Marine freight ship navigation trajectory analysis method and device based on large text model

Abstract

The invention discloses a marine freighter navigation trajectory analysis method and device based on a large text model. The method comprises the steps of obtaining a marine freighter navigation data information set; the marine cargo ship navigation data information set comprises marine cargo ship text information and corresponding cargo ship track information; preprocessing the marine freight ship navigation data information set to obtain a training data set; processing a preset marine cargo ship track analysis model by using the training data set to obtain a trained marine cargo ship track analysis model; and processing the acquired cargo ship navigation text information by using the trained marine cargo ship track analysis model to obtain a cargo ship navigation track. Based on a text large model technology, key information, such as entities and relationships, is extracted from a large number of sea transportation data texts, support is provided for knowledge graph construction, intelligent question and answer systems and other systems, and instant, accurate and refined intelligent question and answer analysis of sea transportation is achieved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and text data processing, and in particular to a method and device for analyzing the navigation trajectory of a sea freighter based on a large text model. Background Art

[0002] As the cornerstone of international trade, shipping carries more than 90% of the world's trade volume. As of October 2024, the number of global container ships has grown to 7,124. Against this background, it is urgent to carry out shipping supervision in combination with the spatiotemporal distribution and behavioral characteristics of cargo ships. Existing methods for analyzing the navigation trajectory of shipping cargo ships usually rely on pre-set data analysis algorithms, which lack flexibility and accuracy when dealing with complex scenarios. Automatic Identification System (AIS) data is an important guarantee for ensuring the safe operation of shipping cargo ships. Based on AIS data, it is possible to conduct analysis of ship traffic flow, cargo flow, ship entry and exit, and sea passages. AIS dynamic data mainly includes ship latitude and longitude coordinates, speed, heading and other information.

[0003] How to effectively analyze and extract shipping data, realize instant, accurate and refined intelligent question-and-answer analysis of shipping, and obtain the navigation track of freight is an issue that needs to be solved urgently. Summary of the invention

[0004] The present invention mainly solves the problem of how to effectively analyze and extract shipping data and realize instant, accurate and refined intelligent question-answering analysis of shipping. The present invention discloses a method and device for analyzing the navigation trajectory of shipping cargo ships based on a large text model.

[0005] In a first aspect of the embodiment of the present application, a method for analyzing the navigation trajectory of a sea freighter based on a large text model is disclosed, comprising:

[0006] Obtaining a set of navigation data information of a sea freighter; the set of navigation data information of a sea freighter includes text information of the sea freighter and corresponding track information of the freighter;

[0007] Preprocessing the ocean freighter navigation data information set to obtain a training data set;

[0008] Using the training data set, processing the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model;

[0009] The trained ocean freighter track analysis model is used to process the collected freighter navigation text information to obtain the freighter navigation track.

[0010] The preprocessing of the marine cargo ship navigation data information set to obtain a training data set includes:

[0011] Performing data cleaning on the ocean freighter navigation data information set to obtain a first freighter navigation data information set;

[0012] Performing a category consistency check on the first cargo ship navigation data information set to obtain a second cargo ship navigation data information set;

[0013] The second freighter navigation data information set is subjected to data enhancement processing to obtain a training data set.

[0014] The step of performing data enhancement processing on the second cargo ship navigation data information set to obtain a training data set includes:

[0015] Using the large language model, rewriting each ocean freighter text information in the second freighter navigation data information set to obtain rewritten ocean freighter text information;

[0016] All the rewritten ocean freighter text information and the corresponding ocean freighter text information's freighter track information are added to the second freighter navigation data information set to obtain a training data set.

[0017] The ocean freighter track analysis model includes: a text recognition module, a semantic analysis module, a feature extraction module, and a track extraction module;

[0018] The text recognition module is used to convert text information input into the ocean freighter track analysis model into digital vector information;

[0019] The semantic analysis module is connected to the text recognition module and the feature extraction module respectively, and is used to perform semantic analysis on the input digital vector information to obtain a semantic information sequence;

[0020] The feature extraction module is used to extract features from the input semantic information sequence to obtain mutual attention features;

[0021] The track extraction module is connected to the feature extraction module and is used to perform track extraction processing on the mutual attention feature to obtain track information.

