Classification and identification method and system for ocean and related industries

Abstract

The present invention provides a method and system for classifying and identifying marine and related industries, comprising: step S1: establishing an enterprise credit information database; step S2: obtaining a standard document and performing text preprocessing on the standard document and a manual identification sample; step S3: establishing a marine-related enterprise identification model based on the manual identification sample and the standard document; step S4: using the marine-related enterprise identification model to identify the enterprise to be identified, obtaining a model identification sample, and calculating the identification accuracy and additional identification ratio of each marine industry by comparing the model identification sample with the manual identification sample; step S5: adjusting the marine-related enterprise identification model, using the adjusted marine-related enterprise identification model to identify the enterprise to be identified, and obtaining an enterprise classification result; and step S6: calculating an evaluation score for each identified marine industry in the enterprise classification result. The present invention can improve the accuracy of the identification result, has interpretability, and can display the classification basis.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing technology, and specifically, to a method for classifying and identifying ocean and related industries based on the public credit information of marine business entities, and more particularly to a method and system for classifying and identifying ocean and related industries. Background Art

[0002] When it comes to enterprise classification, manual investigation and identification are currently the most common method used in this field. This is labor-intensive, time-consuming, and subject to subjective factors, which can lead to biased recognition results. The more text a document contains, the less desirable this method is. When analyzing the connections between enterprises and industries, manually classifying enterprises into their corresponding categories is a time-consuming and labor-intensive task, especially when faced with a large number of enterprise samples requiring classification or when emerging industries face a lack of relevant classification experience. Automatic classification methods that use simple rules to specify keywords cannot guarantee classification accuracy, often resulting in misclassifications due to incomplete keyword coverage.

[0003] In existing related technologies, such as the invention patent with publication number CN112686043A, a word vector-based classification method for enterprises' emerging industries is disclosed. The invention obtains an input emerging industry and obtains relevant information on the Internet based on its name; uses the Textrank algorithm based on the emerging industry related information to obtain its candidate keywords; uses the K-means algorithm to cluster the candidate keywords to obtain emerging industry cluster keywords; obtains the business scope of the enterprise from the official website and obtains the enterprise business vocabulary based on the business scope; expands the emerging industry cluster keywords based on the enterprise business vocabulary to obtain the emerging industry keyword vocabulary; obtains the inverse document frequency weight of the words based on the enterprise business vocabulary; obtains the basic evaluation score, comprehensive evaluation score and enterprise classification score based on the business scope of the enterprise to be classified and the emerging industry keyword vocabulary; and obtains the emerging industry classification result based on the enterprise classification score. This invention only applies to one feature, the business scope of the enterprise, and does not use relevant national standards.

[0004] Patent publication number CN110019769A discloses an intelligent enterprise classification algorithm, comprising a text preprocessing process and a classification algorithm. The preprocessing process includes feature selection, word segmentation, stop word removal, and representative word selection. The classification algorithm is a machine learning algorithm that requires training with existing correctly classified data to obtain a reliable classifier for classifying new descriptive text. This invention can lead to biased recognition results when manual annotation samples are unevenly sampled, when the annotation results are inconsistent with the standard, or when samples of a certain category are missing. Summary of the Invention

[0005] In view of the defects in the prior art, the present invention provides a method and system for classifying and identifying ocean and related industries.

[0006] According to a method and system for classifying and identifying ocean and related industries provided by the present invention, the scheme is as follows:

[0007] In a first aspect, a method for classifying and identifying ocean and related industries is provided, the method comprising:

[0008] Step S1: Establish an enterprise credit information database, divide the enterprises therein into samples for manual identification and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification in the enterprise credit information database;

[0009] Step S2: obtaining a standard document, and performing text preprocessing on the standard document and the manual recognition sample respectively;

[0010] Step S3: Based on the manual identification samples and the standard documents, the feature combinations and association rules of the three feature fields of enterprise name, business scope, and national economic industry classification of enterprises in various marine industries are learned to establish a marine enterprise identification model;

[0011] Step S4: using the marine enterprise identification model to identify the enterprises to be identified, dividing them into n marine industries, obtaining model identification samples, and calculating the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples;

[0012] Step S5: adjusting the marine-related enterprise identification model according to the recognition accuracy and the additional recognition ratio, using the adjusted marine-related enterprise identification model to identify the enterprises to be identified, and classifying them into n marine industries to obtain enterprise classification results;

[0013] Step S6: Calculate an evaluation score for each identified marine industry for the enterprises in the enterprise classification results.

[0014] Preferably, step S2 includes:

[0015] Step S2.1: performing word segmentation extraction on the standard document to screen out standard marine industry characteristic words;

[0016] Step S2.2: Perform word segmentation and extraction on the business scope and company name in the manually identified samples, and use the TF-IDF algorithm to filter out sample marine industry characteristic words from the company keyword library.

[0017] Preferably, step S3 includes:

[0018] Step S3.1: combining standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words;

[0019] Step S3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents that the business scope text of the enterprise does not contain the characteristic word, and 1 represents that the characteristic word exists. After performing the above conversion on all samples, a one-hot matrix is ​​obtained.

[0020] Step S3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable, thus obtaining a classification model that can identify marine industries based on feature words.

[0021] Step S3.4: Derive the weight of the independent variables in the random forest model, which is called the feature weight of each feature word. According to the feature weight of the feature word, formulate multiple identification rules for each marine industry to form a marine-related enterprise identification model.

