The company's main commodity forecasting method, device, computer equipment and storage medium

Abstract

The present invention discloses a method, device, computer equipment and storage medium for predicting the main products of an enterprise. The method comprises: extracting and standardizing product information through multiple data sources to form a total product pool; obtaining the business data of the enterprise to be predicted; constructing a corporate profile for the business data of the enterprise to be predicted; and predicting the main products of the enterprise based on the corporate profile, the total product pool and the set knowledge graph. By implementing the method of the present invention, the accuracy and practicality of the prediction can be improved, and the technical problems faced by the existing methods for predicting the main products of unauthorized enterprises, such as limited data sources, processing complexity and insufficient model generalization ability, can be solved.

Description

Technical Field

[0001] The present invention relates to computers, and more specifically to a method, device, computer equipment and storage medium for predicting the main products of an enterprise. Background Art

[0002] Predicting the main products of unauthorized enterprises has certain benefits, such as understanding the main business and product lines of competitors, helping enterprises identify competitive advantages and disadvantages, and formulating corresponding market strategies. Current technologies mainly rely on artificial intelligence methods such as big data analysis, machine learning, and natural language processing. These technologies help identify the main products of enterprises by analyzing the public data of enterprises such as transaction records, product catalogs, market sales, social media discussions, and news reports. However, these methods also face some challenges.

[0003] The analysis method based on big data uses public information on the Internet to collect data from sources such as e-commerce platforms, corporate websites, industry reports and social media to explore the main business of enterprises. Although this method can extract data such as product information, customer reviews and sales, its coverage is limited, especially the inability to obtain information about companies that do not disclose their operating data. In addition, Internet data is usually messy and contains a lot of noise and invalid information, and processing this data requires a lot of manual cleaning. Due to competition reasons, some companies are reluctant to disclose specific operating information, making accurate predictions based on public data difficult.

[0004] Machine learning models can identify a company's main products by training on historical data. These models take into account a company's procurement, sales, and marketing activities to predict its future business focus. However, the models rely on a large amount of high-quality training data. If the data is insufficient or incomplete, the prediction effect will be affected. In addition, machine learning models often make predictions based on historical data, which easily ignores the impact of market environment, technological changes, or unexpected events. Although complex deep learning models can provide highly accurate predictions, their internal mechanisms are relatively complex, and it is difficult to explain the specific reasons for the model's predictions.

[0005] Natural language processing technology can analyze press releases, product descriptions, and market research reports issued by companies to identify their main products. For example, through information such as corporate official websites and industry news, NLP technology can automatically extract and classify product data. However, for unauthorized companies, a lot of data cannot be obtained, and it is difficult to improve generalization capabilities by relying solely on NLP technology. In addition, the ambiguity of the language itself and contextual factors may lead to misjudgment. For example, when a company mentions multiple products, these products may not all be its core businesses.

[0006] Although existing technologies have made some progress in predicting the main products of unauthorized enterprises, they still face challenges such as limited data sources, privacy protection, data processing complexity, and model generalization capabilities. Future development may need to overcome these shortcomings through more efficient data fusion, privacy-enhancing methods (such as federated learning), more powerful NLP models, and more transparent AI models.

[0007] Therefore, it is necessary to design a new method to improve the accuracy and practicality of predictions and to solve the technical problems faced by existing non-authorized enterprise main commodity prediction methods, such as limited data sources, processing complexity, and insufficient model generalization capabilities. Summary of the invention

[0008] The purpose of the present invention is to overcome the defects of the prior art and provide a method, device, computer equipment and storage medium for predicting the main products of an enterprise.

[0009] To achieve the above purpose, the present invention adopts the following technical solution: a method for predicting the main products of an enterprise, comprising:

[0010] Extract and standardize product information from multiple data sources to form a total product pool;

[0011] Obtain the business data of the enterprise to be predicted;

[0012] Constructing an enterprise profile for the enterprise operation data to be predicted;

[0013] The main products of the enterprise are predicted based on the enterprise portrait, the total product pool, and the set knowledge graph.

[0014] A further technical solution is: extracting and standardizing commodity information through multiple data sources to form a total commodity pool, including:

[0015] Obtaining a variety of data, including standardized descriptions of business scope, patent summaries, names of bidding projects, main commodities on the e-commerce platform, and a list of main commodities;

[0016] Performing data cleaning on the multiple data to obtain a cleaning result;

[0017] Product information is extracted from the cleaning results to form a total product pool.

[0018] A further technical solution is: extracting commodity information from the cleaning results and forming a total commodity pool includes:

[0019] Extract product information from the standardized description of the business scope in the cleaning results, and use the named entity recognition model in natural language processing technology to identify product names to obtain a standard product pool;

[0020] Through text preprocessing and NER model, product information is identified from the patent abstracts in the cleaning results, and outlier detection is performed to obtain detection results;

[0021] Using a large pre-trained language model, the corresponding product names are generated from the irregular bidding project names in the cleaning results to obtain the generated results;

[0022] Use rule design, keyword matching and regular expressions to extract the main product information from the main product names of the e-commerce platform in the cleaning results to obtain the extraction results;

[0023] Combining the detection results, generation results, and extraction results to obtain a single-enterprise commodity pool;

[0024] Standardize the business scope of enterprises through multi-mode matching and similarity matching algorithms, and build a commodity mapping table to obtain an industry statistical commodity pool;

[0025] The standard commodity pool, the single-enterprise commodity pool and the industry statistical commodity pool are combined to obtain a total commodity pool.

[0026] A further technical solution is: the step of constructing an enterprise portrait based on the business data of the enterprise to be predicted includes:

[0027] Standardizing the business scope in the business data of the enterprise to be predicted to obtain a standardized result;

[0028] Mapping the industry classification data in the business data of the enterprise to be predicted to obtain a mapping result;

[0029] The standardized results and the mapping results are integrated and matched with the commodities in the total commodity pool to obtain a corporate portrait.

[0030] Its further technical solution is: predicting the main products of the enterprise according to the enterprise portrait, the total product pool, and the set knowledge graph includes:

[0031] Using cosine similarity and a fast matching algorithm to determine a standard business scope close to the business scope in the enterprise portrait, so as to obtain a set of candidate standard business scopes;

[0032] Selecting commodities involved in the candidate standard business scope set from the total commodity pool and standardizing them to obtain a candidate commodity set;

[0033] Performing intersection analysis on the candidate product set, using a set knowledge graph to perform relationship identification to obtain an identification result, and treating products from different sources as a set according to the identification result;

[0034] The collection is displayed according to the set priority to determine the main products of the enterprise.

