded as the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination.

[0121] It should be noted that the generation of dynamic weights is based entirely on objective data, without any subjectively set fixed thresholds. Factors such as the degree of rule enforcement, release time, path complexity, and multilateral agreements are all reflected through quantitative calculations. The entire process is explainable, and small and medium-sized freight forwarders can clearly understand the origin of the weights.

[0122] For the current order, based on the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination, the attribute values ​​of all cross-border logistics attributes corresponding to each available candidate combination are weighted and averaged to obtain the comprehensive score of each available candidate combination.

[0123] Step S005: Based on the comprehensive score, select the optimal cross-border logistics transportation route for the current order from all available candidate combinations.

[0124] For the current order, the available candidate combination with the highest comprehensive score is recorded as the optimal cross-border logistics transportation route for the current order.

[0125] It should be noted that the overall score takes into account both customer preferences and the dynamic impact of rules, resulting in an optimal route that better reflects actual transportation needs. The optimal cross-border logistics transportation route for the current order is sent to the customer for confirmation. If multiple highest overall scores exist, all available candidate combinations corresponding to these highest overall scores are sent to the customer for selection, thus determining the optimal cross-border logistics transportation route.

[0126] After completing the shipment according to the optimal cross-border logistics route for the current order, analyze the actual transportation results and report to the customer. Specifically: After the current order is completed, the actual cost deviation, timeliness deviation, and compliance results are entered, and the deviation correction factor is calculated by combining the deviation estimated during planning.

[0127] The ratio of the actual deviation to the estimated deviation for each cross-border logistics attribute is denoted as the deviation correction coefficient for each cross-border logistics attribute.

[0128] Example: If the actual cost of combination A exceeds the budget by 12%, and the estimated deviation during planning is 8%; the actual delivery time is delayed by 1 day, and the estimated delay during planning is 0.5 days; and there are no compliance issues, then the cost deviation correction factor is 12% / 8%=1.5, and the delivery time deviation correction factor is 1 / 0.5=2; and there are no compliance issues, so no correction is needed.

[0129] The product of the basic impact coefficient of each overseas rule and the deviation correction coefficient of the cross-border logistics attribute corresponding to each overseas rule is recorded as the basic impact coefficient for updating each overseas rule.

[0130] Example: The basic impact coefficient of EU tax rules has been updated to 0.8 × 1.5 = 1.2.

[0131] It should be noted that by calibrating the rule influence coefficient through actual transportation results, the system can achieve self-optimization. As transportation orders accumulate, the accuracy of the rule influence coefficient will continue to improve, and the route planning deviation rate will gradually decrease.

[0132] Based on the updated basic impact coefficient, the new optimal cross-border logistics transportation route for the current order is obtained again in the manner described above.

[0133] For the current order, if the new optimal cross-border logistics transportation route matches the actual transportation route, only a notification stating "Situation Explanation and Coefficient Updated" will be sent to the customer; if the new optimal cross-border logistics transportation route does not match the actual transportation route, the system needs to generate a "Route Correction Proposal," which includes: Comparison of the original optimal path and its predicted deviation with the actual deviation; Corrected scores for each available candidate combination and the new optimal path; Customers can choose to: accept the proposed changes and use the new route for subsequent similar batches of orders; or maintain the original route but be fully aware of the risks.

[0134] The present invention also provides a cross-border logistics intelligent multimodal transport platform, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned cross-border logistics intelligent multimodal transport method.

[0135] In this embodiment, the cross-border logistics intelligent multimodal transport platform consists of 6 core modules: (1) Order information collection module: collects customer order information through web pages, mini programs or API interfaces, and supports visual filling and data import; (2) Capacity database docking module: docks with the capacity data of mainstream global logistics service providers (such as sea freight schedules, railway schedules, and air freight space) to obtain transportation resource information in real time; (3) Overseas rule structured processing module: crawls official announcements through web crawlers and extracts structured rule data by combining natural language processing technology; (4) Rule-weight adaptation calculation module: the core algorithm module, responsible for rule matching, influence coefficient calculation and dynamic weight generation; (5) Path adaptation score module: calculates and sorts the adaptation scores of each candidate combination; (6) Rule calibration and self-optimization module: updates the rule influence coefficient with actual transportation results to form a closed-loop optimization. The module flowchart of the cross-border logistics intelligent multimodal transport platform is as follows. Figure 2 As shown.

[0136] This invention is now complete.

[0137] In summary, in this embodiment of the invention, for the current order, the enforcement strength and clarity coefficients of each overseas rule in each available candidate combination are obtained, the basic influence coefficient of each overseas rule is determined, the transmission coefficient of each cascading attribute of each overseas rule is obtained, the theoretical influence coefficient of each cascading attribute of each overseas rule is determined, the path dependency correction coefficient of each available candidate combination is obtained, the theoretical influence coefficient is corrected using the path dependency correction coefficient to obtain the actual influence coefficient, thereby determining the total correction coefficient of each cross-border logistics attribute, the initial weight of each cross-border logistics attribute is corrected using the total correction coefficient, and the comprehensive score of each available candidate combination is determined by combining the attribute value of each cross-border logistics attribute. Based on the comprehensive score, the optimal cross-border logistics transportation route for the current order is selected from all available candidate combinations. This invention significantly improves the accuracy of route planning by quantifying the dynamic influence of overseas rules on weights.