[0022] Preferably, step S4 includes:

[0023] Step S4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to perform fuzzy recognition of the enterprise name and business category;

[0024]

[0025] When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model;

[0026] Step S4.2: By comparing the model recognition samples and the manual recognition samples, the recognition accuracy and additional recognition ratio of each ocean and related industry are calculated.

[0027] Preferably, the step S4.2 includes:

[0028] The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows:

[0029]

[0030] The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises manually identified as marine industry A. It is used to evaluate the generalization ability of the marine enterprise identification model in real scenarios and is calculated as follows:

[0031]

[0032] The thresholds for the recognition accuracy and additional recognition ratio of each ocean and related industry are different to ensure that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the manually identified samples.

[0033] Preferably, adjusting the marine-related enterprise identification model in step S5 includes:

[0034] Step S5.1: Develop exclusion rules, including performing keyword segmentation on company names from the model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding companies that are not marine-related or whose main business is not in this marine industry, and controlling the excess recognition rate;

[0035] Step S5.2: Adjusting the identification rules, including adjusting the threshold set in the TF-IDF algorithm for screening feature words, adding or subtracting the marine industry feature words, and adjusting feature weights to add or subtract the identification rules for each marine industry;

[0036] Step S5.3: Perform manual sampling, and add or subtract the marine industry exclusion words and the marine industry characteristic words according to the manual sampling results;

[0037] Step S5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain an adjusted marine-related enterprise identification model;

[0038] At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance.

[0039] Preferably, step S6 includes:

[0040] Step S6.1: Using the descriptions of each marine industry in the standard document, a marine industry description vocabulary is formed;

[0041] Step S6.2: Calculating the assessment score of each identified marine industry by the enterprise in the enterprise classification results based on the text similarity between the marine industry description vocabulary and the business scope of the enterprise in the enterprise classification results;

[0042] If an enterprise is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise will be judged as belonging to the marine industry with a higher evaluation score based on the enterprise's evaluation score for each identified marine industry.

[0043] In a second aspect, a classification and identification system for ocean and related industries is provided, the system comprising:

[0044] Module M1: Establish an enterprise credit information database, divide the enterprises in it into manual identification samples and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification in the enterprise credit information database;

[0045] Module M2: obtaining a standard document, and performing text preprocessing on the standard document and the manual recognition sample respectively;

[0046] Module M3: Based on manual identification samples and the standard documents, learn the feature combinations and association rules of the three feature fields of enterprise name, business scope, and national economic industry classification for enterprises in various marine industries, and establish an identification model for marine-related enterprises;

[0047] Module M4: Use the marine enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, obtain model identification samples, and calculate the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples;

[0048] Module M5: adjusting the marine-related enterprise identification model according to the recognition accuracy and the additional recognition ratio, using the adjusted marine-related enterprise identification model to identify the enterprises to be identified, classifying them into n marine industries, and obtaining enterprise classification results;

[0049] Module M6: Calculate an assessment score for each identified marine industry for the enterprises in the enterprise classification results.

[0050] Preferably, the module M2 includes:

[0051] Module M2.1: Perform word segmentation extraction on the standard document to screen out standard marine industry characteristic words;

[0052] Module M2.2: Segment and extract the business scope and company name in the manually identified samples, and use the TF-IDF algorithm to filter out the sample marine industry characteristic words from the company keyword library;

[0053] The module M3 includes:

[0054] Module M3.1: Combine standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words;

[0055] Module M3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents the absence of the characteristic word in the business scope text of the enterprise, and 1 represents the presence of the characteristic word. After performing this conversion on all samples, a one-hot matrix is ​​obtained.

[0056] Module M3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable to obtain a classification model that can identify marine industries based on feature words.

[0057] Module M3.4: Derive the weights of the independent variables in the random forest model, which are called the feature weights of each feature word. Based on the feature weights of the feature words, formulate multiple identification rules for each marine industry to form a marine-related enterprise identification model.

[0058] Preferably, the module M4 includes:

[0059] Module M4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to fuzzy identify the enterprise name and business classification;

[0060]

[0061] When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model;

[0062] Module M4.2: Calculate the recognition accuracy and additional recognition ratio of each ocean and related industry by comparing the model recognition samples with the manual recognition samples;

[0063] The module M4.2 includes:

[0064] The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows:

[0065]

[0066] The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises manually identified as marine industry A. It is used to evaluate the generalization ability of the marine enterprise identification model in real scenarios and is calculated as follows:

[0067]

[0068] The thresholds for the recognition accuracy and additional recognition ratio for each ocean and related industry are different, ensuring that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the samples for manual identification.

[0069] Adjusting the marine enterprise identification model in the module M5 includes:

[0070] Module M5.1: Develop exclusion rules, including performing keyword segmentation and extraction of company names from model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding companies that are not marine-related or whose main business is not in this marine industry, and controlling the excess recognition rate;

[0071] Module M5.2: Adjusting the identification rules, including adjusting the threshold in the TF-IDF algorithm for filtering feature words, adding or removing feature words for the marine industry, and adjusting feature weights to add or remove identification rules for each marine industry;

[0072] Module M5.3: Perform manual sampling and, based on the results of the manual sampling, add or remove the marine industry exclusion words and the marine industry characteristic words;

[0073] Module M5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain the adjusted marine-related enterprise identification model;

[0074] At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance;

[0075] The module M6 includes:

[0076] Module M6.1: Use the descriptions of each marine industry in the standard documents to develop a marine industry description vocabulary;

[0077] Module M6.2: Calculate the assessment score of each identified marine industry by the enterprise in the enterprise classification results based on the textual similarity between the marine industry description vocabulary and the business scope of the enterprise in the enterprise classification results;

[0078] If an enterprise is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise will be judged as belonging to the marine industry with a higher evaluation score based on the enterprise's evaluation score for each identified marine industry.