[0035] A further technical solution is: the cosine similarity and fast matching algorithm are used to determine the standard business scope close to the business scope in the enterprise portrait, so as to obtain a set of candidate standard business scopes, including:

[0036] Combined with the Aho-Corasick algorithm, the standard business scope corresponding to the business scope in the enterprise portrait is quickly identified to obtain a quick identification result;

[0037] Determining whether the quick identification result is to find a standard business scope;

[0038] If the quick identification result finds a standard business scope, the standard business scope corresponding to the identification result is used as a candidate standard business scope set;

[0039] If the rapid identification result does not find the standard business scope, the business scope in the enterprise portrait and the standard business scope in the business scope specification statement are vectorized, and the cosine similarity of the two vectors is calculated, and the candidate standard business scope set is determined based on the cosine similarity.

[0040] Its further technical solution is: the set priority includes sequentially displaying the commodities in the identification results belonging to the single-enterprise commodity pool, the commodities in the identification results belonging to the industry statistical commodity pool, and the commodities in the identification results belonging to the standard commodity pool.

[0041] The present invention also provides a device for predicting the main products of an enterprise, comprising:

[0042] A total commodity pool building unit, used for extracting and standardizing commodity information through multiple data sources to form a total commodity pool;

[0043] An acquisition unit, used for acquiring the business data of the enterprise to be predicted;

[0044] A portrait construction unit, used to construct an enterprise portrait for the enterprise operation data to be predicted;

[0045] The prediction unit is used to predict the main products of the enterprise based on the enterprise portrait, the total product pool, and the set knowledge graph.

[0046] The present invention further provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor implements the above method when executing the computer program.

[0047] The present invention also provides a storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0048] Compared with the prior art, the present invention has the following beneficial effects: the present invention integrates multiple data sources, extracts and standardizes commodity information, and thus forms a comprehensive total commodity pool. The commodity information in this pool provides rich background information for the business data of the enterprise to be predicted. Then, a detailed corporate portrait is constructed for the business data of the enterprise to be predicted to better reflect its business characteristics and market positioning. By combining the corporate portrait with the total commodity pool, an in-depth analysis is performed based on the set knowledge graph to improve the accuracy of the prediction, effectively addressing the problems of limited data sources and processing complexity in the prediction of the main commodities of unauthorized enterprises.

[0049] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0051] Figure 1 A schematic diagram of an application scenario of the method for predicting the main products of an enterprise provided by an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of a process flow of a method for predicting a main commodity of an enterprise provided by an embodiment of the present invention;

[0053] Figure 3 A schematic block diagram of a device for predicting main products of an enterprise provided by an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of the process of building a total commodity pool provided by an embodiment of the present invention;

[0055] Figure 5 A schematic diagram of a process for building an enterprise portrait provided by an embodiment of the present invention;

[0056] Figure 6 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0059] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0060] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0061] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of an application scenario of the method for predicting the main products of an enterprise provided in an embodiment of the present invention. Figure 2 A schematic flow chart of a method for predicting the main products of an enterprise provided by an embodiment of the present invention. The method for predicting the main products of an enterprise is applied to a server. The server interacts with the terminal to form a rich total product pool by obtaining multiple data sources such as standardized descriptions of business scope, patent abstracts, and e-commerce platform products, thereby ensuring the extensiveness of data sources; the collected data is cleaned and standardized to eliminate noise and improve data quality, laying the foundation for subsequent analysis; the business data of the predicted enterprise is standardized and mapped to industry classifications, and the information is comprehensively matched with the total product pool to form an accurate enterprise portrait; named entity recognition (NER) and a large pre-trained language model are used to extract and generate product information from multiple data to ensure the accuracy of product identification; cosine similarity and fast matching algorithms are used to identify the similarity between the enterprise portrait and the standard business scope, thereby improving the accuracy of candidate products; intersection analysis is performed on the candidate product set, and relationship identification is performed in combination with the knowledge graph to identify more relevant main products; products from different sources are displayed in sequence according to the set priority to ensure that the enterprise can preferentially identify the most likely main products. It can effectively solve problems such as limited data sources, high processing complexity and insufficient model generalization capabilities, and improve the overall effect of the company's main product forecasting.

[0062] Figure 2 FIG. 1 is a flow chart of a method for predicting the main products of an enterprise provided by an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110 to S150.

[0063] S110. Extract and standardize commodity information through multiple data sources to form a total commodity pool.

[0064] In this embodiment, the total commodity pool is a comprehensive commodity information library that brings together standardized commodity information from multiple data sources. The commodity information in the pool is cleaned and processed to provide enterprises with accurate commodity identification and classification, which helps to optimize business decisions such as supply chain management, market analysis and sales forecasting.

[0065] The entire construction process is as follows Figure 4 In one embodiment, the above step S110 may include steps S111 to S113.

[0066] S111. Obtain a variety of data, wherein the data includes a standardized description of the business scope, patent abstracts, bidding project names, main commodities on the e-commerce platform, and a main commodity table.

[0067] In this embodiment, the business scope specification statement is a document submitted by an enterprise to the relevant department when registering, which states the business scope and activities that the enterprise is allowed to engage in. These statements usually include a description of the industry, product or service.

[0068] A patent abstract is a brief description of the content of a patent application, usually including the background of the invention, the technical solution, and the problem it solves. It can quickly provide the core information of the invention.

[0069] The bidding project name is the project name used by the enterprise when participating in the bidding process, which usually describes the work to be completed or the products / services to be provided.

[0070] The main commodities on e-commerce platforms refer to the list of commodities that are most popular or have high sales volume on major e-commerce platforms, which reflects consumers' purchasing preferences and market trends.

[0071] The main product table is an internal document of the company that lists the main products or services sold by the company, usually compiled based on sales data and market analysis.

[0072] S112: performing data cleaning on the multiple data to obtain a cleaning result.

