Method for realizing investment attraction recommendation based on enterprise portrait

Abstract

The invention particularly relates to a method for realizing investment attraction recommendation based on an enterprise portrait. According to the method for realizing investment attraction recommendation based on the enterprise portrait, enterprise information is collected for integration, standardized fields are constructed, enterprise names are unified, self-defined sensitive information is subjected to desensitization processing, and an enterprise unique identifier is established; data cleaning is carried out, and indexes with business significance are extracted from cleaned enterprise data in a user-defined mode; converting the cleaned data into an enterprise label system which can be quantified and updated; and analyzing the investment attraction demand, screening enterprise resources based on the enterprise portrait, and generating a recommendation result based on the large model capability, the content including a structured recommendation list and explanatory analysis. According to the method for realizing investment attraction recommendation based on the enterprise portraits, the enterprise portraits can be dynamically updated in real time, accurate docking of investment attraction demands and enterprise resources and scientific sorting of candidate enterprises are realized, and decision transparency and credibility are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method for implementing investment recommendation based on enterprise portraits. Background Art

[0002] With the urgent need to modernize and upgrade the industrial chain, attracting investment has become a core engine for high-quality urban economic development. However, intensified competition among regions has led to frequent instances of "homogeneous competition for companies." Some cities, due to blind attraction and rising costs for companies to establish businesses, have fallen into a dilemma of "hot signings, cold implementation."

[0003] Under the traditional investment promotion model, information asymmetry makes it difficult to accurately identify high-quality companies that are compatible with the regional industrial chain. Furthermore, reliance on manual experience causes investment promotion decisions to lag behind market dynamics.

[0004] Enterprise portrait refers to a comprehensive, systematic and structured digital description of an enterprise through multi-dimensional data collection, feature extraction and modeling analysis, thereby forming a quantifiable and analyzable "data image".

[0005] In recent years, urban investment promotion departments have actively explored digital transformation paths using enterprise portraits. Many areas have integrated industrial and commercial, tax, and patent data to build basic enterprise information databases to achieve initial enterprise screening. While this type of technology has initially solved the problem of data fragmentation and improved investment attraction efficiency to a certain extent, it still has certain limitations:

[0006] First, the data dimension is single, relying on one-sided data while ignoring unstructured indicators such as public opinion, supply chain, and innovation potential, resulting in a "static snapshot" feature of the enterprise portrait;

[0007] Second, the algorithm is not adaptable enough, and general recommendation models are difficult to capture the particularities of regional industrial chains;

[0008] Third, there is a lack of decision-making explainability, and black-box recommendation results reduce the trust of decision makers.

[0009] In this context, there is an urgent need to build an intelligent investment recommendation method that integrates multi-source heterogeneous data, dynamically perceives corporate status, and embeds regional economic logic, so as to reconstruct the accuracy and scientific nature of investment decision-making from the technical bottom layer.

[0010] Based on the above technical situation, the present invention proposes a method for implementing investment recommendation based on enterprise portrait. Summary of the Invention

[0011] In order to remedy the deficiencies of the prior art, the present invention provides a simple and efficient method for implementing investment recommendation based on enterprise portraits.

[0012] The present invention is achieved through the following technical solutions:

[0013] A method for implementing investment recommendation based on enterprise portrait, characterized by comprising the following steps:

[0014] Step S1: Multi-source data collection and integration

[0015] Collect enterprise information from various data sources with access rights, including structured data, unstructured data, and real-time dynamic data;

[0016] Integrate the collected enterprise data, build standardized fields, unify enterprise names, desensitize customized sensitive information, and establish a unique enterprise identifier;

[0017] Step S2: Data cleaning and feature engineering

[0018] Perform data cleaning on the integrated enterprise data to eliminate noise, fill missing values, correct contradictory data, and remove stop words and irrelevant symbols in text data;

[0019] Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features;

[0020] Step S3: Dynamic modeling of enterprise portraits

[0021] The process of converting cleaned data into a quantifiable and updateable enterprise labeling system is as follows:

[0022] First, we use K-means clustering to classify enterprise types, and verify the rationality of the classification based on industry rules.

