Methods, apparatus, computer equipment and storage media for optimizing cross-border trade processes

Abstract

This application relates to the field of cross-border trade process optimization technology, and discloses a method, apparatus, computer equipment, and storage medium for optimizing cross-border trade processes. The method includes: acquiring information about goods to be traded across borders and customs clearance information; generating multiple preset dimensions based on the goods information and customs clearance information; collecting dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein the state space is the space corresponding to all preset dimensions; and optimizing the state space using a preset learning algorithm to obtain an optimized customs clearance process. The beneficial effects of this invention are: making cross-border trade processes more flexible and adaptable, improving overall operational efficiency, shortening customs clearance time, enabling enterprises to optimize resource allocation, reduce operating costs, and ultimately improve overall market responsiveness and customer satisfaction.

Description

Technical Field

[0001] This invention relates to the field of cross-border trade process optimization technology, and in particular to a method, apparatus, computer equipment, and storage medium for cross-border trade process optimization. Background Technology

[0002] In recent years, although information technology and automation have been continuously introduced into the industry, most systems still cannot efficiently integrate multi-source data, nor can they achieve dynamic and intelligent decision support. Currently, cross-border trade involves a large number of complex factors such as policies and regulations, cargo information, and route planning, which traditional systems struggle to handle. At the same time, existing process decomposition methods often lack flexibility and cannot adapt to policy changes and real-time market demands. Summary of the Invention

[0003] Therefore, it is necessary to propose a method for optimizing existing cross-border trade processes.

[0004] A method for optimizing cross-border trade processes, the method comprising:

[0005] Obtain information on goods to be traded across borders, as well as customs clearance information;

[0006] Multiple preset dimensions are generated based on the cargo information and the customs clearance information;

[0007] Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0008] The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process.

[0009] Furthermore, before the step of collecting the dimension values ​​corresponding to the preset dimension from the preset knowledge graph to form the state space, the method further includes:

[0010] Obtain historical customs clearance data in cross-border trade;

[0011] The historical customs clearance data is categorized into structured data and unstructured data according to data type.

[0012] A two-stream converter architecture is adopted to extract field-level semantic features from the structured data using a pattern-aware encoder, and to perform entity recognition on the unstructured data using a pre-trained language model to obtain each target entity.

[0013] Based on the aforementioned field-level semantic features, a spatial-semantic association matrix is ​​established by aligning the logistics trajectory coordinates with the target entity through a cross-modal attention mechanism.

[0014] Each target entity and the spatial-semantic association matrix are imported into a preset graph database to obtain the preset knowledge graph.

[0015] Furthermore, after the step of obtaining historical customs clearance data in cross-border trade, the method further includes:

[0016] Detect the customs clearance language of the historical customs clearance data;

[0017] Obtain the corresponding translation model based on the customs clearance language;

[0018] The historical customs clearance data is translated based on the translation model to obtain historical customs clearance data in a specified language.

[0019] Furthermore, after the step of importing each of the target entities and the spatial-semantic association matrix into a preset graph database to obtain the preset knowledge graph, the method further includes:

[0020] The historical customs clearance data is monitored in real time to see if any changes occur through a pre-set crawler.

[0021] If changes occur, obtain the first target text after the change and the second target text before the change;

[0022] Calculate the similarity between the first target text and the second target text;

[0023] Determine whether the similarity is greater than a threshold;

[0024] If the similarity is greater than the threshold, a GNN network is constructed to simulate the transmission effect, obtain the predicted entities affected, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph.

[0025] Furthermore, the step of generating multiple preset dimensions based on the cargo information and the customs clearance information includes:

[0026] The cargo information and customs clearance information are processed using a preset multimodal artificial intelligence model to generate a process tree structure;

[0027] The nodes in the process tree structure are identified using a preset deep learning model, thereby obtaining multiple preset dimensions.

[0028] Furthermore, the step of optimizing the state space using a preset learning algorithm to obtain the optimized customs clearance process includes:

[0029] The relevant target model is obtained based on the customs clearance information;

[0030] The state space is input into the target model, and a preset value function is set in the target model to obtain the optimized customs clearance process.

[0031] Furthermore, before the step of obtaining the relevant target simulator based on the customs clearance information, the method further includes:

[0032] Customs clearance categories for obtaining various customs clearance information;

[0033] The simulator is used to generate category datasets for each customs clearance category; wherein, the category datasets include customs clearance category data for multiple customs clearance categories;

[0034] The datasets for each category are input into a pre-set model for training, resulting in target models corresponding to each customs clearance category.

[0035] A cross-border trade process optimization device, the device comprising:

[0036] The acquisition module is used to acquire information about goods to be traded across borders and customs clearance information;

[0037] The generation module is used to generate multiple preset dimensions based on the cargo information and the customs clearance information;

[0038] The data acquisition module is used to collect dimension values ​​corresponding to the preset dimensions from the preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0039] The optimization module is used to optimize the state space using a preset learning algorithm to obtain an optimized customs clearance process.

[0040] A computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps:

[0041] Obtain information on goods to be traded across borders, as well as customs clearance information;

[0042] Multiple preset dimensions are generated based on the cargo information and the customs clearance information;

[0043] Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0044] The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process.

[0045] A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps:

[0046] Obtain information on goods to be traded across borders, as well as customs clearance information;

[0047] Multiple preset dimensions are generated based on the cargo information and the customs clearance information;

[0048] Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0049] The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process.