[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cross-border logistics intelligent multimodal transport method, characterized in that, The method includes the following steps: For the current order, obtain the initial weight of each cross-border logistics attribute of the customer's fuzzy demand conversion, each available candidate combination of cross-border logistics transportation routes, and the attribute value of each cross-border logistics attribute corresponding to each available candidate combination; Obtain the enforcement strength and clarity coefficient of each foreign rule in each available candidate combination; determine the basic impact coefficient of each foreign rule based on the magnitude of the enforcement strength and clarity coefficient; Obtain the transmission coefficient of each cascading attribute of each overseas rule; each overseas rule and each cascading attribute correspond to a cross-border logistics attribute; determine the theoretical impact coefficient of each cascading attribute of each overseas rule based on the magnitude of the basic impact coefficient and the transmission coefficient; obtain the path dependency correction coefficient of each available candidate combination; use the path dependency correction coefficient to correct the theoretical impact coefficient to obtain the actual impact coefficient of each cascading attribute of each overseas rule; For each available candidate combination, obtain the basic influence coefficient of all overseas rules corresponding to the same cross-border logistics attribute and the actual influence coefficient of all cascaded attributes, determine the total correction coefficient of each cross-border logistics attribute, use the total correction coefficient to correct the initial weight, and combine the attribute value to determine the comprehensive score of each available candidate combination. Based on the comprehensive score, the optimal cross-border logistics transportation route for the current order is selected from all available candidate combinations.

2. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps involved in determining the basic impact coefficient for each overseas rule are as follows: The product of the normalized value of the enforcement strength of each foreign rule and the clarity coefficient is recorded as the basic impact coefficient of each foreign rule.

3. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps for determining the theoretical impact coefficient of each cascading attribute of each overseas rule are as follows: The product of the basic impact coefficient of each foreign rule and the transmission coefficient of each cascade attribute of each foreign rule is denoted as the theoretical impact coefficient of each cascade attribute of each foreign rule.

4. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps for obtaining the actual impact coefficient of each cascading attribute of each overseas rule are as follows: The product of the path dependency correction coefficient of each available candidate combination and the theoretical influence coefficient of each cascade attribute of each overseas rule in each available candidate combination is denoted as the actual influence coefficient of each cascade attribute of each overseas rule in each available candidate combination.

5. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps for determining the total correction factor for each cross-border logistics attribute are as follows: For each available candidate combination, the sum of the basic influence coefficients of all overseas rules corresponding to any cross-border logistics attribute and the actual influence coefficients of all cascaded attributes corresponding to any cross-border logistics attribute is denoted as the attribute influence coefficient of any cross-border logistics attribute. Obtain the path complexity adjustment value and multilateral agreement correction coefficient for each available candidate combination; Based on the attribute influence coefficient, the path complexity adjustment value, and the multilateral agreement correction coefficient, determine the total correction coefficient for each cross-border logistics attribute corresponding to each available candidate combination.

6. The intelligent multimodal transport method for cross-border logistics according to claim 5, characterized in that, The specific steps for determining the total correction coefficient for each cross-border logistics attribute corresponding to each available candidate combination based on the attribute influence coefficient, the path complexity adjustment value, and the multilateral agreement correction coefficient are as follows: Obtain the sum of the attribute influence coefficient of each cross-border logistics attribute corresponding to each available candidate combination and the path complexity adjustment value of each available candidate combination. Multiply the sum by the multilateral agreement correction coefficient of each available candidate combination and record it as the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination.

7. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps for determining the comprehensive score of each available candidate combination are as follows: The product of the initial weight of each cross-border logistics attribute and the total correction coefficient of each cross-border logistics attribute corresponding to each available candidate combination is denoted as the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination. The overall score for each available candidate combination is determined based on the dynamic weight and attribute value of each cross-border logistics attribute corresponding to each available candidate combination.

8. The intelligent multimodal transport method for cross-border logistics according to claim 7, characterized in that, The specific steps for determining the comprehensive score of each available candidate combination based on the dynamic weight and attribute value of each cross-border logistics attribute corresponding to each available candidate combination are as follows: Based on the dynamic weight of each cross-border logistics attribute corresponding to each available candidate combination, the attribute values ​​of all cross-border logistics attributes corresponding to each available candidate combination are weighted and averaged to obtain the comprehensive score of each available candidate combination.

9. The intelligent multimodal transport method for cross-border logistics according to claim 1, characterized in that, The specific steps for selecting the optimal cross-border logistics transportation route for the current order from all available candidate combinations based on the comprehensive score are as follows: The candidate combination with the highest overall score is designated as the optimal cross-border logistics transportation route for the current order.

10. A cross-border logistics intelligent multimodal transport platform, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of a cross-border logistics intelligent multimodal transport method as described in any one of claims 1-9.

Images