A smart customs method and system based on artificial intelligence digital port
Through artificial intelligence and Internet of Things technology, customs declaration status tracking, manifest declaration and data chain integrity issues in electronic customs declaration services are solved, customs declaration status prediction, risk assessment and data analysis are realized, and operational efficiency and accuracy are improved.
Patent Information
- Application Number
- CN202411287628.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-09-14
AI Technical Summary
The existing electronic customs declaration services have problems such as difficulty in obtaining and tracking customs declaration status, manual repeated filling of manifest declarations, incomplete data chain and lack of customs declaration data analysis, resulting in low operational efficiency, poor accuracy and insufficient decision-making support.
Adopting a digital port smart customs service method based on artificial intelligence, through data cleaning, multi-source heterogeneous data fusion decision-making model, intelligent identification technology and Internet of Things equipment, customs declaration status prediction, risk assessment and automatic data entry, real-time monitoring and statistical analysis .
It realizes unified query, automated processing of customs declaration status, completeness and information transparency of data chains, reduces labor costs and error rates, and improves operational efficiency and decision-making support capabilities.
Smart Images

Figure CN118797449B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of artificial intelligence recommendation algorithms, and in particular to an artificial intelligence-based digital port smart customs method and system. Background Art
[0002] As the global economy becomes increasingly close, ports, as important hubs connecting domestic and foreign trade, have a direct impact on the smoothness of cross-border trade and the country's economic security through their operational efficiency and security. Port operations face many challenges. On the one hand, the traditional manual operation mode is inefficient and difficult to meet the growing demand for cross-border trade. On the other hand, the information island phenomenon between the various departments of the port is serious, resulting in the inability to effectively integrate and share data, affecting the accuracy and efficiency of supervision. In addition, cross-border trade involves many links and complex risks, which are difficult to deal with by traditional risk assessment and approval methods. Therefore, there is an urgent need to build a digital port smart customs method and system based on artificial intelligence, with the aim of improving port operation efficiency and ensuring trade security by integrating advanced artificial intelligence technology.
[0003] The existing electronic customs declaration service uses electronic information technology to complete customs declaration procedures, simplifies the cumbersome process of traditional paper customs declaration, can automatically verify data, and reduce human errors. Through platforms such as electronic ports and single windows, enterprises can submit customs declaration materials online and complete tax payment and other operations, but the existing electronic customs declaration service has the following problems.
[0004] Firstly, it is not easy for them to obtain and track the customs declaration status. Under the current customs declaration business model, operating companies are unable to track, query and obtain customs declaration status data. The customs declaration status node data (such as arrival, release, departure, tallying, customs cut-off, etc.) of most of the businesses of some regional companies are declared in the customs declaration system owned by the customs broker or in the single window of each local version, and the New Land-Sea Corridor Company cannot uniformly query the status of the customs declaration form; secondly, the manifest declaration is filled in manually and repeatedly. Under the current customs declaration business model, the manifest information still needs to be manually declared. At present, the manifest declaration data cannot be automatically generated through the business system and still relies on the manual filling of freight forwarders, which has a large workload, high labor costs, low manual input efficiency and high error rate; at the same time, under the current customs declaration business model, the data information is not transparent, and there is a lack of cloud-based customs declaration system. The operating company cannot supervise and analyze the customs declaration data of other regional companies from the overall management level, and the entire logistics and port data chain is incomplete and unsmooth; finally, the existing customs declaration system lacks customs declaration statistical data analysis. Since the current customs declaration business model cannot obtain the customs declaration data of various regional companies in a timely manner, the system lacks the ability to generate relevant data reports and conduct statistical analysis, and cannot analyze and summarize the business operations related to customs declaration of the operating company and regional companies and provide decision support. Summary of the invention
[0005] The purpose of the present invention is to solve the shortcomings of the above-mentioned prior art, and to provide a method and system for intelligent customs affairs based on artificial intelligence digital port, which introduces artificial intelligence into the traditional electronic customs declaration service to solve the problems of obtaining customs declaration status, manually filling in the manifest declaration repeatedly, incomplete data chain, and lack of customs declaration data. It realizes the intelligent provision of various reports and statistical analysis to meet the traceability of data at all levels. Enterprises can realize intelligent data docking through artificial intelligence customs affairs technology, reduce the amount of manual customs clearance data entry, improve accuracy, and further improve service levels.
[0006] On the one hand, the said intelligent customs method based on artificial intelligence digital port includes the following steps:
[0007] S1: Collecting historical data of customs declaration status to generate a customs declaration status information set, and performing data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, division of the customs declaration status information set, and labeling processing;
[0008] S2: Select a model for multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, input the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploy the customs declaration status prediction and classification model;
[0009] S3: Generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character, image and voice recognition technology, wherein the key information is set according to customs declaration requirements, and the customs declaration information is automatically entered into the manifest after data analysis, and the transportation status and location of the goods are automatically analyzed by inputting the customs declaration information into the customs declaration status prediction and classification model, and accurate customs declaration status information of the goods is provided;
[0010] S4: IoT devices are deployed at key locations of the customs to collect real-time data of goods, and the collected real-time data is transmitted to the IoT cloud platform through wireless communication technology for further analysis and sharing, and a risk assessment model is constructed based on historical data, customs declaration status information of goods calculated by the customs declaration status prediction and classification model, and the real-time data to perform real-time risk prediction and assessment of goods;
[0011] S5: Obtain customs declaration data from various sources, generate customs declaration data reports after standardization, feature selection and extraction operations on the customs declaration data from various sources, perform trade statistical analysis on the customs declaration data reports, and analyze and summarize the business operations related to customs declaration and provide decision support.