[0022] The feature extraction module includes a first residual multi-head self-attention submodule, a second residual multi-head self-attention submodule, a third residual multi-head self-attention submodule, a fourth residual multi-head self-attention submodule and a multi-head mutual attention submodule;

[0023] The first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are connected to the multi-head mutual attention submodule; the first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are respectively used to receive a semantic information sequence, and perform self-attention feature extraction on the received semantic information sequence to obtain corresponding feature information;

[0024] The multi-head mutual attention submodule is used to extract multi-head mutual attention features from the feature information output by each residual multi-head self-attention submodule to obtain mutual attention features;

[0025] The track extraction module includes a two-dimensional convolution submodule, an activation submodule, and a residual submodule; the two-dimensional convolution submodule, the activation submodule, and the residual submodule are connected in sequence; the input end of the two-dimensional convolution submodule serves as the input end of the track extraction module; the output end of the residual submodule serves as the output end of the track extraction module.

[0026] The method of using the training data set to process the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model includes:

[0027] Initialize the number of training iterations;

[0028] Inputting the text information of the ocean freighter in the training data set as input data into the ocean freighter track analysis model;

[0029] Using the ocean freighter track analysis model, the input data is processed to obtain a predicted value;

[0030] Performing difference calculation processing on the obtained predicted value and the cargo ship track information corresponding to the input data to obtain a difference value;

[0031] Determine whether the difference value satisfies a convergence condition, and obtain a first determination result;

[0032] When the first judgment result is no, judging whether the training iteration number value is equal to the training number threshold, and obtaining a second judgment result;

[0033] When the second judgment result is no, determining that the model training state does not meet the training termination condition;

[0034] When the second judgment result is yes, determining that the model training state satisfies the training termination condition;

[0035] When the first judgment result is yes, determining that the model training state satisfies a training termination condition;

[0036] When the model training status does not meet the termination training condition, the parameter update model is used to update the parameters of the ocean freighter track analysis model, the training iteration number value is increased by 1, and the execution is triggered to input the ocean freighter text information in the training data set as input data into the ocean freighter track analysis model;

[0037] When the model training status satisfies the training termination condition, the training process of the ocean freighter track analysis model is completed to obtain a trained ocean freighter track analysis model.

[0038] The expression of the parameter updating model is:

[0039]

[0040] θ←θ+v;

[0041] In the formula, is the difference value calculated for the i-th data in the training data set, v is the parameter update value, θ is the parameter of the marine cargo ship track analysis model, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, It represents the partial derivative with respect to the variable θ, and H1() represents the first-order Hermitian polynomial.

[0042] In a second aspect of an embodiment of the present invention, a device for analyzing the navigation trajectory of a shipping cargo ship based on a large text model is disclosed, the device comprising:

[0043] A memory storing executable program code;

[0044] a processor coupled to the memory;

[0045] The processor calls the executable program code stored in the memory to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0046] According to a third aspect of an embodiment of the present invention, a computer storable medium is disclosed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0047] According to a fourth aspect of an embodiment of the present invention, an information data processing terminal is disclosed, wherein the information data processing terminal is used to implement the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0048] The beneficial effects of the present invention are:

[0049] Based on the text big model technology, the present invention extracts key information, such as entities and relationships, from a large amount of shipping data text, provides support for knowledge graph construction, intelligent question and answer systems, and realizes instant, accurate and refined intelligent question and answer analysis of shipping.

[0050] The present invention provides a method and device for analyzing the navigation trajectory of a sea-going cargo ship based on a large text model. The model can adaptively learn and understand the relationship between different external input data and perform fusion processing, and realize flexible and accurate intelligent question and answer.

[0051] The present invention discloses a model for analyzing the track of a shipping freighter, which extracts semantic variables through a semantic analysis module, extracts track-related information from the semantic variables through a feature extraction module specifically defined for analyzing the track of a shipping freighter, and accurately extracts the track through a track extraction module. The model adopts a hierarchical structure to achieve accurate analysis and extraction of track information, and each layer can be updated independently, which has the advantage of a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart for implementing the method of the present invention. DETAILED DESCRIPTION

[0053] In order to better understand the content of the present invention, an embodiment is given here.

[0054] Figure 1 It is a flow chart for implementing the method of the present invention.