[0050] The beneficial effects of this invention are: it not only makes cross-border trade processes more flexible and adaptable, but also improves overall operational efficiency and shortens customs clearance time. Furthermore, the implementation of dynamic decision support enhances the system's intelligence, ensuring that enterprises maintain a competitive advantage in the ever-changing international trade environment. Through this innovative approach, enterprises can optimize resource allocation, reduce operating costs, and ultimately improve overall market responsiveness and customer satisfaction. Attached Figure Description

[0051] 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.

[0052] in:

[0053] Figure 1 This is a diagram illustrating the application environment of a cross-border trade process optimization method in one embodiment.

[0054] Figure 2 This is a flowchart of a cross-border trade process optimization method in one embodiment;

[0055] Figure 3 This is a structural block diagram of a cross-border trade process optimization device in one embodiment;

[0056] Figure 4 This is a structural block diagram of a computer device in one embodiment. Detailed Implementation

[0057] 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.

[0058] Figure 1 This is a diagram illustrating an application environment for cross-border trade process optimization in one embodiment. (Refer to...) Figure 1 This cross-border trade process optimization method is applied to a cross-border trade process optimization system. The system includes a terminal 110 and a server 120. The terminal 110 and server 120 are connected via a network. The terminal 110 can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet, or laptop. The server 120 can be a standalone server or a server cluster consisting of multiple servers. The terminal 110 is used for data collection, and the server 120 is used for cross-border trade process optimization.

[0059] like Figure 2 As shown, in one embodiment, a method for optimizing cross-border trade processes is provided. This method can be applied to both terminals and servers; this embodiment illustrates its application to a server. The specific steps of this cross-border trade process optimization method include:

[0060] S1: Obtain information on goods to be traded across borders and customs clearance information;

[0061] S2: Generate multiple preset dimensions based on the cargo information and the customs clearance information;

[0062] S3: Collect dimension values ​​corresponding to the preset dimensions from the preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0063] S4: Optimize the state space using a preset learning algorithm to obtain an optimized customs clearance process.

[0064] As described in step S1 above, information on the goods to be traded across borders and customs clearance information are obtained. Smooth cross-border trade requires accurate and comprehensive information on the goods and customs clearance. This necessitates collecting relevant data from multiple sources, including but not limited to the basic characteristics of the goods (such as product name, HS code, quantity, value, etc.) and the customs clearance requirements and policies of various countries (such as tariff standards, required documents, inspection procedures, etc.). Information can be acquired through interface integration, data crawling, and static or real-time data transmission from enterprise systems (such as ERP, WMS, etc.). For example, enterprises can obtain the latest tariffs and regulations through API interfaces connected to customs systems, while simultaneously using built-in information stored in databases and logistics systems to automatically aggregate all data, ensuring the timeliness and accuracy of the data. The collected information can lay the foundation for the entire optimization process, aiding subsequent decision-making and analysis.

[0065] As described in step S2 above, multiple preset dimensions are generated based on the cargo information and customs clearance information. After collecting the information, multiple preset dimensions are generated based on this cargo information and customs clearance information. These preset dimensions are the classification and feature extraction of the information, typically including variables such as commodity classification, destination country, mode of transport, declaration status, declaration progress, and inspection probability. This transforms complex, raw data into a multi-dimensional structure that can be quantitatively analyzed, facilitating subsequent processing and optimization. For example, when analyzing a certain type of commodity, dimensions might include the impact of regulations in different countries on the commodity, a comparison of the time cost and economics of different modes of transport, and the different customs clearance documents required in each case. In practice, machine learning or data mining techniques can be used to automatically identify and extract useful features. Data mining tools and statistical analysis can be used to build dimensional models suitable for business needs. Through accurate preset dimension generation, subsequent process analysis and decision-making will become more efficient and precise. It should be noted that the preset dimensions here refer to the dimensions that need to be acquired; the specific dimension information has not yet been obtained.

[0066] As described in step S3 above, the dimension values ​​corresponding to the preset dimensions are collected from the preset knowledge graph to form a state space. After generating the preset dimensions, the corresponding dimension information is obtained and transformed into a state space. In this step, the collected preset dimension data needs to be mapped to form a complete state space. The state space is the set of all possible states that the entire system can be in, providing a foundation for subsequent strategy optimization. A knowledge graph is a knowledge network represented graphically, which can associate different information. In this process, different dimension information becomes nodes in the knowledge graph, and the relationships between nodes are represented by edges. For example, the HS code of a product can be linked to the customs clearance requirements of the destination country, thus forming a more complex graph. In this state space, different states reflect various situations and scenarios that may exist under different conditions, and the system can also effectively identify potential risks and opportunities through the state space. Through this method, different data can be integrated into a unified model, facilitating subsequent data acquisition.

[0067] As described in S4 above, the state space is optimized using a preset learning algorithm to obtain an optimized clearance process. After establishing the state space, the preset learning algorithm is used to optimize the policy within that space, thereby refining a more efficient clearance process. The preset learning algorithm can be reinforcement learning, supervised learning, or other methods, aiming to continuously learn and optimize decisions through the exploration and utilization of the state space. For example, the Proximal Policy Optimization (PPO) reinforcement learning algorithm can be used. PPO is a reinforcement learning algorithm proposed by OpenAI and widely used in the field of deep reinforcement learning. It aims to optimize the policy while improving training stability by limiting the update magnitude. The core of this algorithm is to prevent the policy update from being too drastic through proximal optimization, thus making the training process more stable and reliable. Reinforcement learning strategies can be applied, and a reward function can be defined to drive the model to learn to prioritize the optimal clearance path and method. By training the model to perform multiple simulations and adjustments in the state space, the optimal operation mode for each step can be discovered. The optimized customs clearance process will be reflected in the found optimal strategy set, which will not only improve clearance speed but also enhance compliance and reduce costs. In this process, the coordination of accurate status representation, real-time data feedback, and learning algorithms is crucial. Through continuous learning and iteration, the process is optimized, enabling enterprises to more efficiently cope with complex international environments and policy changes in cross-border trade.