[0012] Furthermore, in step S1, the historical data related to customs declaration includes customs declaration information, commodity information, logistics information, transportation status and location information, and data cleaning and preprocessing of the customs declaration status information set further includes:
[0013] Performing data cleaning on the customs declaration status information set, including automatically filling in missing values, removing duplicates, and performing outlier detection and correction;
[0014] The cleaned data is preprocessed, including format conversion and normalization, data set division and labeling processing, wherein:
[0015] The format conversion and normalization include: using one-hot encoding to process categorical variables, converting non-numeric data in the data after the data cleaning into numeric data, and standardizing and normalizing the numeric data so that the various types of customs declaration-related data collected and managed meet unified naming and format requirements;
[0016] The data set division includes: dividing the data after the above processing into a training set, a validation set and a test set, and the division ratio of each data set is adjusted according to the specific situation;
[0017] The labeling process includes: labeling each customs declaration data with a corresponding label according to the customs declaration status, so as to perform supervised learning when training the customs declaration status prediction and classification model.
[0018] Furthermore, in step S2, selecting the multi-source heterogeneous data fusion decision model to build a customs declaration status prediction and classification model further includes:
[0019] S21: Use the model of multi-source heterogeneous data fusion decision-making to design the customs declaration status prediction and classification model, select the long short-term memory network LSTM model to process time series data, select the convolutional neural network CNN model to process the geographic image data of cargo transportation, and input the information processed by the two parts of the neural network structure into the feedforward neural network FNN model for customs declaration status prediction and classification, where:
[0020] Using the training set to train the customs declaration status prediction and classification model, adjusting the weights and biases in the network through a back propagation algorithm, and gradually improving the model;
[0021] The trained customs declaration status prediction and classification model is deployed to the production environment.
[0022] Further, in step S21, the long short-term memory network LSTM model is selected to process the time series data, the convolutional neural network CNN model is selected to process the geographic image data of cargo transportation, and the information processed by the two parts of the neural network structure is comprehensively input into the feedforward neural network FNN model for customs declaration status prediction and classification, which further includes:
[0023] S211: The time series data includes logistics information, customs declaration information, and location information. The collected time series data is preprocessed, and the preprocessed time series data is input into the long short-term memory network LSTM model using a sliding window method to extract sequence information, and the result is output to the feedforward neural network FNN model;
[0024] S212: The geographic image data includes video data collected on site, and the geographic image data is preprocessed, including normalizing pixel values and enhancing image quality, and the preprocessed geographic image data is input into the convolutional neural network CNN model, and image features are extracted after several convolutional layers, activation layers, and pooling layers are calculated and processed, and the results are output to the feedforward neural network FNN model;
[0025] S213: Concatenate the high-dimensional feature vectors processed by the long short-term memory network LSTM model and the convolutional neural network CNN model to form a unified feature representation, input it into the feedforward neural network FNN model, increase the nonlinear processing capability of the feedforward neural network FNN model through the Sigmoid function, and perform final prediction and classification.
[0026] Further, in step S3, the key information in the original customs declaration document is identified and extracted by intelligent character, image and voice recognition technology to generate customs declaration information, and the customs declaration information is automatically entered into the manifest after data analysis, specifically including:
[0027] S31: extracting text from the customs declaration document using OCR optical character recognition technology, and performing character recognition using an OCR model;
[0028] S32: Using semantic analysis technology to perform text segmentation, semantic analysis, and intent recognition on the text extracted by the OCR optical character recognition;
[0029] S33: Using AI image recognition technology to identify and extract data features from the image material of the customs declaration document;
[0030] S34: Use speech-to-text technology to record, upload, transcribe and process the declaration speech into text;
[0031] S35: Analyze the contents identified and extracted in steps S31 to S34, obtain the key information for storage, and automatically enter the key information into the corresponding position of the manifest.
[0032] Furthermore, in step S4, IoT devices are deployed at key locations of the customs to collect real-time data of goods, and the collected real-time data is transmitted to the IoT cloud platform through wireless communication technology for further analysis and sharing, specifically including:
[0033] S41: Deploy and use IoT devices in customs supervision areas, transport vehicles, and key locations of containers, and use artificial intelligence technologies such as image recognition and natural language processing to automatically identify goods and extract information;
[0034] S42: collecting real-time information of the goods, and transmitting the collected real-time data to the Internet of Things cloud platform through wireless communication technology, and the Internet of Things cloud platform processes and verifies the real-time data after receiving it;
[0035] S43: forwarding the processed data to a cloud computing platform, and using big data processing technology on the cloud computing platform to quickly analyze the real-time data;
[0036] S44: The cloud computing platform is used to integrate data from different sources to form a unified data view, and a unified data interface standard is established on the platform to facilitate real-time query and sharing of data by customs departments and enterprises.
[0037] Preferably, in step S4, based on historical data, the customs declaration status information of the goods calculated by the customs declaration status prediction and classification model, and the real-time data, a risk assessment model is constructed to perform real-time risk prediction and assessment on the goods, specifically including:
[0038] Set the historical data to , , the corresponding response variable is ,in, It means the customs declaration documents are normal. Indicates an abnormal state, uses the historical data to train a feedforward neural network f, and for the newly input real-time data X, the feedforward neural network f outputs a value If the output value Y of the feedforward neural network exceeds the threshold value 0.5, it is considered that an abnormal situation has occurred and the system issues a real-time warning;
[0039] At the same time, the customs declaration status information calculated by the customs declaration status prediction and classification model is compared If the deviation from the real-time data X exceeds the set prediction deviation, it is considered that an abnormal situation has occurred and the system issues a real-time warning.