[0055] In a first aspect of the embodiment of the present application, a method for analyzing the navigation trajectory of a sea freighter based on a large text model is disclosed, comprising:

[0056] Obtaining a set of navigation data information of a sea freighter; the set of navigation data information of a sea freighter includes text information of the sea freighter and corresponding track information of the freighter;

[0057] Preprocessing the ocean freighter navigation data information set to obtain a training data set;

[0058] Using the training data set, processing the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model;

[0059] Using the trained ocean freighter track analysis model, the collected freighter navigation text information is processed to obtain the freighter navigation track;

[0060] The preprocessing of the marine cargo ship navigation data information set to obtain a training data set includes:

[0061] Performing data cleaning on the ocean freighter navigation data information set to obtain a first freighter navigation data information set;

[0062] Performing a category consistency check on the first cargo ship navigation data information set to obtain a second cargo ship navigation data information set;

[0063] Performing data enhancement processing on the second cargo ship navigation data information set to obtain a training data set;

[0064] The data cleaning process includes filling missing values, smoothing noise data, and smoothing or deleting outliers; the smoothing of noise data is to first identify the noise data, and then smooth the noise data according to the data before and after the noise data; the outlier point can be identified by using a preset data value range, and the data that is not within the value range is an outlier point. The filling value of the missing value can be determined by averaging the measured values ​​within a certain sampling interval before and after the missing value.

[0065] The category consistency check process is to determine whether the data attributes of each category of data in the first freighter navigation data information set belong to the data attributes of the category data, and obtain a judgment result; the data with a judgment result of no is deleted from the first freighter navigation data information set.

[0066] The data attributes include text attributes, numeric attributes, interval attributes, exhaustive attributes, etc.;

[0067] The step of performing data enhancement processing on the second cargo ship navigation data information set to obtain a training data set includes:

[0068] Using the large language model, rewriting each ocean freighter text information in the second freighter navigation data information set to obtain rewritten ocean freighter text information;

[0069] Adding all the rewritten ocean freighter text information and the corresponding ocean freighter text information's freighter track information to the second freighter navigation data information set to obtain a training data set;

[0070] The use of the large language model to rewrite each ocean freighter text information in the second freighter navigation data information set to obtain the rewritten ocean freighter text information may be:

[0071] Input the following sentence into the large language model to get the rewritten text information of the ocean freighter:

[0072] Please rewrite the following sentence and give three ways to describe it: + [Text information of ocean freighter]

[0073] Please reorganize the following sentences and provide three different ways to describe them: + [Text information of ocean freighter];

[0074] Please rewrite the following statements and provide three different ways of describing them: + [Text message of ocean freighter];

[0075] Please rephrase the following sentence and list three different ways to describe it: + [Text information of ocean freighter];

[0076] Please reorganize the following description and provide three different narrative modes: + [Text information of ocean freighter].

[0077] The large language model can be the Wenxinyiyan large model, the Tongyi large model, and the Zhipu large model.

[0078] The ocean freighter track analysis model includes: a text recognition module, a semantic analysis module, a feature extraction module, and a track extraction module;

[0079] The text recognition module is used to convert text information input into the ocean freighter track analysis model into digital vector information;

[0080] The semantic analysis module is connected to the text recognition module and the feature extraction module respectively, and is used to perform semantic analysis on the input digital vector information to obtain a semantic information sequence;

[0081] The feature extraction module is used to extract features from the input semantic information sequence to obtain mutual attention features;

[0082] The track extraction module is connected to the feature extraction module and is used to perform track extraction processing on the mutual attention feature to obtain track information.

[0083] The feature extraction module includes a first residual multi-head self-attention submodule, a second residual multi-head self-attention submodule, a third residual multi-head self-attention submodule, a fourth residual multi-head self-attention submodule and a multi-head mutual attention submodule;

[0084] The first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are connected to the multi-head mutual attention submodule; the first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are respectively used to receive a semantic information sequence, and perform self-attention feature extraction on the received semantic information sequence to obtain corresponding feature information;

[0085] The multi-head mutual attention submodule is used to extract multi-head mutual attention features from the feature information output by each residual multi-head self-attention submodule to obtain mutual attention features;

[0086] The track extraction module includes a two-dimensional convolution submodule, an activation submodule, and a residual submodule; the two-dimensional convolution submodule, the activation submodule, and the residual submodule are connected in sequence; the input end of the two-dimensional convolution submodule serves as the input end of the track extraction module; the output end of the residual submodule serves as the output end of the track extraction module.