[0023] Secondly, the Prophet time series model is used to analyze enterprise trend indicators and dynamically adjust label weights;

[0024] Step S4: Matching investment needs with resources

[0025] Analyze investment needs, screen enterprise resources based on enterprise profiles, and accurately match regional investment needs with enterprise resources to ensure that recommended enterprises are in line with the direction of local industrial development and have high feasibility of implementation;

[0026] Step S5: Generate and explain recommendation results

[0027] Generate recommendation results based on large model capabilities, including structured recommendation lists and explanatory analysis.

[0028] In step S1, structured data provides a trustworthy benchmark, including business registration information, equity structure, qualification certification, listing information, financial statements, tax records, social security payment, patent data, trademarks and copyrights, R&D investment, litigation records, administrative penalties and credit ratings;

[0029] Mining deep connections in unstructured data, including news media reports, social media discussions, industry research reports, corporate recruitment information, and multimedia information;

[0030] Real-time dynamic data ensures timeliness, including procurement bid announcements, logistics data, real-time equity changes, secondary market data, financing dynamics, IoT sensor data and emergencies.

[0031] In step S2, basic features are directly extracted from the original data fields without complex calculations, including enterprise attributes and operating indicators;

[0032] Derived characteristics: generated through statistics, mathematical transformations, or business rules, including R&D investment intensity (R&D expenses / revenue) and per capita output value (revenue / number of employees).

[0033] Relationship characteristics: Based on the mining of associations between enterprises or between enterprises and external entities, including the number of upstream suppliers, the concentration of downstream customers, and the geographical distance from the investment park;

[0034] Semantic features: extracted from unstructured text based on natural language processing (NLP) technology, including public opinion sentiment, technical keywords, and rule matching.

[0035] In step S3, K-means clustering technology is combined with Prophet time series prediction technology, and new data access is maintained to achieve dynamic updating of the enterprise portrait.

[0036] The K-means clustering implementation process is as follows:

[0037] First, we selected numerical features that are strongly correlated with company type, including registered capital, R&D investment ratio, receivables growth rate over the past three years, total number of patents, and supply chain complexity, and then performed data standardization.

[0038] Secondly, the elbow rule is used to determine the K value and then generate labels. The K value is the number of categories.

[0039] Taking four categories as an example, the K value is 4: including labor-intensive (low R&D, high supply chain), technology-leading (high patents, high R&D), capital-driven (high registered capital, low growth) and growth potential (high growth, low scale);

[0040] Finally, manual review is conducted and the clustering results are corrected by superimposing industry rules.

[0041] The Prophet time series forecasting process is as follows:

[0042] First, prepare the data for time series fields, including quarterly corporate revenue, monthly patent numbers, and annual R&D investment.

[0043] Then, model training and prediction are carried out, trends are scored, and companies are divided into three categories: high-growth, stable growth, and recession risk.

[0044] In step S4, the investment demand analysis includes explicit demand structuring and implicit demand mining;

[0045] Among them, explicit demand structuring involves extracting keywords from investment promotion rules and documents, while implicit demand mining involves analyzing gaps in the industry chain and past successful investment promotion cases.

[0046] Enterprise resource screening includes SQL-style rule filtering for hard condition initial screening and soft condition scoring;

[0047] Among them, the initial screening of hard conditions is achieved through SQL-style rule filtering, and the soft condition scoring is obtained by calculating the similarity between the company's technology and the investment promotion rule requirements.

[0048] In step S5, the matching results are converted into understandable recommendation solutions, and multi-dimensional explanations are used to assist investment promotion personnel in making decisions;

[0049] The weighted scoring model assigns comprehensive scores to companies based on industry compatibility, industrial chain coordination, investment promotion rule matching, and risk factor, and sorts them from high to low to generate a recommendation list in a table format;

[0050] Automatically generate structured natural language reports, converting complex matching logic into intuitive text descriptions.

[0051] A system for implementing investment recommendation based on enterprise portraits, used to implement the above method, including a multi-source data collection and integration module, a data cleaning and feature engineering module, an enterprise portrait dynamic modeling module, an investment demand and resource matching module, and a recommendation result generation and interpretation module;

[0052] The multi-source data collection and integration module is responsible for collecting enterprise information from various data sources with access rights. The data sources include structured data, unstructured data, and real-time dynamic data.