[0040] Furthermore, in step S5, the customs declaration data from each source is subjected to standardization processing and feature selection and extraction operations, specifically including:
[0041] S51: First, all numerical features are identified in the customs declaration data. After cleaning the data, the data is standardized using the Z-score standardization method in the pandas library in Python. The Z-score standardization formula is as follows:
[0042] ,
[0043] Among them, x is the original data, is the mean of the feature, is the standard deviation of the feature;
[0044] S52: Use natural language processing technology (NLP) to extract text features, segment the text into separate phrases, calculate the frequency of each phrase in the text, use the TF-IDF algorithm to weight the word frequency to evaluate the importance of the phrase in the document, use the phrase embedding model to convert the text into a vector representation, and use the vector representation to capture the semantic relationship between phrases;
[0045] S53: Preliminarily screen out the features closely related to the problem, then calculate the correlation between the features and the target variable, evaluate the feature importance by calculating the Pearson correlation coefficient combined with the SHAP value of the feature, and then select the best feature subset based on the correlation and feature importance evaluation results. The calculation formula of the Pearson correlation coefficient is as follows:
[0046] ,
[0047] in, is the i-th data point of the selected feature, is the mean of the feature, is the i-th data point of the target variable, is the mean of the target variable;
[0048] For a given model f, feature i and a feature subset S without feature i are selected, and the SHAP value calculation formula of the feature is as follows:
[0049] ,
[0050] Where N is the number of all features.
[0051] Furthermore, in step S5, trade statistics analysis is performed on the customs declaration data report, and analysis and summary of the business operation related to customs declaration and decision support further include:
[0052] By analyzing the total import and export volume over a period of time, we can assess the expansion or contraction of trade scale;
[0053] Determine the growth rate of trade by calculating the year-on-year or month-on-month growth rate of the total volume of goods;
[0054] By drawing a trading time distribution chart, we can analyze the time distribution and mode of trading and understand the market activity and trading habits;
[0055] Through Geographic Information System (GIS) and big data analysis, we can intelligently analyze the distribution of trading partners and import and export ports, and then use the path optimization algorithm to optimize trade routes and logistics arrangements;
[0056] Through correlation analysis and based on machine learning models, we build price prediction models to analyze price trends and fluctuations and understand the dynamics and influencing factors of market prices.
[0057] On the other hand, the said artificial intelligence-based digital port smart customs system includes:
[0058] A customs declaration data preprocessing module, which is used to collect customs declaration status historical data to generate a customs declaration status information set, and to perform data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, data set division, and labeling processing;
[0059] A model building module selects a model of multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, inputs the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploys the model;
[0060] A customs declaration information entry module, which is used to generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character and image recognition technology, automatically entering the customs declaration information into the manifest after data analysis, automatically analyzing the transportation status and location of the goods by inputting the customs declaration information into the customs declaration status prediction and classification model, and providing accurate customs declaration status information;
[0061] A cloud platform sharing and risk assessment module is used to deploy IoT devices at key locations of the customs to collect real-time data of goods, transmit the collected real-time data to the IoT cloud platform through wireless communication technology for further analysis and sharing, and build a risk assessment model based on historical data, customs declaration status information of goods calculated by the customs declaration status prediction and classification model, and the real-time data to perform real-time risk prediction and assessment on the goods;
[0062] The customs declaration statistical data analysis module is used to timely obtain customs declaration data from various sources, generate customs declaration data reports after standardization, feature selection and extraction operations on the customs declaration data from various sources, conduct trade statistical analysis on the customs declaration data reports, and analyze and summarize the business operations related to customs declaration and provide decision support.
[0063] Compared with the prior art, the present invention has the following beneficial effects:
[0064] (1) The present invention tracks and monitors goods in real time and combines a neural network structure to build a customs declaration status prediction and classification model, automatically analyzes the transportation status and location of goods, and provides accurate customs declaration status information, thereby realizing unified inquiry of customs declaration status;
[0065] (2) The present invention automatically processes customs declaration documents by integrating multiple technical means, identifies and extracts key information from the documents, and automatically enters the identified key information into the manifest, thereby reducing workload and labor costs, improving entry efficiency, and reducing error rates;
[0066] (3) The present invention inputs massive customs declaration data into the risk assessment model for risk prediction and assessment, combines cloud computing and Internet of Things technology to conduct in-depth mining, analysis and interactive sharing, automatically detects violations, forged documents and fraudulent behavior, provides real-time risk warnings, realizes information transparency through cloud computing, and makes the entire logistics and port data chain complete and unobstructed;
[0067] (4) The present invention timely obtains customs declaration data from various sources, performs standardization processing and feature selection and extraction operations on the customs declaration data from various sources, generates relevant data reports, uses an intelligent analysis system to perform statistical analysis, and analyzes and summarizes the business operations related to customs declaration and provides decision support. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0069] Figure 1 This is a flow chart of a smart customs service method based on artificial intelligence digital port in the present invention;
[0070] Figure 2 A schematic diagram of the structure of a customs declaration status prediction and classification model of the present invention;
[0071] Figure 3 This is a schematic diagram of the structure of a long short-term memory network model of the present invention;
[0072] Figure 4 A schematic diagram of the structure of a convolutional neural network model of the present invention;
[0073] Figure 5 A neural network model training flow chart of the present invention;
[0074] Figure 6 This is a schematic diagram of the intelligent entry principle of manifest declaration information of the present invention;
[0075] Figure 7 This is a structural diagram of an intelligent customs system for digital ports based on artificial intelligence in the present invention;
[0076] Figure 8 This is a schematic diagram of an export customs declaration business process of the present invention;
[0077] Fig. 9 The present invention is a schematic diagram of an import customs declaration business process. DETAILED DESCRIPTION
[0078] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0079] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.
[0080] Introducing artificial intelligence into traditional electronic customs declaration services solves problems such as obtaining customs declaration status, manually filling in repeated manifest declarations, incomplete data chains, and lack of customs declaration data. Through artificial intelligence technology, various reports and statistical analysis can be intelligently provided to meet the traceability of data at all levels. Enterprises can use artificial intelligence customs technology to achieve intelligent data docking, reduce the amount of manual customs clearance data entry, improve accuracy, and further improve service levels.