[0087] The semantic analysis module is used to convert the variables corresponding to the natural language into formal semantic representation variables so that the machine can understand and process the meaning of the sentence. It can be implemented by using the Hidden Markov Model (HMM) and Conditional Random Field (CRF) model for semantic analysis. This type of model infers the semantic representation of the sentence by learning the frequency and probability in a large-scale corpus.

[0088] The semantic analysis module can be implemented using the scanner() function and the parser() function in Python.

[0089] The text recognition module can be implemented using a word2vector model;

[0090] The first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule can be a Transformer model.

[0091] The multi-head mutual attention sub-module is implemented using the Mutli Self-Attention module.

[0092] The activation submodule is used to implement an activation function, and the activation function is a ReLU activation function;

[0093] The residual submodule can be implemented using a Resnet network;

[0094] The method of using the training data set to process the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model includes:

[0095] Initialize the number of training iterations;

[0096] Inputting the text information of the ocean freighter in the training data set as input data into the ocean freighter track analysis model;

[0097] Using the ocean freighter track analysis model, the input data is processed to obtain a predicted value;

[0098] Performing difference calculation processing on the obtained predicted value and the cargo ship track information corresponding to the input data to obtain a difference value;

[0099] Determine whether the difference value satisfies a convergence condition, and obtain a first determination result;

[0100] When the first judgment result is no, judging whether the training iteration number value is equal to the training number threshold, and obtaining a second judgment result;

[0101] When the second judgment result is no, determining that the model training state does not meet the training termination condition;

[0102] When the second judgment result is yes, determining that the model training state satisfies the training termination condition;

[0103] When the first judgment result is yes, determining that the model training state satisfies a training termination condition;

[0104] When the model training status does not meet the termination training condition, the parameter update model is used to update the parameters of the ocean freighter track analysis model, the training iteration number value is increased by 1, and the execution is triggered to input the ocean freighter text information in the training data set as input data into the ocean freighter track analysis model;

[0105] When the model training status satisfies the training termination condition, the training process of the ocean freighter track analysis model is completed to obtain a trained ocean freighter track analysis model.

[0106] The difference value satisfies the convergence condition, which means that the difference value is less than a preset convergence threshold; the difference value does not satisfy the convergence condition, which means that the difference value is not less than a preset convergence threshold. The convergence threshold may be 0.5.

[0107] The difference calculation process may be implemented using a loss function.

[0108] The loss function may adopt a cross entropy loss function.

[0109] The expression of the parameter updating model is:

[0110]

[0111] θ←θ+v;

[0112] In the formula, is the difference value calculated for the i-th data in the training data set, v is the parameter update value, θ is the parameter of the marine cargo ship track analysis model, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, It means partial derivative with respect to variable θ, H1() means first-order Hermitian polynomial;

[0113] The method of using the trained ocean freighter track analysis model to process the collected freighter navigation text information to obtain the freighter navigation track includes:

[0114] The collected cargo ship navigation text information includes text sequences at several moments;

[0115] Get the text sequence at the current moment;

[0116] Using the trained ocean freighter track analysis model, the text sequence at the current moment is processed to obtain the freighter navigation track at N subsequent moments after the current moment;

[0117] Using the trained ocean freighter track analysis model, the text sequence between the current moment and the N moments before it is processed to obtain the freighter navigation track at the N+1 moment;

[0118] A circulant matrix is ​​constructed using the navigation trajectory of the cargo ship at the current moment and the N moments before and after the current moment; the i-th row of the circulant matrix is ​​the navigation trajectory of the cargo ship at the N+1 moment obtained by the text sequence of the i-1 moment after the current moment and the N moments before the i-1 moment;

[0119] Perform data enhancement processing on the circulant matrix to obtain the cargo ship's navigation trajectory at the current moment;

[0120] The data enhancement processing includes:

[0121] Extract the sub-diagonal vector of the circulant matrix;

[0122] The sampling sequence of the normalized Gaussian distribution function is used to perform inner product operation with the sub-diagonal vector to obtain the navigation trajectory of the cargo ship at the current moment.

[0123] Taking the current moment as the 10th moment as an example, the expression of the constructed circulant matrix is:

[0124]

[0125] Among them, the numbers in the matrix represent the location information of the cargo ship's navigation track at the current time data of the serial time.