[0073] In this embodiment, the collected data are processed and cleaned to ensure the data quality used in the subsequent steps. This process is intended to eliminate redundant information, errors and inconsistencies in the data, thereby improving the accuracy and reliability of the data. Through effective data cleaning, the validity of the subsequent analysis results can be ensured, providing a solid foundation for subsequent decision-making.

[0074] S113, extracting commodity information from the cleaning results and forming a total commodity pool.

[0075] In one embodiment, the above-mentioned step S113 may include steps S1131 - S1137 .

[0076] S1131. Extract product information from the standardized description of the business scope in the cleaning results, and use the named entity recognition model in natural language processing technology to identify the product name to obtain a standard product pool.

[0077] In this embodiment, the standard commodity pool is a database containing various standardized commodities, which is usually constructed according to industry norms, laws and regulations, and market standards. The commodities in the pool have clear definitions and classifications, which facilitates standardized management and reference in commercial activities.

[0078] Specifically, the business scope standard expression directory needs to be loaded as structured data to ensure that each field is clearly segmented.

[0079] Natural language processing technology, especially named entity recognition (NER) model, is used to identify key entities such as product names from the business scope description.

[0080] Input text T, and output a series of entities E through the entity recognition model NER(T), each entity corresponds to a different category. For example:

[0081] Text (T): Refers to the planting of crops mainly for the harvest of seeds. This includes the planting of crops such as rice, wheat, corn, and the planting of grains used as feed and industrial raw materials.

[0082] Entity (E): ["rice", "wheat", "corn"];

[0083] Perform post-processing on the generated results, such as deduplication and formatting.

[0084] S1132. Identify product information from the patent abstract in the cleaning result through text preprocessing and NER model, and perform outlier detection to obtain the detection result.

[0085] In this embodiment, the patent abstract text is cleaned to remove irrelevant words and formats.

[0086] Entity recognition: Use the NER model to identify key entities, such as product names, and concatenate modified products with products, as shown in Table 1.

[0087] Table 1. Key entities identified using NER models

[0088]

[0089] Outlier detection: Perform outlier detection on the identified commodity entities and remove commodities with large differences, such as removing "broiler chickens".

[0090] S1133. Using a large pre-trained language model, generate corresponding product names from irregular bidding project names in the cleaning results to obtain generated results.

[0091] Specifically, since the project names in the bidding data are irregular, it is impossible to extract products using unified rules or entity recognition. Therefore, a large pre-trained language model is used to infer and generate product names.

[0092] Input text B and generate product name S through language model LM(B) reasoning, as shown in Table 2.

[0093] Table 2. Product names generated by language model LM(B) from input text

[0094]

[0095] S1134. Use rule design, keyword matching and regular expressions to extract main product information from the main product names of the e-commerce platform in the cleaning results to obtain extraction results.

[0096] In this embodiment, based on the structural design rules of commodity names, such as 1688 commodity names usually contain information such as brand, model, category, etc., a keyword library is constructed to extract the main commodities through keyword matching. Regular expressions are used to identify specific commodity formats.

[0097] S1135. Combine the detection results, generation results and extraction results to obtain a single-enterprise commodity pool.

[0098] In this embodiment, the single-enterprise commodity pool is a commodity collection for a specific enterprise, including all commodity information operated by the enterprise. This pool can help enterprises with internal management, inventory control, and sales analysis, and support enterprises in making strategic decisions in the market.

[0099] S1136. Standardize the business scope of enterprises through multi-mode matching and similarity matching algorithms, and build a commodity mapping table to obtain an industry statistical commodity pool.

[0100] In this embodiment, the industry statistical commodity pool aggregates commodity information in a specific industry, mainly used for industry analysis and statistics. By sorting out the classification, quantity and sales of commodities in the industry, it can help enterprises and researchers understand industry trends, market demand and competition.

[0101] Specifically, the business scope of the main commodity data is standardized (Aho-Corasick, similarity matching), and the commodity mapping table is constructed using the standardized business scope and industry.

[0102] First, a multi-pattern matching algorithm is used to identify the standardized business scope of an enterprise. For the unmatched parts, a similarity calculation is performed to ensure that the unidentified parts can be processed through approximate matching.

[0103] Text preprocessing: The business scope of an enterprise is usually described in natural language and may include a variety of goods, services and industry types. To this end, the standard business scope set needs to be preprocessed first to form a list of keywords or phrases, and then the Aho-Corasick algorithm is used for fast scanning and matching.

[0104] Keyword set construction: Establish a standard library containing standard product and service descriptions as the basis for matching.

[0105] Trie tree construction: The standardized business scope description is constructed as a Trie tree, where each node represents a character, and character-by-character matching is performed to find multiple keywords.

[0106] Failure pointer construction: A fail link is set in the Trie tree so that when a character mismatch occurs during the matching process, the fail link can be used to transfer to the next matching state, thus avoiding re-matching from the beginning.

[0107] Matching process: After preprocessing, the input business scope text is input into the Aho-Corasick algorithm character by character for matching. If a standard business scope is matched, it is marked as a standardized business scope. The time complexity of this process is O(n), where n is the length of the input text, and it can efficiently process large data sets.

[0108] Specifically, if the current character c exists in the child node of the current state p, then transfer to the child node: goto(p,c)=q;

[0109] Otherwise, transfer goto(p,c)=fail(p) according to the failure pointer;

[0110] If the child node q of the current state p cannot match the character c, the fail pointer will point to another node r in the Trie tree so that it can continue to match the same character sequence fail(q)=r.

[0111] Similarity matching: For business scopes that cannot be identified by the multi-mode matching algorithm, the similarity matching algorithm is used to process them. By calculating the similarity between the input text and the standardized business scope library, the closest standardized description is found and matched.

[0112] Text similarity calculation: Calculate the similarity between the unmatched business scope and the standardized business scope set, and use cosine similarity to measure the directional similarity of the two texts.

[0113] Text similarity calculation: Calculate the similarity between the unmatched business scope and the standardized business scope set. Among them, A and B are the word vectors of the input text and the standardized business scope, and the cosine similarity is used to measure the directional similarity of the two texts.

[0114] Mapping construction: Finally, a mapping relationship between commodities and their corresponding products is constructed based on the standard business scope and industry, i.e. (standard business scope, industry)-commodity.

[0115] S1137. Combine the standard commodity pool, the single-enterprise commodity pool, and the industry statistical commodity pool to obtain a total commodity pool.