[0081] The specific implementation of the present invention is described below with reference to the accompanying drawings and embodiments.
[0082] Example 1
[0083] like Figure 1As shown, a smart customs service method based on artificial intelligence digital port provided in this embodiment includes the following steps:
[0084] S1: Collecting historical data of customs declaration status to generate a customs declaration status information set, and performing data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, data set division, and labeling processing;
[0085] S2: Select a model for multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, input the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploy the customs declaration status prediction and classification model;
[0086] S3: Generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character, image and voice recognition technology, wherein the key information is set according to customs declaration requirements, and the customs declaration information is automatically entered into the manifest after data analysis, and the transportation status and location of the goods are automatically analyzed by inputting the customs declaration information into the customs declaration status prediction and classification model, and accurate customs declaration status information of the goods is provided;
[0087] S4: deploy IoT devices at key locations of the customs to collect real-time data of goods, transmit the collected real-time data to the IoT cloud platform through wireless communication technology for further analysis and sharing, and build a risk assessment model based on historical data, customs declaration status information of goods calculated by the customs declaration status prediction and classification model, and the real-time data to perform real-time risk prediction and assessment on the goods;
[0088] S5: timely obtain customs declaration data from various sources, generate customs declaration data reports after standardization, feature selection and extraction operations on the customs declaration data from various sources, conduct trade statistical analysis on the customs declaration data reports, and analyze and summarize the business operations related to customs declaration and provide decision support.
[0089] Wherein, in step S1, the historical data related to customs declaration includes customs declaration information, commodity information, logistics information, transportation status and location information, and a machine learning algorithm is used to clean the data, automatically fill in missing values, remove duplicates, and perform outlier detection and correction to improve data quality and accuracy;
[0090] The cleaned data is preprocessed, including format conversion and normalization, data set division and labeling. According to the customs informationization standard system, all types of data collected and managed by the customs are uniformly named and formatted to ensure the accuracy, completeness and comparability of the data.
[0091] The format conversion and normalization include: using one-hot encoding to process categorical variables, converting non-numeric data in the data after the data cleaning into numeric data, and standardizing and normalizing the numeric data so that the various types of customs declaration-related data collected and managed meet unified naming and format requirements;
[0092] The data set division includes: dividing the data after the above processing into a training set, a validation set and a test set, and the division ratio of each data set is adjusted according to the specific situation;
[0093] The labeling process includes: labeling each customs declaration data with a corresponding label according to the customs declaration status, so as to perform supervised learning when training the customs declaration status prediction and classification model.
[0094] Afterwards, artificial intelligence technology is used to automatically query the customs declaration status. Just enter relevant information (such as customs declaration number, bill of lading number, etc.), and the system can automatically obtain and integrate customs declaration status information from multiple data sources. By real-time tracking and monitoring of goods, combined with artificial intelligence algorithms such as decision trees, random forests, neural networks, etc., a customs declaration status prediction and classification model is built to automatically analyze the transportation status and location of goods and provide accurate customs declaration status information.
[0095] Furthermore, the model of multi-source heterogeneous data fusion decision-making is selected, and a customs declaration status prediction and classification model is constructed in combination with a neural network structure, such as Figure 2 As shown, the specific steps include:
[0096] S21: Use the model of multi-source heterogeneous data fusion decision-making to design the customs declaration status prediction and classification model, select the long short-term memory network LSTM model to process time series data, use the convolutional neural network CNN model to process the geographic image data of cargo transportation, and input the information processed by the two parts of the neural network structure into the feedforward neural network FNN model for customs declaration status prediction and classification, where:
[0097] Using the training set to train the customs declaration status prediction and classification model, adjusting the weights and biases in the network through a back propagation algorithm, and gradually improving the model;
[0098] The trained customs declaration status prediction and classification model is deployed to the production environment.
[0099] The innovation of this intelligent customs declaration prediction model is that it uses multi-source heterogeneous data for prediction. The prediction model can include various historical time series data and video data collected on site. Compared with the traditional prediction model that only uses a single data source for prediction, it has higher accuracy and stability.
[0100] Specifically, in this embodiment, the long short-term memory network LSTM model is selected to process time series data. The network structure is as follows Figure 3 As shown, For the input historical time series data, including logistics information, customs declaration information, location information, etc., the collected time series data is preprocessed, and the preprocessed time series data is input into the long short-term memory network using the sliding window method to extract sequence information, and the output data It will be connected to the subsequent feedforward neural network FNN model.
[0101] A convolutional neural network (CNN) model is selected to process the geographic image data of cargo transportation. The network structure is as follows: Figure 4 As shown, the geographic image data is first preprocessed, including normalizing pixel values and enhancing image quality, and the geographic image data is input into the convolutional neural network CNN model. Image features are extracted through several convolutional layers, activation layers and pooling layers, and the output data is connected to the subsequent feedforward neural network FNN model.
[0102] The high-dimensional feature vectors processed by the long short-term memory network and the convolutional neural network are spliced to form a unified feature representation, which is input into the feedforward neural network FNN model. The nonlinear processing capability of the feedforward neural network FNN model is increased through the Sigmoid function to perform final prediction and classification.
[0103] The training process of all the above neural network models is as follows Figure 5 As shown in the figure, a loss function is defined to measure the difference between the predicted value and the actual value, and the gradient of the loss function with respect to the weights and biases of the last layer is calculated. The gradient is back-propagated through the chain rule, and the weights and biases of each layer are updated to reduce the loss. An optimization algorithm such as the stochastic gradient descent algorithm is used to adjust the weights and biases in the network in order to minimize the loss function. We need to repeat the process of forward propagation, loss calculation, back-propagation, and parameter update multiple times, that is, multiple training iterations.