[0126] The present invention uses a digital enhancement method to perform data enhancement processing on the detection results of a single text by constructing a circulant matrix. By comprehensively utilizing information from different sources, it can provide a more comprehensive data view than a single data, thereby increasing the accuracy and reliability of the system's decision-making. Data at different times complement each other's deficiencies, reducing the deviations or errors that may be introduced by relying on a single data source. In addition, data enhancement processing can rely on other modes to continue to operate normally when one mode is disturbed or fails, ensuring the continuity and robustness of the system's prediction.

[0127] In a second aspect of an embodiment of the present invention, a device for analyzing the navigation trajectory of a shipping cargo ship based on a large text model is disclosed, the device comprising:

[0128] A memory storing executable program code;

[0129] a processor coupled to the memory;

[0130] The processor calls the executable program code stored in the memory to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0131] According to a third aspect of an embodiment of the present invention, a computer storable medium is disclosed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called by a computer, they are used to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0132] According to a fourth aspect of an embodiment of the present invention, an information data processing terminal is disclosed, wherein the information data processing terminal is used to implement the method for analyzing the navigation trajectory of a sea freighter based on a large text model.

[0133] In a fifth aspect of the embodiments of the present invention, a method for analyzing the navigation trajectory of a shipping cargo ship based on a large text model is disclosed. The model can adaptively learn and understand the relationship between different external input data and perform fusion processing, and realize flexible and accurate intelligent question answering. The purpose of the present invention is achieved through the following technical solutions:

[0134] Construct a large text model for analyzing the navigation trajectory of ocean freighters, including:

[0135] The data input module is used to receive external questions, query the associated cargo ship AIS system data, and clean and classify the data;

[0136] Semantic analysis module, used to analyze and understand input questions, decompose questions into key elements such as time, object, and scope;

[0137] A transformation rule learning module, used to learn transformation rules from historical transformation data;

[0138] AIS data conversion module, which converts cargo ship AIS data into text based on the learned conversion rules;

[0139] The result output module combines the related information and is used to output the converted freighter navigation trajectory analysis results.

[0140] The above-mentioned large text model for analyzing the navigation trajectory of a shipping vessel includes:

[0141] The semantic analysis module uses deep learning technology to automatically identify and analyze question semantic information;

[0142] The conversion rule learning module adopts machine learning methods and can continuously optimize the conversion rules to improve conversion accuracy.

[0143] Model training data preparation: Based on the cargo ship route information publicly available on the Internet, an unsupervised dataset describing the shipping ship's navigation trajectory is constructed, and the dataset is imported using the dataset input module;

[0144] Based on the manual labeling method, "question-answer" pairs are constructed to provide a detailed description of the navigation trajectory of the ocean-going cargo ship.

[0145] Model training and tuning: Based on the self-attention mechanism and position encoding mechanism of the Transformer architecture, the semantic information and grammatical structure in the text are captured; the model is trained on a large amount of unlabeled data using self-supervised learning, so that it can learn data feature representation from the data;

[0146] Using the "pre-training + fine-tuning" method, based on manually annotated data sets, fine-tuning is performed on labeled data of specific question-answering tasks to train the transformation rule learning module to improve the model's generalization ability and rule generation accuracy.

[0147] Model reasoning analysis: Based on the usage requirements of the mission scenario, the model is used to conduct question-and-answer analysis on the navigation trajectory of the ocean-going cargo ship.

[0148] The above is only the embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for analyzing the navigation trajectory of a shipping vessel based on a large text model, characterized in that: include: Obtain a collection of navigation data information of ocean freighters; The ocean freighter navigation data information set includes ocean freighter text information and corresponding freighter track information; Preprocessing the ocean freighter navigation data information set to obtain a training data set; Using the training data set, processing the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model; The trained ocean freighter track analysis model is used to process the collected freighter navigation text information to obtain the freighter navigation track.

2. The method for analyzing the navigation trajectory of a shipping vessel based on a large text model according to claim 1, characterized in that: The preprocessing of the marine cargo ship navigation data information set to obtain a training data set includes: Performing data cleaning on the ocean freighter navigation data information set to obtain a first freighter navigation data information set; Performing a category consistency check on the first cargo ship navigation data information set to obtain a second cargo ship navigation data information set; The second freighter navigation data information set is subjected to data enhancement processing to obtain a training data set.