[0116] In this embodiment, the benefits of combining the standard commodity pool, the single-enterprise commodity pool, and the industry statistical commodity pool into the total commodity pool include:

[0117] Combining product pools from different sources can cover a wider range of product categories and ensure that no important product information is missed.

[0118] By combining standard commodities with specific commodities of individual enterprises, the accuracy of commodity classification and matching can be improved, and redundancy and inconsistency can be reduced.

[0119] The Industry Statistics commodity pool provides trends and demands within the industry, which, combined with the other two, can better adapt to market changes and customer needs.

[0120] The data in different pools complement each other and can form richer product information when combined, which is convenient for analysis and decision-making.

[0121] The total product pool can be used to optimize the product recommendation system, improve user experience, and increase sales opportunities.

[0122] By integrating resources, companies can better identify market opportunities and gain an advantage over the competition.

[0123] An integrated commodity pool can simplify the data analysis process, improve decision-making efficiency, and support more complex data mining and analysis.

[0124] In summary, combining these commodity pools can provide companies with more powerful commodity management and market responsiveness capabilities.

[0125] S120. Obtaining the business data of the enterprise to be predicted.

[0126] In this embodiment, the business data of the enterprise to be predicted includes data of multiple dimensions such as the business scope and industry category of the enterprise.

[0127] S130, constructing an enterprise portrait for the enterprise business data to be predicted.

[0128] like Figure 5 As shown in Figure 1, the corporate portrait integrates information from multiple dimensions, such as business scope and industry category. Through natural language processing (NLP) and similarity matching technology, information can be extracted from relevant fields of the company (such as business scope and industry category) and mapped to a standardized commodity pool. This process ensures an effective match between the corporate description and the commodities in the total commodity pool.

[0129] Business scope: The business scope of an enterprise is processed and extracted through standardized expressions to more clearly reflect the business activities of the enterprise.

[0130] Industry classification: The industry to which a company belongs corresponds to the standardized entries of its business scope, helping to more accurately locate the company's market environment.

[0131] After integrating this information, a complete portrait of the enterprise can be constructed, providing a solid foundation for subsequent product forecasts.

[0132] In one embodiment, the above-mentioned step S130 may include steps S131 - S133 .

[0133] S131. Standardize the business scope in the business data of the enterprise to be predicted to obtain a standardized result.

[0134] In this embodiment, the business scope of the enterprise is standardized to ensure that its description conforms to a unified format, thereby obtaining a clear standardized result. The benefit of this process is that standardization can reduce information inconsistency and improve the accuracy of subsequent analysis and matching.

[0135] S132. Map the industry classification data in the business data of the enterprise to be predicted to obtain a mapping result.

[0136] In this embodiment, the industry classification data of the enterprise is mapped and matched with the predefined industry categories to obtain a mapping result. The advantage of this step is that accurate industry classification helps to better understand the market positioning and competitive environment of the enterprise.

[0137] S133. Integrate the information of the standardization result and the mapping result, and match them with the commodities in the total commodity pool to obtain a corporate portrait.

[0138] In this embodiment, the standardized results are integrated with the mapping results, and then matched with the products in the total product pool to build a complete profile of the enterprise. The benefit of this step is that through effective matching, more accurate product recommendations and market forecasts can be achieved, improving the decision support capabilities of the enterprise.

[0139] Through these three steps, the characteristics of the enterprise to be predicted can be fully and accurately portrayed, laying a solid foundation for subsequent commodity forecasting.

[0140] S140, predicting the main products of the enterprise based on the enterprise portrait, the total product pool, and the set knowledge graph.

[0141] In one embodiment, the above step S140 may include steps S141-

[0142] S141, using cosine similarity and a fast matching algorithm to determine a standard business scope that is close to the business scope in the enterprise portrait, so as to obtain a set of candidate standard business scopes;

[0143] Specifically, in order to predict the products of an enterprise, a similarity calculation algorithm, such as cosine similarity, can be used. The specific process is as follows:

[0144] Combined with the Aho-Corasick algorithm, the business scope of the enterprise is efficiently scanned and the standardized business scope is marked. If no matching standard entry is found, the system will use similarity calculation to find the closest standard business scope for matching.

[0145] First, the business scope of the enterprise is vectorized with the list of standard business scopes. Then, the closest match between the business scope of the enterprise and the standard business scope is found by calculating the cosine similarity. A represents the vector of the business scope of the enterprise, while B is the vector of a standard business scope in the standard expression catalog. Cosine similarity is used to measure the similarity between the two vectors to determine the accuracy of the match.

[0146] Through the above steps, the business scope of the enterprise can be compared with the standard efficiently and accurately, thus providing support for commodity forecasting.

[0147] In one embodiment, the above-mentioned step S141 may include steps S1411 - S1414 .

[0148] S1411. Use the Aho-Corasick algorithm to quickly identify the standard business scope corresponding to the business scope in the enterprise portrait to obtain a quick identification result.

[0149] In this embodiment, the Aho-Corasick algorithm is used to quickly scan the business scope in the corporate portrait to find the standardized business scope, thereby improving the recognition speed, being able to quickly process a large amount of data, and ensuring timely feedback.

[0150] S1412: Determine whether the rapid identification result is to find a standard business scope.

[0151] In this embodiment, it is checked whether the rapid identification result has found the standard business scope; by making a clear judgment, the possibility of wrong matching is reduced and the accuracy is improved.

[0152] S1413: If the quick identification result finds a standard business scope, the standard business scope corresponding to the identification result is used as a candidate standard business scope set.

[0153] In this embodiment, if a standard business scope is found, it is added to the candidate set; ensuring that the candidate set contains a real and valid business scope is helpful for subsequent product screening.

[0154] S1414. If the rapid identification result does not find the standard business scope, the business scope in the enterprise portrait and the standard business scope in the business scope specification statement are vectorized, and the cosine similarity of the two vectors is calculated, and a set of candidate standard business scopes is determined based on the cosine similarity.

[0155] In this embodiment, if the standard business scope is not found, the closest standard business scope is determined using vectorization and cosine similarity algorithm, which allows the relevant business scope to be found through similarity analysis in the absence of direct matching, thereby improving the flexibility and adaptability of the system.