[0104] Furthermore, in step S3, the customs declaration documents are automatically processed by integrating multiple technical means, such as Figure 6 As shown in the figure, avoiding subjective human errors in key information not only improves efficiency, but also reduces the tediousness and repetitiveness of manual operations, including:
[0105] S31: extracting text from the customs declaration document using OCR optical character recognition technology, and performing character recognition using an OCR model;
[0106] S32: Using semantic analysis technology to perform text segmentation, semantic analysis, and intent recognition on the text extracted by the OCR optical character recognition;
[0107] S33: Using AI image recognition technology to identify and extract data features from the image material of the customs declaration document;
[0108] S34: Use speech-to-text technology to record, upload, transcribe and process the declaration speech into text;
[0109] S35: Analyze the contents identified and extracted in steps S31 to S34, obtain the key information for storage, and automatically enter the key information into the corresponding position of the manifest.
[0110] In this embodiment, artificial intelligence technology is used to automatically process a large amount of customs declaration documents and data, extract and organize key information, and quickly and accurately fill in customs declaration documents. With the help of big data analysis and machine learning technology, artificial intelligence can conduct in-depth mining and analysis of massive customs declaration data, and can automatically detect violations, forged documents and fraudulent behavior, and provide real-time risk warnings and decision support.
[0111] In step S4, IoT devices are deployed at key locations of the customs to collect real-time data of goods, and the collected real-time data is transmitted to the IoT cloud platform through wireless communication technology for further analysis and sharing. Specifically, it includes:
[0112] S41: Deploy and use IoT devices in customs supervision areas, transport vehicles, and key locations of containers, and use artificial intelligence technologies such as image recognition and natural language processing to automatically identify goods and extract information;
[0113] S42: collecting real-time information of the goods, and transmitting the collected real-time data to the Internet of Things cloud platform through wireless communication technology, and the Internet of Things cloud platform receives the real-time data and processes and verifies it;
[0114] S43: forwarding the processed data to a cloud computing platform, and using big data processing technology on the cloud computing platform to quickly analyze the real-time data;
[0115] S44: The cloud computing platform is used to integrate data from different sources to form a unified data view, and a unified data interface standard is established on the platform to facilitate real-time query and sharing of data by customs departments and enterprises.
[0116] This real-time information transmission and collaborative work enables customs departments and enterprises to work together more efficiently to solve problems in customs declaration and ensure the integrity of the data chain. Artificial intelligence can also optimize supply chain management, track the location of goods in real time, monitor inventory levels, and achieve close collaboration with suppliers and logistics companies, thereby further ensuring the integrity of the data chain.
[0117] Among them, we build a risk assessment model based on historical data, the customs declaration status information of the goods calculated by the customs declaration status prediction and classification model, and the real-time data, and train the risk assessment model to predict and assess the risks of real-time goods, specifically including:
[0118] Set the historical data to , , the corresponding response variable is ,in, It means the customs declaration documents are normal. Indicates an abnormal state, uses the historical data to train a feedforward neural network f, and for the newly input real-time data X, the feedforward neural network f outputs a value If the output value Y of the feedforward neural network exceeds the threshold value 0.5, it is considered that an abnormal situation has occurred and the system issues a real-time warning;
[0119] At the same time, the customs declaration status information calculated by the customs declaration status prediction and classification model is compared If the deviation from the real-time data X exceeds the set prediction deviation, it is considered that an abnormal situation has occurred and the system issues a real-time warning.
[0120] In step S5, customs declaration data is first collected through multiple channels such as a single window, enterprise ERP system, and third-party data providers to ensure the accuracy and completeness of the data source.
[0121] Furthermore, the standardization processing, feature selection and extraction operations of the customs declaration data from various sources specifically include:
[0122] S51: First, all numerical features such as amount, quantity, weight, volume, etc. are identified in the customs declaration data. After cleaning the data, in this embodiment, the Z-score standardization method is used to standardize the data in the pandas library in Python. The Z-score standardization formula is as follows:
[0123] ,
[0124] Among them, x is the original data, is the mean of the feature, is the standard deviation of the feature;
[0125] S52: Use natural language processing technology (NLP) to extract text features, segment the text into separate phrases, calculate the frequency of each phrase in the text, use the TF-IDF algorithm to weight the word frequency to evaluate the importance of the phrase in the document, use phrase embedding models such as Word2Vec, GloVe, and BERT to convert the text into a vector representation, and use the vector representation to capture the semantic relationship between phrases;
[0126] S53: Preliminarily screen out the features closely related to the problem, then calculate the correlation between the features and the target variable, evaluate the feature importance by calculating the Pearson correlation coefficient combined with the SHAP value of the feature, and then select the best feature subset based on the correlation and feature importance evaluation results. The calculation formula of the Pearson correlation coefficient is as follows:
[0127] ,
[0128] in, is the i-th data point of the selected feature, is the mean of the feature, is the i-th data point of the target variable, is the mean of the target variable;
[0129] For a given model f, feature i and a feature subset S without feature i are selected, and the SHAP value calculation formula of the feature is as follows:
[0130] ,
[0131] Where N is the number of all features.
[0132] Furthermore, in step S5, the intelligent analysis system is used to perform statistical analysis. The intelligent analysis system is used to determine the scale and growth rate of trade based on the total volume and growth rate of import and export trade, analyze the time distribution and mode of transactions, understand the market activity and transaction habits, analyze the distribution of trading partners and import and export ports, optimize trade routes and logistics arrangements, analyze price trends and fluctuations, and understand the dynamics and influencing factors of market prices, including:
[0133] By analyzing the total import and export volume over a period of time, we can assess the expansion or contraction of trade scale;
[0134] Determine the growth rate of trade by calculating the year-on-year or month-on-month growth rate of the total volume of goods;
[0135] By drawing a trading time distribution chart, we can analyze the time distribution and mode of trading and understand the market activity and trading habits;
[0136] Through Geographic Information System (GIS) and big data analysis, we can intelligently analyze the distribution of trading partners and import and export ports, and then use the path optimization algorithm to optimize trade routes and logistics arrangements;
[0137] Through correlation analysis and based on machine learning models, we build price prediction models to analyze price trends and fluctuations and understand the dynamics and influencing factors of market prices.