[0068] In one embodiment, before step S3 of collecting dimension values ​​corresponding to the preset dimension from a preset knowledge graph to form a state space, the method further includes:

[0069] S201: Obtain historical customs clearance data in cross-border trade;

[0070] S202: The historical customs clearance data is divided into structured data and unstructured data according to data type;

[0071] S203: Using a two-stream converter architecture, a pattern-aware encoder is used to extract field-level semantic features from the structured data, and a pre-trained language model is used to perform entity recognition on the unstructured data to obtain each target entity;

[0072] S204: Based on the aforementioned field-level semantic features, align the logistics trajectory coordinates with the target entity through a cross-modal attention mechanism to establish a spatial-semantic association matrix;

[0073] S205: Import each of the target entities and the spatial-semantic association matrix into a preset graph database to obtain the preset knowledge graph.

[0074] As described in step S201 above, historical customs clearance data for cross-border trade is collected from multiple channels. This data may cover a wide range of content, including declaration documents, encountered tariff policies, inspection and release status, and specific requirements of customs in different countries. Historical customs clearance data can be obtained from customs system API interfaces, the company's internal ERP system, document management system, or online trade platforms. During the information collection process, it is necessary to ensure the diversity and timeliness of the data to avoid information lag or inaccuracies in subsequent steps. For example, it is essential to track the latest tariff policies and import regulations in real time and identify the various customs clearance documents required. After obtaining the customs clearance data, it is necessary to perform preliminary data processing to facilitate subsequent analysis and processing.

[0075] As described in step S202 above, after acquiring the customs clearance data, the collected data is categorized into structured and unstructured data. This step is crucial because different types of data require different processing techniques and models. Structured data typically refers to data with a fixed format, such as tables in a database or CSV files, containing standardized information such as quantity, product type, and HS code. This type of data is easy to automate and analyze, and can be processed using a pattern-aware encoder. Unstructured data originates from text, documents, emails, etc., such as customs clearance application letters, contracts, email communications, and PDF documents. Since this data often lacks a specific format, it is more complex to process and requires the use of natural language processing (NLP) techniques to extract information. For example, NLP methods can be used for text analysis and information extraction. After this categorization, the data processing team can formulate corresponding processing strategies to ensure the smooth progress of subsequent feature extraction and data analysis. Clearly defining and categorizing data in the initial stage helps improve the accuracy and efficiency of subsequent processes.

[0076] As described in step S203 above, a two-stream transformer architecture is employed. A pattern-aware encoder is used to extract field-level semantic features from the structured data, and a pre-trained language model is used for entity recognition of the unstructured data to obtain various target entities. The two-stream transformer architecture, consisting of two independent processing paths, is used to process both structured and unstructured data respectively. The purpose of this architecture is to fully utilize the characteristics of each data type to extract more informative features. For structured data, a pattern-aware encoder is used to extract field-level semantic features. This encoder can understand the relationships between fields based on field patterns in the data table and extract key information. For example, it extracts information such as the name, quantity, and tax rate of goods, and transforms these features into high-dimensional vector representations for subsequent processing. For unstructured data, a pre-trained language model (such as BERT, RoBERTa, etc.) is used for entity recognition. This step uses NLP technology to identify key entities in the text, such as extracting customs clearance-related terms, product names, and shipping locations, thereby transforming scattered information into structured data. This provides rich content support for the subsequent construction of a knowledge graph.

[0077] As described in step S204 above, based on the field-level semantic features, a spatial-semantic association matrix is ​​established by aligning the logistics trajectory coordinates with the target entity through a cross-modal attention mechanism. This effectively combines data from different sources (i.e., logistics trajectories and customs clearance targets) to improve the relevance and consistency between data. The spatial-semantic association matrix provides an abstract expression of spatial relationships between data. Specifically, the logistics trajectory coordinates provide geographical location information about the goods during transportation, while the target entity refers to the specific items and requirements related to the customs clearance process. Through the cross-modal attention mechanism, the association between the two types of information can be analyzed, such as determining the transportation route of a product within a specific time period and its corresponding customs clearance requirements. For example, by comparing the coordinates in the logistics trajectory with the location of the target entity, the system can identify the customs clearance operations required at each transportation point and establish connections between various tasks. The resulting spatial-semantic association matrix not only provides data support for subsequent decision-making but also helps detect potential problems and risks, laying the foundation for optimizing cross-border trade processes.