[0079] Compared with the prior art, the present invention has the following beneficial effects:

[0080] 1. This invention solves the problem of poor interpretability of machine learning algorithms by combining them with keywords, and can also demonstrate the basis for identifying enterprise classifications.

[0081] 2. The present invention solves the problem of uneven sampling of manually labeled samples and inconsistency between labeling results and standards by taking into account both manually labeled samples and relevant national standard documents.

[0082] Other beneficial effects of the present invention will be explained through the introduction of specific technical features and technical solutions in the specific implementation methods. Those skilled in the art should be able to understand the beneficial technical effects brought about by the introduction of these technical features and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0084] Figure 1 It is a schematic diagram of the overall process of the present invention. DETAILED DESCRIPTION

[0085] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0086] The embodiment of the present invention provides a classification and identification method for marine and related industries, which uses a machine learning algorithm combined with a keyword method to automatically classify enterprises into 28 marine and related industry classifications and non-marine-related enterprises. Taking into account relevant national standards and manually labeled samples, multiple features such as enterprise name, business classification, and national economic industry classification are used. The recognition results are highly accurate and interpretable, and can display the basis for classification, which can solve the problems of biased manually labeled samples and incomplete classification results. Figure 1 As shown, the method specifically includes:

[0087] Step S1: Establish an enterprise credit information database, divide the enterprises therein into manually identified samples and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification.

[0088] Step S2: Obtain a standard document, and perform text preprocessing on the standard document and the manual recognition sample respectively.

[0089] Specifically, the standard documents in this step include the "Classification of Marine and Related Industries" (GB / T20794-2021) and the "Technical Regulations for Identifying Marine-Related Entities" (Draft for Comment). Word segmentation and extraction are performed on these standard documents, and standard marine industry feature words are screened using a large language model (LLM) understanding and manual screening. Word segmentation and extraction are performed on the business scope and company name in the manually identified samples, and the TF-IDF algorithm is used to screen sample marine industry feature words from the company keyword library.

[0090] Step S3: Based on manual identification samples and standard documents, learn the feature combinations and association rules of the three feature fields of enterprise name, business scope, and national economic industry classification for each marine industry enterprise, and establish a marine enterprise identification model. This step specifically includes:

[0091] Step S3.1: combining standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words;

[0092] Step S3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents that the business scope text of the enterprise does not contain the characteristic word, and 1 represents that the characteristic word exists. After performing the above conversion on all samples, a one-hot matrix is ​​obtained.

[0093] Step S3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable to obtain a classification model that can identify marine industries based on feature words.

[0094] Step S3.4: Derive the weights of the independent variables in the random forest model, which are called the feature weights of each feature word. Based on the feature weights of the feature words, multiple identification rules are developed for each marine industry to form a marine enterprise identification model.

[0095] Step S4: Use the marine enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, obtain model identification samples, and calculate the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples.

[0096] Step S4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to perform fuzzy recognition of the enterprise name and business category;

[0097]

[0098] When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model;

[0099] Step S4.2: By comparing the model recognition samples and the manual recognition samples, the recognition accuracy and additional recognition ratio of each ocean and related industry are calculated.

[0100] The step S4.2 includes:

[0101] The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows:

[0102]

[0103] The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises manually identified as marine industry A. It is used to evaluate the generalization ability of the marine enterprise identification model in real scenarios and is calculated as follows:

[0104]

[0105] The thresholds for the recognition accuracy and additional recognition ratio of each ocean and related industry are different to ensure that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the manually identified samples.

[0106] Step S5: Adjust the marine-related enterprise identification model, use the adjusted marine-related enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, and obtain enterprise classification results.

[0107] Specifically, adjusting the marine enterprise identification model in step S5 includes:

[0108] Step S5.1: Develop exclusion rules, including performing keyword segmentation on company names from the model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding companies that are not marine-related or whose main business is not in this marine industry, and controlling the excess recognition rate;

[0109] Step S5.2: Adjusting the recognition rules, including adjusting the set threshold in the TF-IDF algorithm for screening feature words, adding or subtracting the marine industry feature words, and adjusting the feature weights to add or subtract the recognition rules for each marine industry;

[0110] Step S5.3: Perform manual sampling, and add or subtract the marine industry exclusion words and the marine industry characteristic words according to the manual sampling results;

[0111] Step S5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain an adjusted marine-related enterprise identification model;

[0112] At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance.

[0113] Step S6: Calculate an assessment score for each identified marine industry for the enterprises in the enterprise classification results.

[0114] Specifically, step S6 includes:

[0115] Step S6.1: Using the descriptions of each marine industry in the standard document, a marine industry description vocabulary is formed;

[0116] Step S6.2: Calculate the enterprise's assessment score for each identified marine industry in the enterprise classification results based on the textual similarity between the marine industry description vocabulary and the business scope of the enterprise in the enterprise classification results. If an enterprise is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise's assessment score for each identified marine industry is used to determine whether the enterprise belongs to the marine industry with the highest assessment score.

[0117] The present invention also provides a system for classifying and identifying marine and related industries. This system can be implemented by executing the process steps of the method for classifying and identifying marine and related industries. Those skilled in the art can understand the method as a preferred embodiment of the system for classifying and identifying marine and related industries. The system specifically includes the following:

[0118] Module M1: Establish an enterprise credit information database, divide the enterprises therein into manually identified samples and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification.