[0053] Integrate the collected enterprise data, build standardized fields, unify enterprise names, desensitize customized sensitive information, and establish a unique enterprise identifier;

[0054] The data cleaning and feature engineering module is responsible for cleaning the integrated enterprise data, eliminating noise, filling missing values, correcting contradictory data, and removing stop words and irrelevant symbols in text data;

[0055] Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features;

[0056] The enterprise portrait dynamic modeling module is responsible for converting the cleaned data into a quantifiable and updateable enterprise label system. The process is as follows:

[0057] First, we use K-means clustering to classify enterprise types, and verify the rationality of the classification based on industry rules.

[0058] Secondly, the Prophet time series model is used to analyze enterprise trend indicators and dynamically adjust label weights;

[0059] The investment demand and resource matching module is responsible for analyzing investment demand, screening enterprise resources based on enterprise profiles, and accurately matching regional investment demand with enterprise resources to ensure that the recommended enterprises are both in line with the direction of local industrial development and have high feasibility of implementation;

[0060] The recommendation result generation and interpretation module is responsible for generating recommendation results based on the large model capabilities, including structured recommendation lists and explanatory analysis.

[0061] A device for implementing investment recommendation based on corporate portraits, characterized in that it includes a memory and a processor; the memory is used to store computer programs, and the processor is used to implement the above-mentioned method steps when executing the computer program.

[0062] A readable storage medium, characterized in that: a computer program is stored on the readable storage medium, and the computer program implements the above method steps when executed by a processor.

[0063] The beneficial effect of the present invention is that the method for implementing investment recommendation based on enterprise portrait can dynamically update enterprise portrait in real time, realize the precise connection between investment demand and enterprise resources and scientific ranking of candidate enterprises, and improve the transparency and credibility of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Attachment Figure 1 This is a schematic diagram of the method for implementing investment recommendation based on enterprise portraits in the present invention. DETAILED DESCRIPTION

[0066] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0067] The method for implementing investment recommendation based on enterprise portrait includes the following steps:

[0068] Step S1: Multi-source data collection and integration

[0069] Collect enterprise information from various data sources with access rights, including structured data, unstructured data and real-time dynamic data; multi-source data collection and integration is the basic link in building enterprise portraits, aiming to obtain comprehensive and multi-dimensional enterprise information through multiple channels.

[0070] Integrate the collected enterprise data, build standardized fields, unify enterprise names, consider data security and compliance, desensitize customized sensitive information, and establish a unique enterprise identifier;

[0071] Step S2: Data cleaning and feature engineering

[0072] Perform data cleaning on the integrated enterprise data to eliminate noise, fill missing values, and correct contradictory data, such as removing duplicate enterprise records and fixing incorrectly formatted fields. Remove stop words and irrelevant symbols from text data;

[0073] Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features;

[0074] Step S3: Dynamic modeling of enterprise portraits

[0075] The process of converting cleaned data into a quantifiable and updateable enterprise labeling system is as follows:

[0076] First, we use K-means clustering to classify enterprises into different types (e.g., “technology-intensive” and “labor-intensive”), and verify the rationality of the classification by combining industry rules.

[0077] Secondly, the Prophet time series model is used to analyze corporate trend indicators (such as revenue fluctuations and patent growth) and dynamically adjust label weights (such as lowering the "growth" score for companies in recession);

[0078] Step S4: Matching investment needs with resources

[0079] Analyze investment needs, screen enterprise resources based on enterprise profiles, and accurately match regional investment needs with enterprise resources to ensure that recommended enterprises are in line with the direction of local industrial development and have high feasibility of implementation;

[0080] Step S5: Generate and explain recommendation results

[0081] Generate recommendations based on the capabilities of the large model, including a structured list of recommendations and explanatory analysis. This allows investment promotion personnel to clearly understand why this company is recommended and provides support for implementation decisions.

[0082] In step S1, a multimodal database with enterprises as core nodes is formed to provide raw materials for subsequent analysis. The data sources mainly include three categories:

[0083] Structured data provides a trusted benchmark to answer the question of "who the company is." It includes business registration information, equity structure, qualification certification, listing information, financial statements, tax records, social security payments, patent data, trademarks and copyrights, R&D investment, litigation records, administrative penalties, and credit ratings.

[0084] Unstructured data mining to uncover deep connections and answer the question "What potential / risk does the company have?" includes news media reports, social media discussions, industry research reports, corporate recruitment information, and multimedia information.