[0078] As described in step S205 above, each target entity and the spatial-semantic association matrix are imported into a preset graph database to obtain the preset knowledge graph. The extracted target entities and the established spatial-semantic association matrix are then imported into the preset graph database to form a complete preset knowledge graph. A knowledge graph is an information database represented by a graph structure, allowing information to be stored, queried, and analyzed in a more flexible and dynamic way. By expressing information in the form of nodes and connections, a network is formed, making the relationships and associations between data visible. In the process of importing the target entities and association matrix into the graph database, each target entity is treated as a node, while the spatial-semantic association matrix is ​​used to define the edges between nodes, representing the relationships and interactions between different nodes. Using graph databases (such as Neo4j) enables efficient relational queries and data management, improving the efficiency and convenience of data access. By constructing a knowledge graph, relevant information of cross-border trade is systematized, enabling various queries, reasoning, and decisions to be implemented quickly, thus providing strong support for subsequent process optimization. Ultimately, the completed knowledge graph will lay the foundation for subsequent state-space-based decision optimization and strategy generation, improving the overall efficiency of cross-border trade processes.

[0079] In one embodiment, after step S201 of obtaining historical customs clearance data in cross-border trade, the method further includes:

[0080] S2021: Detect the customs clearance language of the historical customs clearance data;

[0081] S022: Obtain the corresponding translation model based on the customs clearance language;

[0082] S2023: Translate the historical customs clearance data based on the translation model to obtain historical customs clearance data in a specified language.

[0083] As described in steps S2021-S2023 above, identifying and determining the language used in the customs clearance data is crucial for accurate language detection, which forms the basis for subsequent processing. Since cross-border trade involves different countries and regions, customs clearance data may be presented in multiple languages; therefore, identifying the language of the data is essential. The language detection process typically uses language recognition technology, which can be implemented using Natural Language Processing (NLP) tools or libraries (such as LangDetect, Google Cloud Natural Language API, etc.). The system analyzes the input customs clearance text, using statistical features or pre-trained models to determine the specific language used. Based on language features, such as vocabulary distribution and syntactic structure, the system can quickly determine the language type of the text. This process not only improves the quality of subsequent translation, avoiding inaccuracies due to incorrect language identification, but also provides input for selecting and setting relevant parameters for subsequent translation models. Accurate language identification is a prerequisite for the effective processing and utilization of historical customs clearance data. In one embodiment, if the input is voice, the NVIDIA Riva framework can be integrated to customize acoustic models for cross-border technical terms (such as INCOTERMS2020), supporting end-to-end recognition of Southeast Asian dialects, thereby converting voice input into text and then analyzing the text.

[0084] After detecting the language of the customs clearance data, a corresponding translation model is obtained based on that language. A suitable translation tool is selected, and the data is translated into the original language of the historical customs clearance data to ensure its effectiveness and accuracy. Obtaining a translation model typically involves the following aspects: First, determining the required translation model for different languages ​​may require integrating multiple translation systems. For example, for foreign language customs clearance data, possible translation models include Google Translate, Microsoft Translator, or open-source translation frameworks (such as OpenNMT, Fairseq, etc.). Second, considering the applicability of the translation model ensures that the selected model can effectively handle specific industry and customs clearance terms, which usually requires specialized vocabulary tuning. Finally, when introducing deep learning-based translation models, the models can be trained and fine-tuned to enhance their performance in specific domains. Specifically, a multilingual BERT variant (XLM-RoBERTa) can be pre-built, and the semantic space of tariff terms in 107 languages ​​can be aligned through contrastive learning. Then, the T5-3B model is used for zero-shot translation, and the context-aware generation of templates such as "tax refund application - Spanish" can be achieved by combining the domain memory.

[0085] After determining the translation model, the final step is to translate the historical customs clearance data based on the model to obtain historical customs clearance data in the specified language. It should be noted that the translation process involves selecting a translation model corresponding to the original language of the customs clearance data and translating the text according to preset translation rules and parameters. The translation model processes the input customs clearance data sentence by sentence or paragraph by paragraph to generate corresponding target language content. In this process, the model does not simply translate word by word, but needs to consider context, grammatical structure, and technical terminology to ensure that the translation result meets customs clearance requirements in both semantics and usage.

[0086] Furthermore, a post-processing stage can be incorporated into the translation process to optimize the results. This includes correcting potential grammatical errors, improper punctuation, and further verifying industry-specific terminology to ensure the final output text meets customs clearance requirements. This step, which yields historical customs clearance data in the specified language, will greatly facilitate subsequent decision-making, process optimization, and efficient cross-border trade activities, enabling seamless data transmission and exchange and providing a guarantee for enterprises' international operations.

[0087] In one embodiment, after step S204 of importing each of the target entities and the spatial-semantic association matrix into a preset graph database to obtain the preset knowledge graph, the method further includes:

[0088] S2051: Monitor the historical customs clearance data in real time to see if there are any changes by using a preset crawler;

[0089] S2052: If a change has occurred, obtain the changed first target text and the original second target text;

[0090] S2053: Calculate the similarity between the first target text and the second target text;

[0091] S2054: Determine whether the similarity is greater than the threshold;

[0092] S2055: If the similarity is greater than the threshold, a GNN network is constructed to simulate the transmission effect, obtain the predicted entities affected, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph.