[0138] Example 2
[0139] In step S5, the customs declaration data from each source is standardized. In addition to the Z-score standardization method in Example 1, the Min-Max standardization method can also be used. The Min-Max standardization formula is as follows:
[0140] ,
[0141] Among them, x is the original data, and are the minimum and maximum values of the feature respectively.
[0142] Example 3
[0143] This embodiment provides an intelligent customs system based on artificial intelligence digital port, such as Figure 7 As shown, it includes several parts:
[0144] Comprehensive service portal: Build a comprehensive service portal for customs declaration service system for cargo owners / freight forwarders, provide online one-stop customs declaration entrustment services, adopt cloud technology and big data to create a SaaS model enterprise service system, establish an intelligent customs center with the whole process of customs clearance as the core, and realize the integrated service of the whole process of customs declaration business;
[0145] Decision analysis system: Build a decision analysis system for customs declaration service operating companies, provide query and acquisition of customs declaration data of companies in each region, so that operating companies can supervise and analyze the customs declaration data of companies in other regions from the overall management level, and promote the integrity and smoothness of the entire logistics and port data chain. Mainly including highway customs declaration data, railway customs declaration data, air transport customs declaration data, cargo customs declaration data, customs declaration node data, customs declaration statistical data;
[0146] Business operating system: The customs declaration service system provides business operating systems for companies in each region. Its main functions include order management, highway customs declaration management, railway customs declaration management, ocean customs declaration management, air customs declaration management, customs declaration document management, customs declaration status management, financial management, user center, etc.
[0147] Customs declaration management: Supports customers to fill in declarations online by themselves, provides customs declaration management for road, rail, sea and air transport channels, generates electronic documents online, and declares import and export declarations, original manifests, pre-allocated manifests and other documents through the national single window, so as to achieve unified coordination and management of port operations related to import and export enterprises and logistics business;
[0148] Customs declaration document management: Provide customs brokers with maintenance and query of relevant electronic documents pushed by import and export traders, realize one-stop logistics business entrustment, process the data information submitted by the enterprise once, and make customs work clear. Effectively connect the electronic documents such as work lists, cargo lists, customs declarations, etc. in the customs clearance process, realize the upload and download of various customs clearance documents, real-time synchronization, automatic data conversion and declaration. Mainly includes: cargo list management, packing list management, import customs declaration management, export customs declaration management, commercial invoice management, document intelligent identification;
[0149] Customs declaration status management: Provides comprehensive management of the customs declaration status of import and export goods. By connecting to the national single window, the customs declaration status information is updated and obtained synchronously, and can be queried by combination conditions such as order number and bill of lading number. It mainly includes customs declaration status acquisition, customs declaration status feedback, customs declaration status query, customs declaration status tracking, and customs declaration status statistics;
[0150] Financial management: Provide financial management for foreign trade logistics related enterprises to achieve transparency and visualization of enterprise expense settlement. Mainly used for customs brokers to maintain service bill information, set corresponding customs declaration billing rules, compare the statements sent by traders with the statements generated by themselves, and perform operations such as sending and confirmation;
[0151] Report management: The system provides customs brokers with statistical analysis of import and export traders’ customs declaration data. The main functions include: customs declaration quantity analysis, customs declaration cost analysis, customs declaration structure analysis, customs declaration trend analysis, and trade data analysis;
[0152] Third-party connection: information exchange with external systems such as the national single window, electronic port, customs, customs brokers, shippers / freight forwarders, etc.
[0153] The method of performing the method in Examples 1 and 2 mainly includes:
[0154] A customs declaration data preprocessing module, which is used to collect customs declaration status historical data to generate a customs declaration status information set, and to perform data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, data set division, and labeling processing;
[0155] A model building module selects a model of multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, inputs the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploys the model;
[0156] A customs declaration information entry module, which is used to generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character and image recognition technology, automatically entering the customs declaration information into the manifest after data analysis, automatically analyzing the transportation status and location of the goods by inputting the customs declaration information into the customs declaration status prediction and classification model, and providing accurate customs declaration status information;
[0157] A cloud platform sharing and risk assessment module is used to deploy IoT devices at key locations of the customs to collect real-time data of goods, transmit the collected real-time data to the IoT cloud platform through wireless communication technology for further analysis and sharing, and build a risk assessment model based on historical data, customs declaration status information of goods calculated by the customs declaration status prediction and classification model, and the real-time data to perform real-time risk prediction and assessment on the goods;
[0158] The customs declaration statistical data analysis module is used to timely obtain customs declaration data from various sources, generate customs declaration data reports after standardization, feature selection and extraction operations on the customs declaration data from various sources, conduct trade statistical analysis on the customs declaration data reports, and analyze and summarize the business operations related to customs declaration and provide decision support.
[0159] Among them, in this embodiment, in the more critical part, the process of export customs declaration business is as follows Figure 8 As shown in the figure, import customs declaration business is as follows Fig. 9 shown.
[0160] Through the application of artificial intelligence technology, the platform can intelligently provide various reports and statistical analysis to meet the traceability of data at all levels. At the same time, the platform can realize one-click generation of verification lists, customs declarations, etc., realizing intelligence and making the entire process more efficient. In addition, the smart port uses digital means to achieve information sharing, break information silos, optimize port trade processes, and thus improve port clearance efficiency.
[0161] Providing "one-stop" services, enterprises can complete all business through one platform, reducing the time and energy of running errands. In addition, the platform can also connect to enterprise ERP and other data systems to achieve intelligent data connection, reduce the amount of customs clearance data entry, improve accuracy, and further improve service levels.