[0085] Real-time dynamic data ensures timeliness and supports the question of "whether the current situation is suitable for investment promotion", including procurement winning bid announcements, logistics data, real-time equity changes, secondary market data, financing dynamics, IoT sensor data and emergencies.

[0086] In step S2, the original data is converted into features that can effectively express business problems and improve model performance. Features include the following:

[0087] Basic features: directly extracted from raw data fields without complex calculations, including enterprise attributes and operating indicators;

[0088] Derived characteristics: generated through statistics, mathematical transformations, or business rules, including R&D investment intensity (R&D expenses / revenue) and per capita output value (revenue / number of employees).

[0089] Relationship characteristics: Based on the mining of associations between enterprises or between enterprises and external entities, including the number of upstream suppliers, the concentration of downstream customers, and the geographical distance from the investment park;

[0090] Semantic features: extracted from unstructured text based on natural language processing (NLP) technology, including public opinion sentiment, technical keywords, and rule matching.

[0091] In step S3, K-means is used to answer the question "What type of company is it?", and Prophet is used to answer the question "What will the company's future look like?" By combining K-means clustering technology with Prophet time series forecasting technology and continuously ingesting new data, the company's profile is dynamically updated.

[0092] The K-means clustering implementation process is as follows:

[0093] First, we selected numerical features that are strongly correlated with company type, including registered capital, R&D investment ratio, receivables growth rate over the past three years, total number of patents, and supply chain complexity, and then performed data standardization.

[0094] Secondly, the elbow rule is used to determine the K value and then generate labels. The K value is the number of categories.

[0095] Taking four categories as an example, the K value is 4: including labor-intensive (low R&D, high supply chain), technology-leading (high patents, high R&D), capital-driven (high registered capital, low growth) and growth potential (high growth, low scale);

[0096] Finally, manual review is conducted and the clustering results are corrected by superimposing industry rules.

[0097] The Prophet time series forecasting process is as follows:

[0098] First, prepare the data for time series fields, including quarterly corporate revenue, monthly patent numbers, and annual R&D investment.

[0099] Then, model training and prediction are carried out, trends are scored, and companies are divided into three categories: high-growth, stable growth, and recession risk.

[0100] In step S4, the investment demand analysis includes explicit demand structuring and implicit demand mining;

[0101] Among them, explicit demand structuring involves extracting keywords from investment promotion rules and documents, while implicit demand mining involves analyzing gaps in the industry chain and past successful investment promotion cases.

[0102] Enterprise resource screening includes SQL-style rule filtering for hard condition initial screening and soft condition scoring;

[0103] Among them, the initial screening of hard conditions is achieved through SQL-style rule filtering, and the soft condition scoring is obtained by calculating the similarity between the company's technology and the investment promotion rule requirements.

[0104] In step S5, the matching results are converted into understandable recommendation solutions, and multi-dimensional explanations are used to assist investment promotion personnel in making decisions;

[0105] The weighted scoring model assigns comprehensive scores to companies based on industry compatibility, industrial chain coordination, investment promotion rule matching, and risk factor, and sorts them from high to low to generate a recommendation list in a table format;

[0106] Automatically generate structured natural language reports, converting complex matching logic into intuitive text descriptions.

[0107] The system for implementing investment recommendation based on enterprise portraits is used to implement the above method, including a multi-source data collection and integration module, a data cleaning and feature engineering module, an enterprise portrait dynamic modeling module, an investment demand and resource matching module, and a recommendation result generation and interpretation module;

[0108] The multi-source data collection and integration module is responsible for collecting enterprise information from various data sources with access rights. The data sources include structured data, unstructured data, and real-time dynamic data.

[0109] Integrate the collected enterprise data, build standardized fields, unify enterprise names, desensitize customized sensitive information, and establish a unique enterprise identifier;

[0110] The data cleaning and feature engineering module is responsible for cleaning the integrated enterprise data, eliminating noise, filling missing values, correcting contradictory data, and removing stop words and irrelevant symbols in text data;

[0111] Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features;

[0112] The enterprise portrait dynamic modeling module is responsible for converting the cleaned data into a quantifiable and updateable enterprise label system. The process is as follows:

[0113] First, we use K-means clustering to classify enterprise types, and verify the rationality of the classification based on industry rules.