[0093] As described in steps S2051-S2055 above, a pre-set web crawler is used to monitor historical customs clearance data in real time to determine if the data has changed. A web crawler is an automated software program used to extract information from specific web pages or data sources. During the customs clearance process, relevant data may change frequently due to policy changes, regulatory updates, or changes in market conditions; therefore, a real-time monitoring mechanism is established. The pre-set web crawler typically accesses designated data sources periodically, such as the official customs website, trade-related databases, and various industry reports, monitoring changes in relevant data by obtaining HTML pages, CSV files, or API interfaces. Monitoring content may include tariff policies, updates to laws and regulations, and document requirements for customs clearance. When determining if data has changed, the crawler compares the currently acquired data with previously stored data. If discrepancies are found, these changes need further processing to ensure that future decisions and process optimizations reflect the latest realities. In this way, the timeliness and accuracy of the knowledge graph are maintained, improving the responsiveness of cross-border trade processes to market changes. After monitoring for data changes, the next step is to obtain the changed first target text and the original second target text. This stage corresponds to two target text data points at different times: the most recently crawled and updated data (first target text) and the previously stored data (second target text). First, the newly extracted data (serial data) is used as the first target text, typically covering the latest customs clearance policies, cargo information, and relevant laws and regulations. Second, the original, unchanged customs clearance information is retained as the second target text for discrepancy analysis. Direct comparison of these two texts effectively identifies specific changes, such as additions, modifications, or deletions of clauses. This detailed information acquisition process provides the necessary basis for subsequent similarity calculations and entity updates, ensuring the system can adapt to new changes in a timely manner, thus providing crucial support for the optimization and maintenance of cross-border trade processes. After acquiring the changed first target text and the unchanged second target text, the next step is to calculate the similarity between them. Similarity calculation is a crucial process for comparing the similarity of two text contents, aiming to determine the degree of change between the updated data and historical data. Calculating similarity typically uses text similarity algorithms from natural language processing, such as cosine similarity, Jaccard coefficient, and word embeddings (e.g., Word2Vec, BERT). Specifically, this involves converting two texts into vector representations and then using mathematical formulas to calculate a similarity score between them. For example, when using cosine similarity, the two texts are treated as vectors in a high-dimensional space, and the cosine of their angle is calculated, resulting in a similarity score ranging from 0 to 1; a higher value indicates greater text similarity.Similarity calculations can help quickly identify differences in key data during text updates, thereby determining whether the current knowledge graph needs to be updated to ensure its real-time performance and accuracy.

[0094] After calculating the similarity between the first and second target texts, the next step is to determine whether the similarity exceeds a threshold. This step is used to determine whether the change between historical and new data is significant, i.e., whether the similarity score exceeds a preset threshold. Setting a threshold avoids ineffective updates for minor changes, ensuring that the system only updates and adjusts the knowledge graph in cases of real and significant changes. If the similarity result is higher than the threshold, it indicates that the change between the two texts is not significant, usually considered a minor content adjustment or format correction, and the graph database does not need to be updated. However, if the similarity is lower than the threshold, it indicates a large difference between the texts, potentially signifying a major policy change, a significant update to customs clearance requirements, etc. In this case, further action is needed to ensure the system can dynamically respond and adjust relevant data in a timely manner. This judgment process is a crucial step in ensuring the effectiveness and accuracy of the knowledge graph, helping to improve the system's adaptability and effectiveness in cross-border trade processes.

[0095] After determining whether the similarity exceeds a set threshold, the final step is to construct a Graph Neural Network (GNN) to simulate the transmission effect if the similarity exceeds the threshold, obtain the predicted entities affected, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph. If the similarity exceeds the threshold, it indicates that there is a substantial change between historical data and new data. Therefore, a Graph Neural Network (GNN) is needed to further analyze the impact of these changes on the entire knowledge graph. The design of GNNs can handle graph-structured data and is very suitable for simulating the links between entities and dynamic information updates. Based on the current knowledge graph and newly acquired changed data, the GNN model analyzes how the changed content affects the associated nodes and uses the graph structure to mine potential affected entities. This step can effectively identify customs clearance policies, process nodes, or related news entities that may be affected by changes. After completing the prediction, the spatial-semantic association matrix in the graph database is updated according to these affected entities. This update ensures the timeliness and accuracy of the data, enabling the knowledge graph to always reflect the latest status of cross-border trade-related information. Through this comprehensive and systematic approach, cross-border trade can react quickly, optimize processes, and improve efficiency when facing policy and regulatory changes. Ultimately, the generated updated knowledge graph will provide solid data support and decision-making basis for subsequent applications.

[0096] In one embodiment, step S2, which generates multiple preset dimensions based on the cargo information and the customs clearance information, includes:

[0097] S211: Use a preset multimodal artificial intelligence model to process the cargo information and the customs clearance information to generate a process tree structure;

[0098] S212: Use a preset deep learning model to identify the nodes in the process tree structure, thereby obtaining multiple preset dimensions.