[0162] At the same time, the platform can set various risk warnings to ensure operational safety. By automatically identifying risks and pre-processing them, the safe and stable operation of the port can be ensured. At the same time, the smart port can grasp the port operation status in real time through big data analysis and artificial intelligence technology, and promptly detect and handle abnormal situations, thereby enhancing the safety of the port.
[0163] Utilize artificial intelligence, machine learning and other technologies to achieve automated management of ports. This can not only improve port security and supervision, but also reduce the risk of human operational errors and omissions. On this basis, it can promote innovation, transformation and upgrading: smart ports are an important means to promote and accelerate innovation, transformation and upgrading of port management. By using the Internet of Things, sensors, cloud computing, decision analysis and optimization and other technical means, the key information of each core of the port can be thoroughly perceived, widely connected and deeply calculated, so as to achieve seamless connection and coordinated linkage between various resources and participants, and make intelligent responses to port management and operation.
[0164] Finally, it should be noted that the above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.
[0165] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described 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.
Claims
1. A smart customs method for digital ports based on artificial intelligence, characterized in that: The steps include: S1: Collecting historical data of customs declaration status to generate a customs declaration status information set, and performing data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, division of the customs declaration status information set, and labeling processing; S2: Select a model for multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, input the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploy the customs declaration status prediction and classification model; S3: Generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character, image and voice recognition technology, wherein the key information is set according to customs declaration requirements, and the customs declaration information is automatically entered into the manifest after data analysis, and the transportation status and location of the goods are automatically analyzed by inputting the customs declaration information into the customs declaration status prediction and classification model, and accurate customs declaration status information of the goods is provided; S4: deploying IoT devices at key locations of the customs to collect real-time data of goods, transmitting the collected real-time data to the IoT cloud platform through wireless communication technology for further analysis and sharing, and building a risk assessment model based on historical data, the customs declaration status information calculated by the customs declaration status prediction and classification model in step S3, and the real-time data to perform real-time risk prediction and assessment of goods; S5: Obtain customs declaration data from various sources, perform standardization, feature selection and extraction on the customs declaration data from various sources, generate a customs declaration data report, perform trade statistics analysis on the customs declaration data report, and analyze and summarize the business operation status related to customs declaration and provide decision support; In step S1, the historical data related to customs declaration includes customs declaration information, commodity information, logistics information, transportation status and location information, and data cleaning and preprocessing of the customs declaration status information set further includes: Performing data cleaning on the customs declaration status information set, including automatically filling in missing values, removing duplicates, and performing outlier detection and correction; The cleaned data is preprocessed, including format conversion and normalization, data set division and labeling processing, wherein: The format conversion and normalization include: using one-hot encoding to process categorical variables, converting non-numeric data in the data after the data cleaning into numeric data, and standardizing and normalizing the numeric data so that the various types of customs declaration-related data collected and managed meet unified naming and format requirements; The data set division includes: dividing the data after the above processing into a training set, a validation set and a test set, and the division ratio of each data set is adjusted according to the specific situation; The labeling process includes: labeling each customs declaration data with a corresponding label according to the customs declaration status, so as to perform supervised learning when training the customs declaration status prediction and classification model; In step S2, selecting the multi-source heterogeneous data fusion decision model to build a customs declaration status prediction and classification model further includes: S21: Use the model of multi-source heterogeneous data fusion decision-making to design the customs declaration status prediction and classification model, select the long short-term memory network LSTM model to process time series data, select the convolutional neural network CNN model to process the geographic image data of cargo transportation, and input the information processed by the two parts of the neural network structure into the feedforward neural network FNN model for customs declaration status prediction and classification, where: Using the training set to train the customs declaration status prediction and classification model, adjusting the weights and biases in the network through a back propagation algorithm, and gradually improving the model; Deploy the trained customs declaration status prediction and classification model into the production environment; In step S21, the long short-term memory network LSTM model is selected to process the time series data, the convolutional neural network CNN model is selected to process the geographic image data of cargo transportation, and the information processed by the two parts of the neural network structure is comprehensively input into the feedforward neural network FNN model for customs declaration status prediction and classification. Further including: S211: The time series data includes logistics information, customs declaration information, and location information. The collected time series data is preprocessed, and the preprocessed time series data is input into the long short-term memory network LSTM model using a sliding window method to extract sequence information, and the result is output to the feedforward neural network FNN model; S212: The geographic image data includes video data collected on site, and the geographic image data is preprocessed, including normalizing pixel values and enhancing image quality, and the preprocessed geographic image data is input into the convolutional neural network CNN model, and image features are extracted after several convolutional layers, activation layers, and pooling layers are calculated and processed, and the results are output to the feedforward neural network FNN model; S213: Concatenate the high-dimensional feature vectors processed by the long short-term memory network LSTM model and the convolutional neural network CNN model to form a unified feature representation, input it into the feedforward neural network FNN model, increase the nonlinear processing capability of the feedforward neural network FNN model through the Sigmoid function, and perform final prediction and classification.
2. The method for intelligent customs affairs at digital ports based on artificial intelligence according to claim 1 is characterized in that: In step S3, the key information in the original customs declaration document is identified and extracted by intelligent character, image and voice recognition technology to generate customs declaration information, and the customs declaration information is automatically entered into the manifest after data analysis, which specifically includes: S31: extracting text from the customs declaration document using OCR optical character recognition technology, and performing character recognition using an OCR model; S32: Using semantic analysis technology to perform text segmentation, semantic analysis, and intent recognition on the text extracted by the OCR optical character recognition; S33: Using AI image recognition technology to identify and extract data features from the image material of the customs declaration document; S34: Use speech-to-text technology to record, upload, transcribe and process the declaration speech into text; S35: Analyze the contents identified and extracted in steps S31 to S34, obtain the key information for storage, and automatically enter the key information into the corresponding position of the manifest.