[0156] S142: Filter out commodities involved in the candidate standard business scope set from the total commodity pool, and standardize them to obtain a candidate commodity set.

[0157] The products matched from the product pool are standardized, mainly by removing irrelevant content through regular expressions. Table 3 shows an example of standardization.

[0158] Table 3. Standardization examples

[0159]

[0160] It is possible to infer the products that meet two or more business scopes at the same time. For example, if a company's business scope is "fastener manufacturing" and "auto parts and accessories manufacturing", it is possible to find the intersection of the products corresponding to these two business scopes;

[0161] Business scope: "Fastener manufacturing" and "Auto parts and accessories manufacturing";

[0162] Product intersection: engine cylinder head bolts, wheel hub bolts, chassis suspension bolts, hexagonal flange nuts, self-locking nuts, self-tapping screws, cross recessed pan head screws, flat washers, spring washers, aluminum rivets, steel rivets.

[0163] Through standardization, redundant information in product descriptions can be eliminated, ensuring data consistency and clarity, which facilitates subsequent analysis and decision-making.

[0164] S143. Perform intersection analysis on the candidate product set, use the set knowledge graph to perform relationship identification to obtain an identification result, and according to the identification result, treat the products from different sources as a set.

[0165] In this embodiment, commodities from different sources are regarded as a set, and the internal knowledge graph is used to identify commodity relationships. Through the knowledge graph, the relationship between commodities can be better understood, potential commodity associations can be identified, and more accurate commodity forecasts can be provided to enterprises.

[0166] For example, {forest spices; rock laurel oil; cypress oil; mushroom seedlings} identifies "rock laurel oil" and "cypress oil" as belonging to "forest spices", and is finally displayed as {forest spices; mushroom seedlings}, which means displaying the product at the highest level in the product relationship.

[0167] Intersection analysis helps to find products that meet multiple business scopes, while knowledge graphs enhance the understanding of the relationships between products and can identify potential product associations, thereby providing companies with more accurate product recommendations.

[0168] S144. Display the collection according to the set priority to determine the main products of the enterprise.

[0169] In this embodiment, the set priority includes sequentially displaying commodities belonging to a single-enterprise commodity pool in the recognition results, commodities belonging to an industry statistical commodity pool in the recognition results, and commodities belonging to a standard commodity pool in the recognition results.

[0170] Through priority display, companies can ensure that the most relevant products are considered first when making decisions, thereby improving decision-making efficiency and effectiveness.

[0171] The above steps can efficiently and accurately extract the information of main products from the enterprise portrait and product pool. This method combines the advantages of fast matching, similarity calculation and knowledge graph, which not only improves the accuracy of prediction, but also can flexibly respond to diverse business scenarios, helping enterprises seize opportunities in competition.

[0172] The method of this embodiment significantly improves the automation level of data processing, especially in the process of business scope, industry classification and commodity matching, and widely uses natural language processing (NLP), named entity recognition (NER) and multi-mode matching algorithms (such as Aho-Corasick). Compared with traditional manual processing methods, this solution greatly reduces manual intervention, improves work efficiency and continuity of data processing, and can quickly provide high-quality commodity prediction results to adapt to the rapidly changing business environment.

[0173] In addition, the method of this embodiment combines multiple algorithms (such as cosine similarity and large pre-trained language models) to effectively extract product information that is not explicitly stated by the enterprise and accurately predict potential main products. By matching the similarity between the product pool and the business scope of the enterprise, the relevance of the prediction results is ensured and the interference of irrelevant products is reduced. At the same time, the standardized extraction method of rule design and regular expressions is adopted to further enhance the accuracy of the prediction results.

[0174] The applicability and flexibility of the method in this embodiment are also very significant, and can meet the needs of enterprises of different sizes and multiple industries. Whether it is from standardized business scope data or complex bidding project names and patent abstracts, this solution can effectively extract and predict through NER models and pre-trained language models. This multi-source data processing capability makes it widely adaptable in various application scenarios.

[0175] In terms of data processing and response speed, the method of this embodiment can quickly process large-scale text data by combining the multi-pattern matching algorithm with similarity calculation, and the processing speed is significantly higher than the traditional manual method. In addition, the time complexity of the multi-pattern matching algorithm is O(n), which ensures efficient operation in large data sets and significantly improves the scalability and real-time processing capabilities of the system.

[0176] By introducing knowledge graph and commodity level identification technology, the method of this embodiment can intelligently classify commodity data from different sources and integrate similar or related commodities into higher-level categories, thereby improving the accuracy of commodity prediction and providing enterprises with deeper business insights and industry analysis support.

[0177] Through accurate commodity forecasting, the method of this embodiment helps enterprises identify potential business expansion opportunities and provides accurate market analysis. At the same time, for unauthorized enterprises, the system can conduct compliance review by comparing the commodities with those of authorized enterprises in the market, thereby enhancing market competitiveness.

[0178] In addition, the method of this embodiment has significant advantages in terms of data coverage, and can effectively identify and process enterprise products and data information from different sources, overcoming the limitations of the existing technology in collecting enterprise product information. Combining multi-mode matching algorithms and similarity matching algorithms, the enterprise business scope text is quickly scanned and unmatched items are accurately processed, achieving efficient and accurate data processing.

[0179] The method of this embodiment also relies on NER technology to accurately extract product names from complex unstructured texts, and combines it with a large model to significantly improve the accuracy of extraction. By using knowledge graph technology to hierarchically identify and classify products, the system can clearly display product categories, making the prediction results more systematic.

[0180] Finally, the method of this embodiment focuses on the automated processing of data cleaning and standardization to ensure high-quality and consistent data. In addition, it is possible to extract product information from a variety of data sources (such as business scope catalogs, bidding project names, etc.) to ensure the compatibility of different structured and unstructured data. The intelligent algorithm realizes the automated mapping of corporate portraits and commodity pools, accelerates data processing, updates commodities in the commodity pool in real time, and provides accurate business forecasting support for enterprises. Through the vectorization of text data, this technology also lays the foundation for subsequent similarity calculation, classification, and prediction.