[0099] As described in steps S211-S212 above, the collected cargo information and customs clearance information are comprehensively processed using a preset multimodal artificial intelligence model, ultimately generating a JSON-formatted process tree structure. The advantage of the multimodal artificial intelligence model lies in its ability to simultaneously process multiple types of data, such as text, images, and audio, making it suitable for extracting and analyzing information in complex business scenarios. Specifically, when cargo information (such as commodity classification, HS code, quantity, shipper, etc.) and customs clearance information (such as customs clearance requirements, customs declaration documents, etc.) are input into the multimodal artificial intelligence model, the model performs feature extraction and representation learning through a multi-layered neural network architecture. This process includes data preprocessing, feature selection, and feature fusion. In addition to extracting keywords and sentences from the text, the model may also combine relevant visual data (such as commodity images) to enhance the depth of understanding. The extracted information is then transformed into a process tree structure. This structure displays each stage and its interrelationships in a hierarchical manner, starting from the root node and progressively showing each sub-task. For example, the root node may represent the entire customs clearance process, while the child nodes will display specific steps or conditions. By generating a process tree structure, the complex customs clearance process can be clearly captured and represented, making subsequent analysis and decision-making more concise and efficient. After generating the process tree structure, a pre-defined deep learning model is used to identify each node in the process tree to extract multiple pre-defined dimensions. The purpose of this process is to simplify the complex structure built through multimodal models into specific dimensions that can effectively support subsequent decision-making and analysis. Deep learning models, especially those using architectures such as Convolutional Neural Networks (CNNs) or Recurrent Neural Networks (RNNs), demonstrate excellent capabilities in identifying specific node information. These models are trained on large amounts of data to automatically identify and classify the types and characteristics of nodes. The model classifies each node and transforms it into specific pre-defined dimensions. These dimensions may include multiple key factors, such as the tax rate of the goods, the documents required for customs clearance, the declaration time, and the legal framework involved. By clearly defining and extracting these dimensions, conditions are created for subsequent optimization and analysis, and the foundation is laid for the efficient management of the entire cross-border trade process.

[0100] In one embodiment, step S4, which optimizes the state space using a preset learning algorithm to obtain an optimized customs clearance process, includes:

[0101] S401: Obtain the relevant target model based on the customs clearance information;

[0102] S402: Input the state space into the target model and set a preset value function in the target model to obtain the optimized customs clearance process.

[0103] As described in steps S401-S402 above, it is necessary to extract a target model related to optimizing the customs clearance process from the customs clearance information. The target model typically refers to a deep learning model. Trained based on historical data, it has captured past customs clearance processes under specific conditions and the numerous variables affecting the process output. Specifically, it can utilize deep learning, reinforcement learning, and other techniques to comprehensively consider different variables (such as cargo type, destination country regulations, freight costs, and timeliness) and simulate their impact on customs clearance efficiency. Through this training, the model can provide relatively accurate predictions for different strategies and their potential outcomes. After determining the target model, the selected state space is input into the target model, and a preset value function is set in the model to obtain the optimized customs clearance process. Specifically, various features in the state space (such as cargo information, customs clearance policies, and logistics information) are input into the target model in the form of input vectors. The model receives this state information and runs simulation algorithms to evaluate potential customs clearance decisions. The value function is used to evaluate the utility of performing a specific action under each state. In a specific implementation, it can combine three indicators: timeliness (R1), cost (R2), and compliance (R3), and sum these three indicators with weights. By setting this value function, the target model can reflect the expected benefits of different decision paths, thereby selecting the strategy most conducive to optimizing the customs clearance process. The optimization process may use policy optimization algorithms in reinforcement learning (such as PPO, Q-learning, etc.), adjusting the strategy selection by repeatedly comparing the benefits of different strategies, and finally determining the optimal customs clearance process.

[0104] In one embodiment, before step S401 of obtaining the relevant target simulator based on the customs clearance information, the method further includes:

[0105] S4001: Customs clearance categories for obtaining various customs clearance information;

[0106] S4002: Use the simulator to generate category datasets for each customs clearance category; wherein, the category datasets include customs clearance category data for multiple categories with the same customs clearance category;

[0107] S4003: Input the datasets of each category into the preset model for training to obtain the target model corresponding to each customs clearance category.

[0108] As described in steps S4001-S4003 above, it is necessary to obtain the customs clearance categories for various customs clearance information. This process involves classifying the collected customs clearance information to identify different types of customs clearance procedures and related requirements. For example, customs clearance categories may include various forms such as import clearance, export clearance, temporary import / export, and bonded zone clearance. Each type of customs clearance procedure has its specific policies, regulations, and operational requirements; therefore, identifying the customs clearance category is the foundation for subsequent data processing and model training. Obtaining customs clearance categories typically involves comparing and analyzing regulatory documents related to customs regulations and industry standards. Data analysis tools or technologies, such as text mining and natural language processing, can be used to identify and classify keywords and information in the documents.

[0109] After obtaining the clearance categories for various clearance information, a category dataset for each clearance category is generated using a simulator. A corresponding training dataset is then created for each clearance category for use in training the target model.

[0110] Generating category datasets using simulators is based on an understanding of real customs clearance data and processes, transforming them into simulated data. By setting rules and scenarios, the simulator can generate legal customs clearance data that conforms to specific category characteristics. For example, for import customs clearance, the simulator might generate corresponding data with real customs document requirements, cargo nature, and different shipping locations. The generated dataset should contain multiple samples, each representing the same category of customs clearance requirements, while also covering diverse cargo types and clearance situations to improve the model's generalization ability. The quantity and diversity of category datasets directly affect the training effect and accuracy of the target model; therefore, forming a sufficient and representative training set is crucial. This helps the model better understand the standards and processes of different customs clearance categories in subsequent steps, improving its application performance in real-world scenarios. Ultimately, the generated category datasets will provide rich data support for training the target models for each customs clearance category, ensuring the models can effectively cope with future customs clearance scenarios. After generating category datasets for each customs clearance category, each category dataset is input into a pre-set model for training, obtaining the target model corresponding to each customs clearance category. This process is the core of the entire customs clearance process optimization, aiming to build an intelligent model capable of effectively handling different customs clearance categories. The process of inputting the category dataset into the pre-defined model typically involves the following aspects: First, selecting a model architecture suitable for the current task, such as a deep learning model (e.g., convolutional neural networks, graph neural networks, or recurrent neural networks), to capture complex patterns in the data. Then, by setting appropriate hyperparameters (e.g., learning rate, batch size), the model is ensured to learn effectively during training. During training, the model continuously adjusts its internal parameters to minimize the loss function, thereby improving its ability to accurately identify and predict customs clearance categories. The training algorithm used can be stochastic gradient descent (SGD) or its variants, such as the Adam optimizer. Finally, the trained target model will correspond to a specific customs clearance category, enabling it to accurately identify, predict, and process future customs clearance requests. This targeted model allows the entire cross-border trade process to dynamically adapt to different customs clearance requirements, achieving optimization in practical applications, thereby improving process efficiency and reducing potential operational risks. Through this series of operations, enterprises can build stronger capabilities in customs clearance, ensuring efficient and compliant cross-border trade processes.