[0119] Module M2: Obtain a standard document, and perform text preprocessing on the standard document and the manual recognition sample respectively.

[0120] Specifically, the module includes standard documents such as the "Classification of Marine and Related Industries" (GB / T20794-2021) and the "Technical Regulations for Identifying Marine-Related Entities" (Draft for Comment). This module performs word segmentation and extraction on these standard documents, using a large language model and manual screening to identify standard marine industry key words. The module also performs word segmentation and extraction on the business scope and company name in the manually identified samples, using the TF-IDF algorithm to filter out sample marine industry key words from the company keyword library.

[0121] Module M3: Based on manual identification samples and standard documents, learn the feature combinations and association rules of the three characteristic fields of enterprise name, business scope, and national economic industry classification for enterprises in various marine industries, and establish an identification model for marine-related enterprises. This module specifically includes:

[0122] Module M3.1: Combine standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words;

[0123] Module M3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents the absence of the characteristic word in the business scope text of the enterprise, and 1 represents the presence of the characteristic word. After performing this conversion on all samples, a one-hot matrix is ​​obtained.

[0124] Module M3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable to obtain a classification model that can identify marine industries based on feature words.

[0125] Module M3.4: Derive the weights of the independent variables in the random forest model, referred to as the feature weights of each feature word. Based on the feature weights of the feature words, develop multiple identification rules for each marine industry to form a marine enterprise identification model.

[0126] Module M4: Use the marine enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, obtain model identification samples, and calculate the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples.

[0127] Module M4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to fuzzy identify the enterprise name and business classification;

[0128]

[0129] When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model;

[0130] Module M4.2: By comparing the model recognition samples and manual recognition samples, calculate the recognition accuracy and additional recognition ratio of each ocean and related industry.

[0131] The module M4.2 includes:

[0132] The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows:

[0133]

[0134] The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises manually identified as marine industry A. It is used to evaluate the generalization ability of the marine enterprise identification model in real scenarios and is calculated as follows:

[0135]

[0136] The thresholds for the recognition accuracy and additional recognition ratio of each ocean and related industry are different to ensure that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the manually identified samples.

[0137] Module M5: Adjust the marine enterprise identification model, use the adjusted marine enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, and obtain the enterprise classification results.

[0138] Specifically, the adjustment of the marine enterprise identification model in module M5 includes:

[0139] Module M5.1: Develop exclusion rules, including performing keyword segmentation and extraction of company names from model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding companies that are not marine-related or whose main business is not in this marine industry, and controlling the excess recognition rate;

[0140] Module M5.2: Adjusting the recognition rules, including adjusting the threshold set in the TF-IDF algorithm for screening feature words, adding or removing the marine industry feature words, and adjusting the feature weights to add or remove the recognition rules for each marine industry;

[0141] Module M5.3: Conduct manual sampling and, based on the results of the manual sampling, add or remove marine industry exclusion words and marine industry characteristic words;

[0142] Module M5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain the adjusted marine-related enterprise identification model;

[0143] At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance.

[0144] Module M6: Calculate the assessment score for each identified marine industry for the enterprises in the enterprise classification results.

[0145] Specifically, the module M6 includes:

[0146] Module M6.1: Use the descriptions of each marine industry in the standard documents to develop a marine industry description vocabulary;

[0147] Module M6.2: Calculate the enterprise's assessment score for each identified marine industry in the enterprise classification results based on the textual similarity between the marine industry description vocabulary and the enterprise's business scope in the enterprise classification results. If a company is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise's assessment score for each identified marine industry will be used to determine whether it belongs to the marine industry with the higher assessment score.

[0148] Next, the present invention will be described in more detail.

[0149] The present invention provides a classification and identification method for marine and related industries. This method has high recognition accuracy, is explainable, can display the classification basis, and can solve the problems of biased manual labeling samples and incomplete classification results. Specifically, it includes the following contents:

[0150] Step S1: Establish an enterprise credit information database and divide the enterprises therein into "manual identification samples" and "enterprises to be identified." Select multiple "feature fields" in the enterprise credit information database, including enterprise name, business scope, and national economic industry classification.

[0151] Step S2: Obtain the standard documents, namely the "Classification of Marine and Related Industries" (GB / T20794-2021) and the "Technical Regulations for the Identification of Marine-Related Units" (Draft for Comment). Use a large language model (LLM) to understand the descriptions and classifications of each marine industry in these documents, and summarize descriptive sentences that match the characteristics of each marine industry. Then, manually select a series of characteristic words that best represent each industry from these descriptive sentences based on semantics, forming the "standard marine industry characteristic words."

[0152] Each marine industry category in the "manually identified samples" was divided into a corresponding "corpus." The text in the "corpus" was segmented to obtain the number of occurrences of each feature word in each "corpus." Based on the semantic validity of the feature words, a manual list of feature words that are not helpful for industry classification was created, including "stop words" such as "project," "general," "is," and punctuation marks. This constituted a "stop word list." Stop words in the "corpus" were excluded based on the stop word list.

[0153] The TF-IDF algorithm is used to screen feature words. The specific steps include calculating the TF-IDF value of each feature word, sorting the feature words according to the calculated TF-IDF value, and screening feature words with high TF-IDF values ​​by setting a threshold to form "sample marine industry feature words."