[0181] The above-mentioned enterprise main commodity forecasting method integrates multiple data sources, extracts and standardizes commodity information, and thus forms a comprehensive total commodity pool. The commodity information in this pool provides rich background information for the business data of the enterprise to be forecasted. Next, a detailed enterprise portrait is constructed for the business data of the enterprise to be forecasted to better reflect its business characteristics and market positioning. By combining the enterprise portrait with the total commodity pool, an in-depth analysis is performed based on the set knowledge graph to improve the accuracy of the forecast, effectively addressing the problems of limited data sources and processing complexity in the forecast of the main commodities of unauthorized enterprises.

[0182] Figure 3 is a schematic block diagram of a device 300 for predicting main products of an enterprise provided by an embodiment of the present invention. Figure 3 As shown, corresponding to the above enterprise main product forecasting method, the present invention also provides an enterprise main product forecasting device 300. The enterprise main product forecasting device 300 includes a unit for executing the above enterprise main product forecasting method, and the device can be configured in a server. Figure 3 The enterprise main product prediction device 300 includes a total product pool building unit 301, an acquisition unit 302, a portrait building unit 303 and a prediction unit 304.

[0183] The total commodity pool building unit 301 is used to extract and standardize commodity information through multiple data sources to form a total commodity pool; the acquisition unit 302 is used to obtain the business data of the enterprise to be predicted; the portrait construction unit 303 is used to construct an enterprise portrait for the business data of the enterprise to be predicted; the prediction unit 304 is used to predict the main commodities of the enterprise based on the enterprise portrait, the total commodity pool, and the set knowledge graph.

[0184] In one embodiment, the total commodity pool construction unit includes:

[0185] A data acquisition subunit is used to acquire a variety of data, wherein the data includes a standardized description of the business scope, patent abstracts, bidding project names, main commodities on the e-commerce platform, and a main commodity table;

[0186] A cleaning subunit, used for performing data cleaning on the multiple data to obtain a cleaning result;

[0187] A subunit is formed, which is used to extract commodity information from the cleaning result and form a total commodity pool.

[0188] In one embodiment, the forming subunit comprises:

[0189] The first recognition module is used to extract product information from the standard description of the business scope in the cleaning result, and use the named entity recognition model in the natural language processing technology to recognize the product name to obtain a standard product pool;

[0190] The second recognition module is used to identify product information from the patent abstract in the cleaning result through text preprocessing and NER model, and perform outlier detection to obtain the detection result;

[0191] A generation module is used to generate corresponding product names from irregular bidding project names in the cleaning results using a large pre-trained language model to obtain a generation result;

[0192] An extraction module is used to extract main product information from the main product names of the e-commerce platform in the cleaning results by using rule design, keyword matching and regular expressions to obtain extraction results;

[0193] A first combination module, used for combining the detection results, the generation results and the extraction results to obtain a single-enterprise commodity pool;

[0194] A mapping table construction module is used to standardize the business scope of enterprises through multi-mode matching and similarity matching algorithms, and to construct a commodity mapping table to obtain an industry statistical commodity pool;

[0195] The second combination module is used to combine the standard commodity pool, the single-enterprise commodity pool and the industry statistical commodity pool to obtain a total commodity pool.

[0196] In one embodiment, the portrait construction unit 303 includes:

[0197] A first standardization subunit is used to standardize the business scope in the business data of the enterprise to be predicted to obtain a standardized result;

[0198] A mapping subunit, used for mapping the industry classification data in the business data to be predicted to obtain a mapping result;

[0199] The integration subunit is used to integrate the information of the standardization result and the mapping result, and match them with the commodities in the total commodity pool to obtain a corporate portrait.

[0200] In one embodiment, the prediction unit 304 includes:

[0201] A scope determination subunit, for determining a standard business scope close to the business scope in the enterprise portrait by using cosine similarity and a fast matching algorithm, so as to obtain a set of candidate standard business scopes;

[0202] A screening subunit, used to screen out commodities involved in a candidate standard business scope set from the total commodity pool, and perform standardization to obtain a candidate commodity set;

[0203] An analysis subunit, configured to perform intersection analysis on the candidate product set, use a set knowledge graph to perform relationship recognition to obtain a recognition result, and classify products from different sources as a set according to the recognition result;

[0204] The display subunit is used to display the collection according to the set priority to determine the main products of the enterprise.

[0205] In one embodiment, the range determination subunit includes:

[0206] A rapid identification module, used to quickly identify the standard business scope corresponding to the business scope in the enterprise portrait in combination with the Aho-Corasick algorithm to obtain a rapid identification result;

[0207] A judgment module, used to judge whether the rapid identification result is to find a standard business scope;

[0208] A first determination module, configured to, if the rapid identification result finds a standard business scope, use the standard business scope corresponding to the identification result as a candidate standard business scope set;

[0209] A calculation module is used to vectorize the business scope in the enterprise portrait and the standard business scope in the business scope specification statement if the rapid identification result does not find the standard business scope, and calculate the cosine similarity of the two vectors, and determine the candidate standard business scope set according to the cosine similarity.

[0210] It should be noted that technicians in the relevant field can clearly understand that the specific implementation process of the above-mentioned enterprise main product prediction device 300 and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, it will not be repeated here.

[0211] The above-mentioned enterprise main product prediction device 300 can be implemented in the form of a computer program. The computer program can be used in Figure 6 Runs on the computer device shown.

[0212] See also Figure 6 , Figure 65 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a server, wherein the server may be an independent server or a server cluster composed of multiple servers.

[0213] See also Figure 6 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0214] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions, and when the program instructions are executed, the processor 502 can execute a method for predicting the main products of an enterprise.

[0215] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0216] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a method for predicting the main products of an enterprise.

[0217] The network interface 505 is used to communicate with other devices over the network. Figure 6 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0218] The processor 502 is used to run the computer program 5032 stored in the memory to implement the following steps:

[0219] Extract and standardize commodity information through multiple data sources to form a total commodity pool; obtain the business data of the enterprise to be predicted; construct a corporate portrait for the business data to be predicted; predict the main commodities of the enterprise based on the corporate portrait, the total commodity pool, and the set knowledge graph.

[0220] In one embodiment, when the processor 502 implements the step of extracting and standardizing commodity information through multiple data sources to form a total commodity pool, the processor 502 specifically implements the following steps:

[0221] Acquire a variety of data, wherein the data include standardized descriptions of business scope, patent abstracts, names of bidding projects, main commodities on the e-commerce platform, and a main commodity table; perform data cleaning on the various data to obtain cleaning results; extract commodity information from the cleaning results and form a total commodity pool.