[0114] Secondly, the Prophet time series model is used to analyze enterprise trend indicators and dynamically adjust label weights;

[0115] The investment demand and resource matching module is responsible for analyzing investment demand, screening enterprise resources based on enterprise profiles, and accurately matching regional investment demand with enterprise resources to ensure that the recommended enterprises are both in line with the direction of local industrial development and have high feasibility of implementation;

[0116] The recommendation result generation and interpretation module is responsible for generating recommendation results based on the large model capabilities, including structured recommendation lists and explanatory analysis.

[0117] The device for implementing investment recommendation based on corporate portraits includes a memory and a processor; the memory is used to store computer programs, and the processor is used to implement the above-mentioned method steps when executing the computer program.

[0118] The readable storage medium stores a computer program, which implements the above method steps when executed by a processor.

[0119] Compared with existing technologies, this method for implementing investment recommendation based on enterprise portraits has the following characteristics:

[0120] (1) Comprehensive coverage of multi-source data: By integrating multi-dimensional data such as industry and commerce, finance, and public opinion, a complete corporate portrait is constructed to avoid the problem of one-sided information in traditional investment promotion and lay a solid foundation for accurate recommendations.

[0121] (2) Real-time update of dynamic portraits: Combining time series analysis and clustering algorithms, the corporate portrait can be dynamically adjusted as business conditions change, maintaining the timeliness and accuracy of the recommendation basis.

[0122] (3) Accurate analysis and matching of demand: Deeply explore explicit and implicit investment needs, and achieve accurate matching of demand and enterprise resources through intelligent algorithms, greatly improving the matching fit.

[0123] (4) Multi-dimensional evaluation and scientific ranking: Establish a multi-dimensional scoring system that includes industrial collaboration, investment promotion rule adaptation, etc., to achieve scientific ranking of candidate companies and optimize recommendation priorities.

[0124] (5) Explainable recommendations enhance trust: Through intuitive visual reports and natural language explanations, the recommendation logic and basis are clearly displayed, improving the transparency and credibility of decision-making.

[0125] The embodiment described above is only one specific implementation of the present invention. Common changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for implementing investment recommendation based on enterprise portrait, characterized by: The following steps are involved: Step S1: Multi-source data collection and integration Collect enterprise information from various data sources with access rights, including structured data, unstructured data, and real-time dynamic data; Integrate the collected enterprise data, build standardized fields, unify enterprise names, desensitize customized sensitive information, and establish a unique enterprise identifier; Step S2: Data cleaning and feature engineering Perform data cleaning on the integrated enterprise data to eliminate noise, fill missing values, correct contradictory data, and remove stop words and irrelevant symbols in text data; Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features; Step S3: Dynamic modeling of enterprise portraits The process of converting cleaned data into a quantifiable and updateable enterprise labeling system is as follows: First, we use K-means clustering to classify enterprise types, and verify the rationality of the classification based on industry rules. Secondly, the Prophet time series model is used to analyze enterprise trend indicators and dynamically adjust label weights; Step S4: Matching investment needs with resources Analyze investment needs, screen enterprise resources based on enterprise profiles, and accurately match regional investment needs with enterprise resources to ensure that recommended enterprises are both in line with the direction of local industrial development and feasible for implementation; Step S5: Generate and explain recommendation results Generate recommendation results based on large model capabilities, including structured recommendation lists and explanatory analysis.

2. The method for implementing investment recommendation based on enterprise portrait according to claim 1, characterized in that: In step S1, structured data provides a trustworthy benchmark, including business registration information, equity structure, qualification certification, listing information, financial statements, tax records, social security payment, patent data, trademarks and copyrights, R&D investment, litigation records, administrative penalties and credit ratings; Mining deep connections in unstructured data, including news media reports, social media discussions, industry research reports, corporate recruitment information, and multimedia information; Real-time dynamic data ensures timeliness, including procurement bid announcements, logistics data, real-time equity changes, secondary market data, financing dynamics, IoT sensor data and emergencies.