3. The method for intelligent customs affairs at digital ports based on artificial intelligence according to claim 1 is characterized in that: In step S4, IoT devices are deployed at key locations of the customs to collect real-time data of goods, and the collected real-time data is transmitted to the IoT cloud platform through wireless communication technology for further analysis and sharing. Specifically, it includes: S41: Deploy and use IoT devices in customs supervision areas, transport vehicles, and key locations of containers, and use artificial intelligence technologies such as image recognition and natural language processing to automatically identify goods and extract information; S42: collecting real-time information of the goods, and transmitting the collected real-time data to the Internet of Things cloud platform through wireless communication technology, and the Internet of Things cloud platform processes and verifies the real-time data after receiving it; S43: forwarding the processed data to a cloud computing platform, and using big data processing technology on the cloud computing platform to quickly analyze the real-time data; S44: The cloud computing platform is used to integrate data from different sources to form a unified data view, and a unified data interface standard is established on the platform to facilitate real-time query and sharing of data by customs departments and enterprises.
4. The method for intelligent customs affairs at digital ports based on artificial intelligence according to claim 3 is characterized in that: In step S4, based on historical data, the customs declaration status information of the goods calculated by the customs declaration status prediction and classification model, and the real-time data, a risk assessment model is constructed to perform real-time risk prediction and assessment on the goods, specifically including: Set the historical data to , , the corresponding response variable is ,in, It means the customs declaration documents are normal. Indicates an abnormal state, and uses the historical data to train a feedforward neural network , for newly input real-time data , the feedforward neural network Output value , if the feedforward neural network Output value If the threshold exceeds 0.5, it is considered an abnormal situation and the system issues a real-time warning; At the same time, the customs declaration status information calculated by the customs declaration status prediction and classification model is compared And the real-time data ,If the deviation exceeds the set prediction deviation, it is considered that an abnormal situation has occurred, and the system issues a real-time warning.
5. The method for intelligent customs affairs at digital ports based on artificial intelligence according to claim 1 is characterized in that: In step S5, the customs declaration data from various sources are subjected to standardization processing and feature selection and extraction operations, specifically including: S51: First, all numerical features are identified in the customs declaration data. After cleaning the data, the data is standardized using the Z-score standardization method in the pandas library in Python. The Z-score standardization formula is as follows: , in, is the original data, is the mean of the feature, is the standard deviation of the feature; S52: Use natural language processing technology (NLP) to extract text features, segment the text into separate phrases, calculate the frequency of each phrase in the text, use the TF-IDF algorithm to weight the word frequency to evaluate the importance of the phrase in the document, use the phrase embedding model to convert the text into a vector representation, and use the vector representation to capture the semantic relationship between phrases; S53: Preliminarily screen out the features closely related to the problem, then calculate the correlation between the features and the target variable, evaluate the feature importance by calculating the Pearson correlation coefficient combined with the SHAP value of the feature, and then select the best feature subset based on the correlation and feature importance evaluation results. The calculation formula of the Pearson correlation coefficient is as follows: , in, is the first data points, is the mean of the feature, is the target variable data points, is the mean of the target variable; For a given model , selected features and does not contain features A subset of features , the SHAP value calculation formula of the feature is as follows: , in, is the number of all features.
6. The method for intelligent customs affairs at digital ports based on artificial intelligence according to claim 5 is characterized in that: In step S5, trade statistics analysis is performed on the customs declaration data report, and analysis and summary of the business operation related to customs declaration and decision support further include: By analyzing the total import and export volume over a period of time, we can assess the expansion or contraction of trade scale; Determine the growth rate of trade by calculating the year-on-year or month-on-month growth rate of the total volume of goods; By drawing a trading time distribution chart, we can analyze the time distribution and mode of trading and understand the market activity and trading habits; Through Geographic Information System (GIS) and big data analysis, we can intelligently analyze the distribution of trading partners and import and export ports, and then use the path optimization algorithm to optimize trade routes and logistics arrangements; Through correlation analysis and based on machine learning models, we build price prediction models to analyze price trends and fluctuations and understand the dynamics and influencing factors of market prices.
7. An artificial intelligence digital port smart customs system, which executes the artificial intelligence digital port smart customs method according to any one of claims 1 to 6, characterized in that: include: A customs declaration data preprocessing module, which is used to collect customs declaration status historical data to generate a customs declaration status information set, and to perform data cleaning and preprocessing on the customs declaration status information set, wherein the preprocessing includes format conversion and normalization, data set division, and labeling processing; A model building module selects a model of multi-source heterogeneous data fusion decision-making to build a customs declaration status prediction and classification model, inputs the cleaned and pre-processed customs declaration status information set into the customs declaration status prediction and classification model for training, and then deploys the model; A customs declaration information entry module, which is used to generate customs declaration information by identifying and extracting key information from the original customs declaration document through intelligent character and image recognition technology, automatically entering the customs declaration information into the manifest after data analysis, automatically analyzing the transportation status and location of the goods by inputting the customs declaration information into the customs declaration status prediction and classification model, and providing accurate customs declaration status information; A cloud platform sharing and risk assessment module is used to deploy IoT devices at key locations of the customs to collect real-time data of goods, transmit the collected real-time data to the IoT cloud platform through wireless communication technology for further analysis and sharing, and build a risk assessment model based on historical data, customs declaration status information of goods calculated by the customs declaration status prediction and classification model, and the real-time data to perform real-time risk prediction and assessment on the goods; The customs declaration statistical data analysis module is used to timely obtain customs declaration data from various sources, generate customs declaration data reports after standardization, feature selection and extraction operations on the customs declaration data from various sources, conduct trade statistical analysis on the customs declaration data reports, and analyze and summarize the business operations related to customs declaration and provide decision support.