[0222] In one embodiment, when the processor 502 implements the step of extracting commodity information from the cleaning result and forming a total commodity pool, the processor 502 specifically implements the following steps:

[0223] Extract commodity information from the standardized description of the business scope in the cleaning results, and use the named entity recognition model in the natural language processing technology to identify the commodity name to obtain a standard commodity pool; identify commodity information from the patent abstract in the cleaning results through text preprocessing and NER model, and perform outlier detection to obtain detection results; use a large pre-trained language model to generate corresponding commodity names from irregular bidding project names in the cleaning results to obtain generation results; use rule design, keyword matching and regular expressions to extract the main commodity information from the main commodity names of the e-commerce platform in the cleaning results to obtain extraction results; combine the detection results, generation results and extraction results to obtain a single-household enterprise commodity pool; standardize the enterprise business scope through multi-mode matching and similarity matching algorithms, and construct a commodity mapping table to obtain an industry statistical commodity pool; combine the standard commodity pool, the single-household enterprise commodity pool and the industry statistical commodity pool to obtain a total commodity pool.

[0224] In one embodiment, when implementing the step of constructing an enterprise profile for the enterprise operation data to be predicted, the processor 502 specifically implements the following steps:

[0225] The business scope in the business data of the enterprise to be predicted is standardized to obtain a standardized result; the industry classification data in the business data of the enterprise to be predicted is mapped to obtain a mapping result; the standardized result and the mapping result are integrated, and matched with the commodities in the total commodity pool to obtain a corporate portrait.

[0226] In one embodiment, when the processor 502 implements the step of predicting the main products of the enterprise according to the enterprise portrait, the total product pool, and the set knowledge graph, it specifically implements the following steps:

[0227] The cosine similarity and fast matching algorithm are used to determine the standard business scope that is close to the business scope in the enterprise portrait to obtain a set of candidate standard business scopes; the commodities involved in the set of candidate standard business scopes are screened out from the total commodity pool and standardized to obtain a set of candidate commodities; an intersection analysis is performed on the set of candidate commodities, and a set knowledge graph is used to perform relationship identification to obtain an identification result, and commodities from different sources are grouped as a set according to the identification result; the set is displayed according to the set priority to determine the main commodities of the enterprise.

[0228] The set priority includes sequentially displaying the commodities in the identification results belonging to the single-enterprise commodity pool, the commodities in the identification results belonging to the industry statistics commodity pool, and the commodities in the identification results belonging to the standard commodity pool.

[0229] In one embodiment, when the processor 502 implements the step of using cosine similarity and a fast matching algorithm to determine a standard business scope close to the business scope in the enterprise portrait to obtain a set of candidate standard business scopes, the processor 502 specifically implements the following steps:

[0230] The standard business scope corresponding to the business scope in the enterprise portrait is quickly identified in combination with the Aho-Corasick algorithm to obtain a rapid identification result; it is determined whether the rapid identification result is a standard business scope; if the rapid identification result finds a standard business scope, the standard business scope corresponding to the identification result is used as a candidate standard business scope set; if the rapid identification result does not find a standard business scope, the business scope in the enterprise portrait and the standard business scope in the business scope specification statement are vectorized, and the cosine similarity of the two vectors is calculated, and the candidate standard business scope set is determined according to the cosine similarity.

[0231] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0232] It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0233] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein when the computer program is executed by a processor, the processor executes the following steps:

[0234] Extract and standardize commodity information through multiple data sources to form a total commodity pool; obtain the business data of the enterprise to be predicted; construct a corporate portrait for the business data to be predicted; predict the main commodities of the enterprise based on the corporate portrait, the total commodity pool, and the set knowledge graph.

[0235] In one embodiment, when the processor executes the computer program to implement the step of extracting and standardizing commodity information through multiple data sources to form a total commodity pool, the processor specifically implements the following steps:

[0236] Acquire a variety of data, wherein the data include standardized descriptions of business scope, patent abstracts, names of bidding projects, main commodities on the e-commerce platform, and a main commodity table; perform data cleaning on the various data to obtain cleaning results; extract commodity information from the cleaning results and form a total commodity pool.

[0237] In one embodiment, when the processor executes the computer program to implement the step of extracting commodity information from the cleaning results and forming a total commodity pool, the processor specifically implements the following steps:

[0238] Extract commodity information from the standardized description of the business scope in the cleaning results, and use the named entity recognition model in the natural language processing technology to identify the commodity name to obtain a standard commodity pool; identify commodity information from the patent abstract in the cleaning results through text preprocessing and NER model, and perform outlier detection to obtain detection results; use a large pre-trained language model to generate corresponding commodity names from irregular bidding project names in the cleaning results to obtain generation results; use rule design, keyword matching and regular expressions to extract the main commodity information from the main commodity names of the e-commerce platform in the cleaning results to obtain extraction results; combine the detection results, generation results and extraction results to obtain a single-household enterprise commodity pool; standardize the enterprise business scope through multi-mode matching and similarity matching algorithms, and construct a commodity mapping table to obtain an industry statistical commodity pool; combine the standard commodity pool, the single-household enterprise commodity pool and the industry statistical commodity pool to obtain a total commodity pool.

[0239] In one embodiment, when the processor executes the computer program to implement the step of constructing an enterprise profile for the business data of the enterprise to be predicted, the processor specifically implements the following steps:

[0240] The business scope in the business data of the enterprise to be predicted is standardized to obtain a standardized result; the industry classification data in the business data of the enterprise to be predicted is mapped to obtain a mapping result; the standardized result and the mapping result are integrated, and matched with the commodities in the total commodity pool to obtain a corporate portrait.

[0241] In one embodiment, when the processor executes the computer program to implement the step of predicting the main products of the enterprise according to the enterprise portrait, the total product pool, and the set knowledge graph, the processor specifically implements the following steps:

[0242] The cosine similarity and fast matching algorithm are used to determine the standard business scope that is close to the business scope in the enterprise portrait to obtain a set of candidate standard business scopes; the commodities involved in the set of candidate standard business scopes are screened out from the total commodity pool and standardized to obtain a set of candidate commodities; an intersection analysis is performed on the set of candidate commodities, and a set knowledge graph is used to perform relationship identification to obtain an identification result, and commodities from different sources are grouped as a set according to the identification result; the set is displayed according to the set priority to determine the main commodities of the enterprise.