[0154] TF-IDF (term frequency–inverse document frequency) is a commonly used weighting technique for information retrieval and text mining. TF-IDF is a statistical method used to assess the importance of a word to a document set or a document in a corpus. The importance of a word increases in direct proportion to the number of times it appears in a document, but decreases in inverse proportion to the frequency of its appearance in the corpus. The main idea of ​​TF-IDF is that if a word appears with a high frequency TF in one article and rarely appears in other articles, it is considered that this word or phrase has good category discrimination ability and is suitable for classification.

[0155] The TF-IDF calculation formula is as follows:

[0156] TF-IDF=TF·IDF

[0157] TF-IDF is the product of TF and IDF; TF stands for term frequency; IDF stands for inverse document frequency.

[0158]

[0159] Among them, ni,j Represents the i-th feature word t i In the jth corpus d j The number of times it appears in TF i,j Represents the i-th feature word t i In the jth corpus d j The frequency of occurrence in ; k represents the total number of feature words in the jth corpus; n k,j Represents the kth feature word t k In the jth corpus d j The number of times it appears in .

[0160]

[0161] Among them, IDF i represents the inverse document frequency of the i-th feature word; |D| represents the number of all corpora, |j:t i ∈d j | indicates that it contains the feature word t i The number of corpora.

[0162] Step S3: Based on the "manual identification samples," the "Classification of Marine and Related Industries," and the "Technical Regulations for Identifying Marine-Related Enterprises," we learn the feature combinations and association rules of the three characteristic fields of enterprise name, business scope, and national economic industry classification for each marine industry enterprise, and establish a "marine-related enterprise identification model." This step specifically includes:

[0163] Step S3.1: Combining the “standard marine industry characteristic words” and the “sample marine industry characteristic words” into “marine industry characteristic words”;

[0164] Step S3.2: Consider each characteristic word in the "Marine Industry Characteristic Words" as a variable. Convert the business scope text of each enterprise in the "Manual Identification Sample" into a vector containing only 0s and 1s, based on whether the characteristic word exists. A 0 indicates that the business scope text does not contain the characteristic word, and a 1 indicates that the characteristic word exists. After performing this conversion on all samples, a matrix containing only 0s and 1s is obtained, called a one-hot matrix.

[0165] Step S3.3: Use the "Random Forest Model" to train the marine industry classification results of the "manual identification sample" as the dependent variable and each feature word in the constructed one hot matrix as the independent variable to obtain a classification model that can determine the marine industry by feature words;

[0166] "Random Forest" is an integrated learning model composed of multiple decision trees. Each decision tree is trained by randomly selecting samples and randomly selecting feature subsets. The results are finally summarized by voting (classification) or averaging (regression) to improve prediction accuracy and generalization ability.

[0167] Step S3.4: Derive the weights of the independent variables in the "Random Forest Model," which are called the "feature weights" of each feature word. Based on the size of the "feature weights" of the feature words, multiple "identification rules" are developed for each ocean and related industry to form the "Ocean-related Enterprise Identification Model."

[0168] The strength of the correlation between different "feature fields" and whether a company belongs to the marine industry, as well as the authority of the judgment basis, varies. Furthermore, the "feature words" of different "feature fields" may have semantic conflicts (for example, the "company name" contains "aquaculture" but the "National Economic Industry Classification" belongs to the construction industry). To balance the judgment basis of multiple "feature fields", through differentiated weight allocation and priority rules, the most direct and authoritative "feature fields" are prioritized, and then gradually combined with other "feature fields" for supplementary verification, thereby improving the accuracy and efficiency of recognition results.

[0169] The "identification rules" include the weights of the three "feature fields" of company name, business scope, and national economic industry classification, the multiple "marine industry characteristic words" contained in each "feature field", and the priority and union and intersection logical relationships between different "marine industry characteristic words".

[0170] For example, the "identification rules" for marine fisheries include: 1. The "National Economic Industry Classification" is classified as "mariculture" or "marine fishing." 2. The "Company Name" contains "marine industry characteristic words" such as "mariculture" or "marine fishery" and the "National Economic Industry Classification" is classified as "agriculture, forestry, animal husbandry, or fishery."

[0171] Step S4: Use the "marine-related enterprise identification model" to identify the "enterprises to be identified" and divide them into 28 two-digit code categories of marine and related industries (the marine industries (marine and related industries) mentioned in this invention refer to the two-digit code categories of marine and related industries in the "Classification of Marine and Related Industries" (GB / T20794-2021), a total of 28), and obtain "model identification samples".

[0172] This step specifically involves segmenting the text in the "enterprises to be identified" using the same method as the "manual identification samples," comparing the resulting feature words with the "marine industry feature words," and calculating the "relative Hamming distance." Fuzzy recognition of company names and business categories is performed by setting a threshold for the "relative Hamming distance."

[0173] The Hamming distance is named after Richard Wesley Hamming. In information theory, the Hamming distance between two strings of equal length is the number of characters that differ in corresponding positions in the two strings. In other words, it is the number of characters required to replace one string to transform it into the other.

[0174] In actual calculation, the present invention introduces the concept of "relative Hamming distance", and the formula is as follows:

[0175]

[0176] When the Hamming distance between the "enterprise to be identified" feature word and the "marine industry feature word" is small, and the "marine industry feature word" is long, the "relative Hamming distance" is small. Feature words with the smallest "relative Hamming distance" are selected, and a manual determination is made as to whether the two feature words can be considered the same. If so, the "marine industry feature word" is used to replace them when input into the "marine-related enterprise identification model."

[0177] For example, when the "characteristic word of the marine industry" is "aquatic product processing", words such as "aquatic product processing", "aquatic product processing", "processing of aquatic products", and "aquatic product production, sales, and processing" are all replaced with "aquatic product processing", thereby fuzzy identification of the company name and business classification.