3. The method for implementing investment recommendation based on enterprise portrait according to claim 1, characterized in that: In step S2, basic features are directly extracted from the original data fields without complex calculations, including enterprise attributes and operating indicators; Derived characteristics: generated through statistics, mathematical transformations, or business rules, including R&D investment intensity (i.e., R&D expenditure / revenue) and per capita output value; Relationship characteristics: Based on the mining of associations between enterprises or between enterprises and external entities, including the number of upstream suppliers, the concentration of downstream customers, and the geographical distance from the investment park; Semantic features: extracted from unstructured text based on natural language processing (NLP) technology, including public opinion sentiment, technical keywords, and rule matching.

4. The method for implementing investment recommendation based on enterprise portrait according to claim 1, characterized in that: In step S3, K-means clustering technology is combined with Prophet time series prediction technology, and new data access is maintained to achieve dynamic updating of the enterprise portrait; The K-means clustering implementation process is as follows: First, we selected numerical features that are strongly correlated with company type, including registered capital, R&D investment ratio, receivables growth rate over the past three years, total number of patents, and supply chain complexity, and then performed data standardization. Secondly, the elbow rule is used to determine the K value and then generate labels. The K value is the number of categories. Finally, manual review is performed and the clustering results are corrected by adding industry rules; The Prophet time series forecasting process is as follows: First, prepare the data for time series fields, including quarterly corporate revenue, monthly patent numbers, and annual R&D investment. Then, model training and prediction are carried out, trends are scored, and companies are divided into three categories: high-growth, stable growth, and recession risk.

5. The method for implementing investment recommendation based on enterprise portrait according to claim 1, characterized in that: In step S4, the investment demand analysis includes explicit demand structuring and implicit demand mining; Among them, explicit demand structuring involves extracting keywords from investment promotion rules and documents, while implicit demand mining involves analyzing gaps in the industry chain and past successful investment promotion cases. Enterprise resource screening includes SQL-style rule filtering for hard condition initial screening and soft condition scoring; Among them, the initial screening of hard conditions is achieved through SQL-style rule filtering, and the soft condition scoring is obtained by calculating the similarity between the company's technology and the investment promotion rule requirements.

6. The method for implementing investment recommendation based on enterprise portrait according to claim 1, characterized in that: In step S5, the matching results are converted into understandable recommendation solutions, and multi-dimensional explanations are used to assist investment promotion personnel in making decisions; The weighted scoring model assigns comprehensive scores to companies based on industry compatibility, industrial chain coordination, investment promotion rule matching, and risk factor, and sorts them from high to low to generate a recommendation list in a table format; Automatically generate structured natural language reports, converting matching logic into intuitive text descriptions.

7. A system for implementing investment recommendation based on enterprise portraits, characterized by: Used to implement the method according to any one of claims 1 to 6, comprising a multi-source data collection and integration module, a data cleaning and feature engineering module, a corporate portrait dynamic modeling module, an investment demand and resource matching module, and a recommendation result generation and interpretation module; The multi-source data collection and integration module is responsible for collecting enterprise information from various data sources with access rights. The data sources include structured data, unstructured data, and real-time dynamic data. Integrate the collected enterprise data, build standardized fields, unify enterprise names, desensitize customized sensitive information, and establish a unique enterprise identifier; The data cleaning and feature engineering module is responsible for cleaning the integrated enterprise data, eliminating noise, filling missing values, correcting contradictory data, and removing stop words and irrelevant symbols in text data; Customize the extraction of business-significant indicators from cleaned enterprise data, including basic features, derived features, relationship features, and semantic features; The enterprise portrait dynamic modeling module is responsible for converting the cleaned data into a quantifiable and updateable enterprise label system. The process is as follows: First, we use K-means clustering to classify enterprise types, and verify the rationality of the classification based on industry rules. Secondly, the Prophet time series model is used to analyze enterprise trend indicators and dynamically adjust label weights; The investment demand and resource matching module is responsible for analyzing investment demand, screening enterprise resources based on enterprise profiles, and accurately matching regional investment demand with enterprise resources to ensure that the recommended enterprises are both in line with the direction of local industrial development and have feasibility of implementation; The recommendation result generation and interpretation module is responsible for generating recommendation results based on the large model capabilities, including structured recommendation lists and explanatory analysis.

8. A device for implementing investment recommendation based on enterprise portraits, characterized by: including memory and processor; The memory is used to store a computer program, and the processor is used to implement the method steps according to any one of claims 1 to 6 when executing the computer program.

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

Images