[0111] Reference Figure 3 The present invention also provides a cross-border trade process optimization device, the device comprising:

[0112] The acquisition module 902 is used to acquire information on goods to be traded across borders and customs clearance information;

[0113] Generation module 904 is used to generate multiple preset dimensions based on the cargo information and the customs clearance information;

[0114] The acquisition module 906 is used to acquire dimension values ​​corresponding to the preset dimensions from the preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0115] The optimization module 908 is used to optimize the state space through a preset learning algorithm to obtain an optimized customs clearance process.

[0116] In one embodiment, the cross-border trade process optimization device further includes:

[0117] The historical customs clearance data acquisition module is used to acquire historical customs clearance data in cross-border trade.

[0118] The historical customs clearance data segmentation module is used to classify the historical customs clearance data into structured data and unstructured data according to data type.

[0119] The feature recognition module is used to extract field-level semantic features from the structured data using a pattern-aware encoder and to perform entity recognition on the unstructured data using a pre-trained language model, thereby obtaining each target entity.

[0120] The matrix building module is used to align the logistics trajectory coordinates with the target entity based on the field-level semantic features and through a cross-modal attention mechanism to build a spatial-semantic association matrix.

[0121] The matrix import module is used to import each of the target entities and the spatial-semantic association matrix into a preset graph database to obtain the preset knowledge graph.

[0122] In one embodiment, the cross-border trade process optimization device further includes:

[0123] The customs clearance language detection module is used to detect the customs clearance language of the historical customs clearance data;

[0124] The translation model acquisition module is used to acquire the corresponding translation model based on the customs clearance language.

[0125] The translation module is used to translate the historical customs clearance data based on the translation model to obtain historical customs clearance data in a specified language.

[0126] In one embodiment, the cross-border trade process optimization device further includes:

[0127] The crawler monitoring module is used to monitor in real time whether the historical customs clearance data has changed through a preset crawler;

[0128] The target text acquisition module is used to acquire the first target text after the change and the second target text before the change if the change has occurred.

[0129] A similarity calculation module is used to calculate the similarity between the first target text and the second target text;

[0130] A similarity determination module is used to determine whether the similarity is greater than a threshold.

[0131] The matrix update module is used to construct a GNN network to simulate the transmission effect if the similarity is greater than a threshold, obtain the predicted entities affected by the prediction, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph.

[0132] In one embodiment, the generation module 904 includes:

[0133] The process tree structure generation submodule is used to process the cargo information and the customs clearance information using a preset multimodal artificial intelligence model to generate a process tree structure.

[0134] The process tree structure recognition submodule is used to identify nodes in the process tree structure using a preset deep learning model, thereby obtaining multiple preset dimensions.

[0135] In one embodiment, the optimization module 908 includes:

[0136] The target model acquisition submodule is used to acquire relevant target models based on the customs clearance information.

[0137] The state space input submodule is used to input the state space into the target model and set a preset value function in the target model to obtain the optimized customs clearance process.

[0138] In one embodiment, the optimization module 908 further includes:

[0139] The customs clearance category acquisition submodule is used to retrieve the customs clearance category for various customs clearance information.

[0140] The category dataset generation submodule is used to generate category datasets for each customs clearance category using the simulator; wherein, the category dataset includes customs clearance category data for multiple identical customs clearance categories;

[0141] The training submodule is used to input the datasets of each category into the preset model for training, so as to obtain the target model corresponding to each customs clearance category.

[0142] Figure 4 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 4 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program. When executed by the processor, this computer program enables the processor to implement a cross-border trade process optimization method. The internal memory may also store a computer program, which, when executed by the processor, enables the processor to implement the cross-border trade process optimization method. Those skilled in the art will understand that… Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0143] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the following steps:

[0144] Obtain information on goods to be traded across borders, as well as customs clearance information;

[0145] Multiple preset dimensions are generated based on the cargo information and the customs clearance information;

[0146] Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0147] The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process.

[0148] This not only makes cross-border trade processes more flexible and adaptable, but also improves overall operational efficiency and shortens customs clearance time. Furthermore, the implementation of dynamic decision support enhances the system's intelligence, ensuring that enterprises maintain a competitive edge in the ever-changing international trade environment. Through this innovative approach, enterprises can optimize resource allocation, reduce operating costs, and ultimately improve overall market responsiveness and customer satisfaction.

[0149] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform the following steps:

[0150] Obtain information on goods to be traded across borders, as well as customs clearance information;

[0151] Multiple preset dimensions are generated based on the cargo information and the customs clearance information;

[0152] Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions;

[0153] The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process.