[0178] By comparing the "model identification samples" and "manual identification samples", the "identification accuracy" and "additional identification ratio" of each ocean and related industry are calculated.

[0179] Because predictions need to be made on a large number of unlabeled samples, traditional supervised learning metrics (such as precision and recall) cannot be directly used. Therefore, the "recognition accuracy" and "extra recognition ratio" metrics are introduced to evaluate model performance.

[0180] "Identification accuracy" is the ratio of manually identified companies in marine industry A to the number of manually identified companies in marine industry A. This indicator is used to evaluate the ability of the "Marine-related Enterprise Identification Model" to capture known marine-related companies. When the "Identification Accuracy" approaches 100%, it indicates that the model has sufficient coverage of manually labeled positive examples and the risk of missed detection is low. By evaluating this indicator, we can ensure that the model maintains consistency with manual judgment in the prediction of unlabeled data.

[0181]

[0182] The "extra recognition ratio" is the ratio of the number of enterprises identified by the model as belonging to marine industry A to the number of enterprises manually identified as belonging to marine industry A. This indicator is used to evaluate the generalization ability of the "marine-related enterprise identification model" in real-world scenarios. If the value is too high, it indicates that the model's judgment logic is too broad and may mistakenly include non-marine-related enterprises; if the value is too low, potential marine-related enterprises that meet national standards but have not been manually labeled may be missed. By evaluating this indicator, the recognition coverage and accuracy can be dynamically balanced during model iteration, especially when processing large amounts of unlabeled data, to avoid systematic biases caused by overfitting manually labeled samples or rigid rules.

[0183]

[0184] The thresholds for "Identification Accuracy" and "Additional Identification Ratio" vary for each marine and related industry. Generally, the "Identification Accuracy" is maintained above 70% for each marine and related industry. The "Additional Identification Ratio" is set based on the ratio of "Unidentified Enterprises" to "Manually Identified Samples." The initial threshold for each marine and related industry is generally around 10. This threshold will be adjusted based on the results of manual sampling for each marine and related industry.

[0185] Step S5: Fine-tune and refine the "marine-related enterprise identification model", use the adjusted "marine-related enterprise identification model" to identify the enterprises to be identified, and divide them into 28 two-digit code categories of marine and related industries to obtain the "enterprise classification results".

[0186] Specifically, adjusting the marine enterprise identification model in step S5 includes:

[0187] Step S5.1: Establish “exclusion rules” to exclude companies that are clearly not involved in the ocean and whose main business is not in this marine industry, and control the “additional identification ratio”;

[0188] This includes segmenting and extracting keywords from company names in the “model recognition samples”, using the TF-IDF algorithm and manual supplementation to screen out “marine industry exclusion words”.

[0189] For example, the “exclusion rules” for “marine fisheries” include: “Company name” contains “characteristic words” such as “construction” and “advertising”, which indicate that the company’s main business is obviously not related to marine fisheries.

[0190] Step S5.2: Adjust the "identification rules", including adjusting the TF-IDF threshold of the selected feature words, adding or subtracting the "marine industry feature words", and adjusting the "feature weights" to add or subtract the identification rules for each marine industry.

[0191] For example: If the "recognition accuracy" is lower than the threshold or the "additional recognition ratio" is too low, the TF-IDF threshold of the selected feature words will be adjusted to add more "marine industry feature words", or the intersection and union logic in the "recognition rules" will be modified to improve the model's generalization ability and identify more potential marine-related enterprises.

[0192] If the “additional recognition ratio” is too high, the TF-IDF threshold of the candidate feature words will be adjusted to reduce the “marine industry feature words” or add more “marine industry exclusion words”.

[0193] Step S5.3: Perform manual sampling, and add or subtract “marine industry exclusion words” and “marine industry characteristic words” based on the manual sampling results;

[0194] Manual sampling results are analyzed, and based on standard documentation, a manual determination is made as to whether the enterprise's primary business is in the marine industry, thereby determining whether the model's identification is correct. If the model identification misses a marine-related enterprise, "marine industry characteristic words" are added to adjust the "identification rules." If an error occurs, "marine industry exclusion words" are added or "marine industry characteristic words" are modified to adjust the "identification rules."

[0195] Step S5.4: Adjust the “marine-related enterprise identification model” to obtain the “adjusted marine-related enterprise identification model”.

[0196] The "Marine-related Enterprise Identification Model" was fine-tuned and refined by manually screening and supplementing "Marine Industry Exclusion Words" and "Marine Industry Characteristic Words," adjusting the weights of "Characteristic Fields," adjusting the intersection and union logic in the "Identification Rules," and adding "Identification Rules." The adjusted "Identification Accuracy" and "Additional Identification Ratio" were also calculated to evaluate model performance.

[0197] Step S6: Calculate the assessment score for each identified marine industry for the enterprises in the “Enterprise Classification Results”.

[0198] Specifically, step S6 includes:

[0199] Step S6.1: Use the descriptions of each marine industry in the Classification of Marine and Related Industries to form a "Marine Industry Description Database";

[0200] Step S6.2: Calculate the text similarity between the business scope of the enterprise in the “Marine Industry Description Lexicon” and the “Enterprise Classification Results” to obtain the enterprise’s “assessment score” for each identified marine industry.

[0201] If an enterprise is classified into multiple marine and related industries by the "Marine Enterprise Identification Model", the enterprise can be judged to belong to the marine industry with a higher "assessment score" based on the "assessment score" of the enterprise for each identified marine industry.