[0243] The set priority includes sequentially displaying commodities belonging to a single-enterprise commodity pool in the identification results, commodities belonging to an industry statistical commodity pool in the identification results, and commodities belonging to a standard commodity pool in the identification results.

[0244] In one embodiment, when the processor executes the computer program to implement the step of using cosine similarity and a fast matching algorithm to determine a standard business scope close to the business scope in the corporate portrait to obtain a set of candidate standard business scopes, the processor specifically implements the following steps:

[0245] The standard business scope corresponding to the business scope in the enterprise portrait is quickly identified in combination with the Aho-Corasick algorithm to obtain a rapid identification result; it is determined whether the rapid identification result is a standard business scope; if the rapid identification result finds a standard business scope, the standard business scope corresponding to the identification result is used as a candidate standard business scope set; if the rapid identification result does not find a standard business scope, the business scope in the enterprise portrait and the standard business scope in the business scope specification statement are vectorized, and the cosine similarity of the two vectors is calculated, and the candidate standard business scope set is determined according to the cosine similarity.

[0246] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0247] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0248] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0249] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0250] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0251] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting the main products of an enterprise, characterized in that: include: Extract and standardize product information from multiple data sources to form a total product pool; Obtain the business data of the enterprise to be predicted; Constructing an enterprise profile for the enterprise operation data to be predicted; Predict the main products of the enterprise based on the enterprise portrait, the total product pool, and the set knowledge graph; The predicting of the main products of the enterprise according to the enterprise portrait, the total product pool, and the set knowledge graph includes: Using cosine similarity and a fast matching algorithm to determine a standard business scope close to the business scope in the enterprise portrait, so as to obtain a set of candidate standard business scopes; Selecting commodities involved in the candidate standard business scope set from the total commodity pool and standardizing them to obtain a candidate commodity set; Performing intersection analysis on the candidate product set, using a set knowledge graph to perform relationship identification to obtain an identification result, and treating products from different sources as a set according to the identification result; The collection is displayed according to the set priority to determine the main products of the enterprise.

2. The method for predicting the main products of an enterprise according to claim 1, characterized in that: The product information is extracted and standardized through multiple data sources to form a total product pool, including: Obtaining a variety of data, including standardized descriptions of business scope, patent summaries, names of bidding projects, main commodities on the e-commerce platform, and a list of main commodities; Performing data cleaning on the multiple data to obtain a cleaning result; Product information is extracted from the cleaning results to form a total product pool.

3. The enterprise main product forecasting method according to claim 2, characterized in that: The step of extracting commodity information from the cleaning results and forming a total commodity pool includes: Extract product information from the standardized description of the business scope in the cleaning results, and use the named entity recognition model in natural language processing technology to identify product names to obtain a standard product pool; Through text preprocessing and NER model, product information is identified from the patent abstracts in the cleaning results, and outlier detection is performed to obtain detection results; Using a large pre-trained language model, the corresponding product names are generated from the irregular bidding project names in the cleaning results to obtain the generated results; Use rule design, keyword matching and regular expressions to extract the main product information from the main product names of the e-commerce platform in the cleaning results to obtain the extraction results; Combining the detection results, generation results, and extraction results to obtain a single-enterprise commodity pool; Standardize the business scope of enterprises through multi-mode matching and similarity matching algorithms, and build a commodity mapping table to obtain an industry statistical commodity pool; The standard commodity pool, the single-enterprise commodity pool and the industry statistical commodity pool are combined to obtain a total commodity pool.

4. The method for predicting the main products of an enterprise according to claim 1, characterized in that: The step of constructing an enterprise profile for the enterprise operation data to be predicted includes: Standardizing the business scope in the business data of the enterprise to be predicted to obtain a standardized result; Mapping the industry classification data in the business data of the enterprise to be predicted to obtain a mapping result; The standardized results and the mapping results are integrated and matched with the commodities in the total commodity pool to obtain a corporate portrait.

5. The method for predicting the main products of an enterprise according to claim 1, characterized in that: The method of using cosine similarity and a fast matching algorithm to determine a standard business scope close to the business scope in the enterprise portrait to obtain a set of candidate standard business scopes includes: Combined with the Aho-Corasick algorithm, the standard business scope corresponding to the business scope in the enterprise portrait is quickly identified to obtain a quick identification result; Determining whether the quick identification result is to find a standard business scope; If the quick identification result finds a standard business scope, the standard business scope corresponding to the identification result is used as a candidate standard business scope set; If the rapid identification result does not find the standard business scope, the business scope in the enterprise portrait and the standard business scope in the business scope specification statement are vectorized, and the cosine similarity of the two vectors is calculated, and the candidate standard business scope set is determined based on the cosine similarity.

6. The method for predicting the main products of an enterprise according to claim 1, characterized in that: The set priority includes sequentially displaying the commodities in the identification results belonging to the single-enterprise commodity pool, the commodities in the identification results belonging to the industry statistics commodity pool, and the commodities in the identification results belonging to the standard commodity pool.

7. The enterprise's main product forecasting device is characterized by: include: A total commodity pool building unit, used for extracting and standardizing commodity information through multiple data sources to form a total commodity pool; An acquisition unit, used for acquiring the business data of the enterprise to be predicted; A portrait construction unit, used to construct an enterprise portrait for the enterprise operation data to be predicted; A prediction unit, used to predict the main products of the enterprise according to the enterprise portrait, the total product pool, and the set knowledge graph; The prediction unit comprises: A scope determination subunit, for determining a standard business scope close to the business scope in the enterprise portrait by using cosine similarity and a fast matching algorithm, so as to obtain a set of candidate standard business scopes; A screening subunit, used to screen out commodities involved in a candidate standard business scope set from the total commodity pool, and perform standardization to obtain a candidate commodity set; An analysis subunit, configured to perform intersection analysis on the candidate product set, use a set knowledge graph to perform relationship recognition to obtain a recognition result, and classify products from different sources as a set according to the recognition result; The display subunit is used to display the collection according to the set priority to determine the main products of the enterprise.

8. A computer device, characterized in that: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images