[0154] This not only makes cross-border trade processes more flexible and adaptable, but also improves overall operational efficiency and shortens customs clearance time. Furthermore, the implementation of dynamic decision support enhances the system's intelligence, ensuring that enterprises maintain a competitive edge in the ever-changing international trade environment. Through this innovative approach, enterprises can optimize resource allocation, reduce operating costs, and ultimately improve overall market responsiveness and customer satisfaction.

[0155] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0156] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0157] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A cross-border trade process optimization method, characterized in that, The method includes: Obtain information on goods to be traded across borders, as well as customs clearance information; Multiple preset dimensions are generated based on the cargo information and the customs clearance information; Collect dimension values ​​corresponding to the preset dimensions from a preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions; The state space is optimized using a preset learning algorithm to obtain an optimized customs clearance process; Before the step of collecting dimension values ​​corresponding to the preset dimension from the preset knowledge graph to form a state space, the method further includes: Obtain historical customs clearance data in cross-border trade; The historical customs clearance data is categorized into structured data and unstructured data according to data type. A two-stream converter architecture is adopted to extract field-level semantic features from the structured data using a pattern-aware encoder, and to perform entity recognition on the unstructured data using a pre-trained language model to obtain each target entity. Based on the aforementioned field-level semantic features, a spatial-semantic association matrix is ​​established by aligning the logistics trajectory coordinates with the target entity through a cross-modal attention mechanism. Each target entity and the spatial-semantic association matrix are imported into a graph database with preset values ​​to obtain the preset knowledge graph; The historical customs clearance data is monitored in real time to see if any changes occur through a pre-set crawler. If changes occur, obtain the first target text after the change and the second target text before the change; Calculate the similarity between the first target text and the second target text; Determine whether the similarity is greater than a threshold; If the similarity is greater than the threshold, a GNN network is constructed to simulate the transmission effect, obtain the predicted entities affected, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph. The step of generating multiple preset dimensions based on the cargo information and the customs clearance information includes: The cargo information and customs clearance information are processed using a preset multimodal artificial intelligence model to generate a process tree structure; The nodes in the process tree structure are identified using a preset deep learning model, thereby obtaining multiple preset dimensions.

2. The cross-border trade process optimization method of claim 1, wherein, Following the step of obtaining historical customs clearance data in cross-border trade, the method further includes: Detect the customs clearance language of the historical customs clearance data; Obtain the corresponding translation model based on the customs clearance language; The historical customs clearance data is translated based on the translation model to obtain historical customs clearance data in a specified language.

3. The cross-border trade process optimization method of claim 1, wherein, The step of optimizing the state space using a preset learning algorithm to obtain the optimized customs clearance process includes: The relevant target model is obtained based on the customs clearance information; The state space is input into the target model, and a preset value function is set in the target model to obtain the optimized customs clearance process.

4. The cross-border trade process optimization method according to claim 3, characterized in that, Before the step of obtaining the relevant target simulator based on the customs clearance information, the method further includes: Customs clearance categories for obtaining various customs clearance information; The simulator is used to generate category datasets for each customs clearance category; wherein, the category datasets include customs clearance category data for multiple customs clearance categories; The datasets for each category are input into a pre-set model for training, resulting in target models corresponding to each customs clearance category.

5. A cross-border trade process optimization device, characterized in that, The device includes: The acquisition module is used to acquire information about goods to be traded across borders and customs clearance information; The generation module is used to generate multiple preset dimensions based on the cargo information and the customs clearance information; The historical customs clearance data acquisition module is used to acquire historical customs clearance data in cross-border trade. The historical customs clearance data segmentation module is used to classify the historical customs clearance data into structured data and unstructured data according to data type. The feature recognition module is used to extract field-level semantic features from the structured data using a pattern-aware encoder and to perform entity recognition on the unstructured data using a pre-trained language model to obtain each target entity. The matrix building module is used to align the logistics trajectory coordinates with the target entity based on the field-level semantic features and through a cross-modal attention mechanism to build a spatial-semantic association matrix. The matrix import module is used to import each of the target entities and the spatial-semantic association matrix into a preset graph database to obtain the preset knowledge graph. The crawler monitoring module is used to monitor in real time whether the historical customs clearance data has changed through a preset crawler; The target text acquisition module is used to acquire the first target text after the change and the second target text before the change if the change has occurred. A similarity calculation module is used to calculate the similarity between the first target text and the second target text; A similarity determination module is used to determine whether the similarity is greater than a threshold. The matrix update module is used to construct a GNN network to simulate the transmission effect if the similarity is greater than a threshold, obtain the predicted entities affected by the prediction, and update the spatial-semantic association matrix in the graph database to obtain the updated knowledge graph. The data acquisition module is used to collect dimension values ​​corresponding to the preset dimensions from the preset knowledge graph to form a state space; wherein, the state space is the space corresponding to all preset dimensions; The optimization module is used to optimize the state space using a preset learning algorithm to obtain an optimized customs clearance process. The generation module includes: The process tree structure generation submodule is used to process the cargo information and the customs clearance information using a preset multimodal artificial intelligence model to generate a process tree structure. The process tree structure recognition submodule is used to identify nodes in the process tree structure using a preset deep learning model, thereby obtaining multiple preset dimensions.

6. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, causes the processor to perform the steps of the cross-border trade process optimization method as described in any one of claims 1 to 4.

7. A computer device, characterized in that, The device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the cross-border trade process optimization method as described in any one of claims 1 to 4.

Images