[0202] The text similarity is calculated using cosine similarity. The specific calculation method is to segment the text of the business scope of the enterprise in the "Marine Industry Description Thesaurus" and "Enterprise Classification Results" to obtain two word lists:

[0203] A=[t1,t2,...t i ]

[0204] B=[t1,t2,...t j ]

[0205] Merge the two word lists to remove duplicates and get all the words in the input sample:

[0206] T(A,B)=T(A)+T(B)=[t1,t2,...t k ]

[0207] Calculate the number of times the kth word appears in A and B as a feature vector:

[0208]

[0209] Finally, substitute the formula to calculate the cosine similarity:

[0210]

[0211] Using the above method, each enterprise can obtain its "cosine similarity" to each marine industry. The "assessment score" for the enterprise's specific identified marine industry is obtained by dividing the cosine similarity value of the enterprise's specific identified marine industry by the sum of the cosine similarity values ​​of all the enterprise's identified marine industries.

[0212] The present invention provides a method and system for identifying marine and related industry classifications. By combining machine learning algorithms with keywords, this method addresses the limited interpretability of machine learning algorithms and can demonstrate the basis for identifying enterprise classifications. By integrating both manually annotated samples and relevant national standards, this method addresses the issues of uneven sampling of manually annotated samples and inconsistency between annotation results and standards.

[0213] Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, and units provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same functions of the system and its various devices, modules, and units provided by the present invention in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; the devices, modules, and units for implementing various functions can also be considered as both software modules implementing the method and structures within the hardware component.

[0214] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A classification and identification method for ocean and related industries, characterized in that: include: Step S1: Establish an enterprise credit information database, divide the enterprises therein into samples for manual identification and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification in the enterprise credit information database; Step S2: obtaining a standard document, and performing text preprocessing on the standard document and the manual recognition sample respectively; Step S3: Based on the manual identification samples and the standard documents, the feature combinations and association rules of the three feature fields of enterprise name, business scope, and national economic industry classification of enterprises in various marine industries are learned to establish a marine enterprise identification model; Step S4: using the marine enterprise identification model to identify the enterprises to be identified, dividing them into n marine industries, obtaining model identification samples, and calculating the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples; Step S5: adjusting the marine-related enterprise identification model according to the recognition accuracy and the additional recognition ratio, using the adjusted marine-related enterprise identification model to identify the enterprises to be identified, and classifying them into n marine industries to obtain enterprise classification results; Step S6: calculating an assessment score for each identified marine industry for the enterprises in the enterprise classification result; The step S4 comprises: Step S4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to perform fuzzy recognition of the enterprise name and business category; When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model; Step S4.2: Compare the model-identified samples with the manually identified samples and calculate the recognition accuracy and additional recognition ratio for each marine and related industry. The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises identified as marine industry A by the manually identified samples. This is used to evaluate the generalization ability of the marine enterprise recognition model in real-world scenarios. The step S5 comprises: Step S5.1: Formulate exclusion rules, including performing keyword segmentation and extraction of enterprise names from the model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding enterprises that are not involved in the sea and whose main business is not in this marine industry, and controlling the excess recognition ratio.

2. The method for classifying and identifying ocean and related industries according to claim 1, characterized in that: The step S2 comprises: Step S2.1: performing word segmentation extraction on the standard document to screen out standard marine industry characteristic words; Step S2.2: Perform word segmentation and extraction on the business scope and company name in the manually identified samples, and use the TF-IDF algorithm to filter out sample marine industry characteristic words from the company keyword library.

3. The method for classifying and identifying ocean and related industries according to claim 1, characterized in that: The step S3 comprises: Step S3.1: combining standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words; Step S3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents that the business scope text of the enterprise does not contain the characteristic word, and 1 represents that the characteristic word exists. After performing the above conversion on all samples, a one-hot matrix is ​​obtained. Step S3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable, thus obtaining a classification model that can identify marine industries based on feature words. Step S3.4: Derive the weight of the independent variables in the random forest model, which is called the feature weight of each feature word. According to the feature weight of the feature word, formulate multiple identification rules for each marine industry to form a marine-related enterprise identification model.

4. The method for classifying and identifying ocean and related industries according to claim 1, characterized in that: The step S4.2 includes: The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows: The thresholds for the recognition accuracy and additional recognition ratio of each ocean and related industry are different to ensure that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the manually identified samples.

5. The method for classifying and identifying ocean and related industries according to claim 3, characterized in that: The adjustment of the marine-related enterprise identification model in step S5 further includes: Step S5.2: Adjusting the identification rules, including adjusting the threshold in the TF-IDF algorithm for filtering feature words, adding or subtracting the marine industry feature words, and adjusting feature weights to add or subtract the identification rules for each marine industry; Step S5.3: Perform manual sampling, and add or subtract the marine industry exclusion words and the marine industry characteristic words according to the manual sampling results; Step S5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain an adjusted marine-related enterprise identification model; At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance.

6. The method for classifying and identifying ocean and related industries according to claim 1, characterized in that: The step S6 comprises: Step S6.1: Using the descriptions of each marine industry in the standard document, a marine industry description vocabulary is formed; Step S6.2: Calculating the assessment score of each identified marine industry by the enterprise in the enterprise classification results based on the text similarity between the marine industry description vocabulary and the business scope of the enterprise in the enterprise classification results; If an enterprise is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise will be judged as belonging to the marine industry with a higher evaluation score based on the enterprise's evaluation score for each identified marine industry.