3. The method for analyzing the navigation trajectory of a shipping vessel based on a large text model as claimed in claim 2, characterized in that: The step of performing data enhancement processing on the second cargo ship navigation data information set to obtain a training data set includes: Using the large language model, rewriting each ocean freighter text information in the second freighter navigation data information set to obtain rewritten ocean freighter text information; All the rewritten ocean freighter text information and the corresponding ocean freighter text information's freighter track information are added to the second freighter navigation data information set to obtain a training data set.

4. The method for analyzing the navigation trajectory of a shipping vessel based on a large text model as claimed in claim 2, characterized in that: The ocean freighter track analysis model includes: a text recognition module, a semantic analysis module, a feature extraction module, and a track extraction module; The text recognition module is used to convert text information input into the ocean freighter track analysis model into digital vector information; The semantic analysis module is connected to the text recognition module and the feature extraction module respectively, and is used to perform semantic analysis on the input digital vector information to obtain a semantic information sequence; The feature extraction module is used to extract features from the input semantic information sequence to obtain mutual attention features; The track extraction module is connected to the feature extraction module and is used to perform track extraction processing on the mutual attention feature to obtain track information.

5. The method for analyzing the navigation trajectory of a sea freighter based on a large text model as claimed in claim 4, characterized in that: The feature extraction module includes a first residual multi-head self-attention submodule, a second residual multi-head self-attention submodule, a third residual multi-head self-attention submodule, a fourth residual multi-head self-attention submodule and a multi-head mutual attention submodule; The first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are connected to the multi-head mutual attention submodule; the first residual multi-head self-attention submodule, the second residual multi-head self-attention submodule, the third residual multi-head self-attention submodule, and the fourth residual multi-head self-attention submodule are respectively used to receive a semantic information sequence, and perform self-attention feature extraction on the received semantic information sequence to obtain corresponding feature information; The multi-head mutual attention submodule is used to extract multi-head mutual attention features from the feature information output by each residual multi-head self-attention submodule to obtain mutual attention features; The track extraction module includes a two-dimensional convolution submodule, an activation submodule, and a residual submodule; the two-dimensional convolution submodule, the activation submodule, and the residual submodule are connected in sequence; the input end of the two-dimensional convolution submodule serves as the input end of the track extraction module; the output end of the residual submodule serves as the output end of the track extraction module.

6. The method for analyzing the navigation trajectory of a sea freighter based on a large text model as claimed in claim 4, characterized in that: The method of using the training data set to process the preset ocean freighter track analysis model to obtain a trained ocean freighter track analysis model includes: Initialize the number of training iterations; Inputting the text information of the ocean freighter in the training data set as input data into the ocean freighter track analysis model; Using the ocean freighter track analysis model, the input data is processed to obtain a predicted value; Performing difference calculation processing on the obtained predicted value and the cargo ship track information corresponding to the input data to obtain a difference value; Determine whether the difference value satisfies a convergence condition, and obtain a first determination result; When the first judgment result is no, judging whether the training iteration number value is equal to the training number threshold, and obtaining a second judgment result; When the second judgment result is no, determining that the model training state does not meet the training termination condition; When the second judgment result is yes, determining that the model training state satisfies the training termination condition; When the first judgment result is yes, determining that the model training state satisfies a training termination condition; When the model training status does not meet the termination training condition, the parameter update model is used to update the parameters of the ocean freighter track analysis model, the training iteration number value is increased by 1, and the execution is triggered to input the ocean freighter text information in the training data set as input data into the ocean freighter track analysis model; When the model training status satisfies the training termination condition, the training process of the ocean freighter track analysis model is completed to obtain a trained ocean freighter track analysis model.

7. The method for analyzing the navigation trajectory of a sea freighter based on a large text model as claimed in claim 6, characterized in that: The expression of the parameter updating model is: θ←θ+v: In the formula, is the difference value calculated for the i-th data in the training data set, v is the parameter update value, θ is the parameter of the marine cargo ship track analysis model, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, It represents the partial derivative with respect to the variable θ, and H1() represents the first-order Hermitian polynomial.

8. A device for analyzing the navigation trajectory of a shipping cargo ship based on a large text model, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model as described in any one of claims 1 to 7.

9. A computer storable medium, characterized in that: The computer storable medium stores computer instructions, which, when called by a computer, are used to execute the method for analyzing the navigation trajectory of a sea freighter based on a large text model as described in any one of claims 1 to 7.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the method for analyzing the navigation trajectory of a sea freighter based on a large text model as described in any one of claims 1 to 7.

Images