7. A classification and identification system for ocean and related industries, characterized by: include: Module M1: Establish an enterprise credit information database, divide the enterprises in it into manual identification samples and enterprises to be identified, and select multiple characteristic fields such as enterprise name, business scope, and national economic industry classification in the enterprise credit information database; Module M2: obtaining a standard document, and performing text preprocessing on the standard document and the manual recognition sample respectively; Module M3: Based on manual identification samples and the standard documents, learn the feature combinations and association rules of the three feature fields of enterprise name, business scope, and national economic industry classification for enterprises in various marine industries, and establish an identification model for marine-related enterprises; Module M4: Use the marine enterprise identification model to identify the enterprises to be identified, divide them into n marine industries, obtain model identification samples, and calculate the identification accuracy and additional identification ratio of each marine industry by comparing the model identification samples with the manual identification samples; Module M5: adjusting the marine-related enterprise identification model according to the recognition accuracy and the additional recognition ratio, using the adjusted marine-related enterprise identification model to identify the enterprises to be identified, classifying them into n marine industries, and obtaining enterprise classification results; Module M6: Calculating an assessment score for each identified marine industry for the enterprises in the enterprise classification results; The module M4 includes: Module M4.1: Segment the text of the enterprise to be identified, compare the obtained feature words with the marine industry feature words, calculate the relative Hamming distance, and set the relative Hamming distance threshold to fuzzy identify the enterprise name and business classification; When the Hamming distance between the characteristic word of the enterprise to be identified and the characteristic word of the marine industry is smaller than the length of the characteristic word of the marine industry, the characteristic word with the relative Hamming distance is selected to determine whether the two characteristic words can be considered as the same characteristic word. If so, the characteristic word of the marine industry is used to replace them in the input of the marine enterprise identification model; Module M4.2: Compare the model-identified samples with the manually identified samples to calculate the recognition accuracy and additional recognition ratio for each marine and related industry. The additional recognition ratio is the ratio of the number of enterprises identified as marine industry A by the model to the number of enterprises manually identified as marine industry A. This is used to evaluate the generalization ability of the marine enterprise recognition model in real-world scenarios. The module M5 includes: Module M5.1: Formulate exclusion rules, including performing keyword segmentation and extraction of enterprise names from model recognition samples, using the TF-IDF algorithm and manual supplementation to screen out marine industry exclusion words, excluding enterprises that are not involved in the sea and whose main business is not in this marine industry, and controlling the excess recognition ratio.

8. The marine and related industry classification and identification system according to claim 7 is characterized in that: The module M2 includes: Module M2.1: Perform word segmentation extraction on the standard document to screen out standard marine industry characteristic words; Module M2.2: Segment and extract the business scope and company name in the manually identified samples, and use the TF-IDF algorithm to filter out the sample marine industry characteristic words from the company keyword library; The module M3 includes: Module M3.1: Combine standard marine industry characteristic words and sample marine industry characteristic words to form marine industry characteristic words; Module M3.2: Take each characteristic word in the marine industry characteristic words as a variable and convert the business scope text of each enterprise in the manual identification sample into a vector containing only 0 and 1 based on whether the characteristic word exists. 0 represents the absence of the characteristic word in the business scope text of the enterprise, and 1 represents the presence of the characteristic word. After performing this conversion on all samples, a one-hot matrix is ​​obtained. Module M3.3: Use the random forest model to train the marine industry classification results of the manually identified samples using each feature word in the constructed one-hot matrix as an independent variable to obtain a classification model that can identify marine industries based on feature words. Module M3.4: Derive the weights of the independent variables in the random forest model, which are called the feature weights of each feature word. Based on the feature weights of the feature words, formulate multiple identification rules for each marine industry to form a marine-related enterprise identification model.

9. The marine and related industry classification and identification system according to claim 8, characterized in that: The module M4.2 includes: The recognition accuracy rate is the ratio of the number of enterprises manually identified as marine industry A to the number of enterprises manually identified as marine industry A. It is used to evaluate the ability of the marine enterprise recognition model to capture known marine enterprises. The calculation method is as follows: The thresholds for the recognition accuracy and additional recognition ratio for each ocean and related industry are different, ensuring that the recognition accuracy of each ocean and related industry is higher than 70%. The additional recognition ratio is set according to the ratio of the enterprises to be identified and the samples for manual identification. The module M5 further includes adjusting the marine enterprise identification model: Module M5.2: Adjusting the identification rules, including adjusting the threshold in the TF-IDF algorithm for filtering feature words, adding or removing feature words for the marine industry, and adjusting feature weights to add or remove identification rules for each marine industry; Module M5.3: Perform manual sampling and, based on the results of the manual sampling, add or remove the marine industry exclusion words and the marine industry characteristic words; Module M5.4: Adjust the marine-related enterprise identification model by screening and supplementing marine industry exclusion words and marine industry characteristic words, adjusting the weights of characteristic fields, adjusting the intersection and union logic in the identification rules, and adding identification rules to obtain the adjusted marine-related enterprise identification model; At the same time, the recognition accuracy and additional recognition ratio of the adjusted marine enterprise identification model are calculated to evaluate the model performance; The module M6 includes: Module M6.1: Use the descriptions of each marine industry in the standard documents to develop a marine industry description vocabulary; Module M6.2: Calculate the assessment score of each identified marine industry by the enterprise in the enterprise classification results based on the textual similarity between the marine industry description vocabulary and the business scope of the enterprise in the enterprise classification results; If an enterprise is classified into multiple marine and related industries by the marine enterprise identification model, the enterprise will be judged as belonging to the marine industry with a higher evaluation score based on the enterprise's evaluation score for each identified marine industry.

Images