A multi-modal and knowledge base-based enterprise site selection driving force analysis method and system

Abstract

The application provides a multi-modal and knowledge base-based enterprise site selection dynamic analysis method, which comprises collecting enterprise information, industry information, park information and policy information and constructing a structured knowledge graph; extracting a historical site selection data set from the knowledge graph with enterprise development time as the axis, analyzing the historical site selection data set and obtaining key factors affecting site selection, integrating the key factors to obtain a first site selection feature set; collecting multi-modal data of the enterprise, extracting a development environment data set from the multi-modal data through text analysis, image analysis and audio analysis, combining the development environment data set with the knowledge graph through a data fusion method, and obtaining a second site selection feature set; calculating the correlation of the first site selection feature set and the second site selection feature set, constructing an enterprise site selection prediction model through a machine learning algorithm; and generating an enterprise-oriented investment attraction strategy according to the prediction result of the enterprise site selection prediction model.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of investment software, and particularly relates to a method and system for analyzing the driving force of enterprise site selection based on multi-modal and knowledge base. BACKGROUND

[0002] In today's competitive business environment, office park marketing faces many challenges. On the one hand, with the rapid development of economy and the continuous advancement of urbanization, various office parks have sprung up like mushrooms, and market competition is becoming increasingly fierce. On the other hand, when choosing an office location, enterprises no longer focus on a single factor such as rent price, but pay more attention to comprehensive consideration of multiple factors, including the convenience of geographical location, the comfort of the surrounding environment, the perfection of facilities and equipment, the industry agglomeration effect, and the availability of talent resources, etc. Accurate understanding of the site selection needs and preferences of enterprises is crucial to improving the success rate of office park marketing. Traditional marketing methods often rely on limited market research and subjective judgment, lacking in-depth insight into the real needs of enterprises. This approach is not only inefficient, but also difficult to accurately attract potential customers that meet the positioning of the park.

[0003] Traditional office park marketing methods usually have the following problems:

[0004] (1) Unclear marketing target, traditional office park marketing methods are usually extensive, lacking in-depth analysis of the site selection driving force of potential enterprises. Without fully understanding the actual needs and decision-making factors of enterprises, marketing work is often like finding a needle in a haystack, blindly promoting and soliciting, resulting in unclear marketing targets.

[0005] (2) Difficult to meet individual needs of enterprises, different enterprises have distinct individual needs when choosing an office park site, but existing marketing models are difficult to accurately grasp these needs, making it difficult to provide targeted marketing solutions.

[0006] (3) Lack of effective analysis tools, existing marketing analysis tools have obvious limitations when faced with complex office park marketing scenarios, and are difficult to consider multi-modal data and enterprise knowledge graphs, making it difficult to provide strong support for marketing decisions. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application proposes a method for analyzing the driving force of enterprise site selection based on multi-modal and knowledge base, comprising

[0008] Step 1: Collect enterprise information, industry information, park information, and policy information and build a structured knowledge graph;

[0009] Step 2: Extract a historical site selection dataset from the knowledge graph with enterprise development time as the axis, analyze the historical site selection dataset, and obtain key factors affecting site selection, and integrate the key factors to obtain a first site selection feature set;

[0010] Step 3: Collect multi-modal data of the enterprise, extract a development environment dataset from the multi-modal data through text analysis, image analysis, and audio analysis, combine the development environment dataset with the knowledge graph through a data fusion method, and obtain a second site selection feature set;

[0011] Step 4: Based on the first site selection feature set and the second site selection feature set, construct an enterprise site selection prediction model through a machine learning algorithm;

[0012] Step 5: According to the prediction result of the enterprise site selection prediction model, generate an enterprise-oriented investment strategy.

[0013] In a further technical solution of the embodiment, in step 2, the historical site selection dataset includes enterprise development information and enterprise address information, and when analyzing the historical site selection dataset and obtaining key factors affecting site selection, the following steps are further included:

[0014] Step 21: Collect historical relocation records of the enterprise;

[0015] Step 22: Analyze the relocation reasons and time regularity in the historical relocation records, and mine the key factors;

[0016] Step 23: Extract development features, recruitment features, leasing features, market promotion and business development features from the key factors through data analysis and processing.

[0017] In a further technical solution of the embodiment, extracting the development features includes the following steps:

[0018] Draw an enterprise revenue growth curve based on time series, mark the relocation time point on the enterprise revenue growth curve, analyze the change trend of the enterprise revenue growth curve before and after the relocation time point, analyze the change trend of the revenue growth change trend before and after the relocation time point, and classify the cities with higher revenue growth change trend;

[0019] Draw a market share change curve based on time series, mark the relocation time point on the market share change curve, analyze the change trend of the market share before and after the relocation time point, and summarize the regional competitors before and after the relocation.

[0020] In a further technical solution of the embodiment, extracting the recruitment features includes the following steps:

[0021] Based on the time series, a recruitment number change curve is drawn, and a relocation time point is marked on the recruitment number change curve, and the recruitment number change of the enterprise before and after the relocation time point is analyzed.

[0022] Based on the enterprise's job demand structure, the relevance of the region after the relocation time point to the job demand structure is analyzed.

[0023] In the further technical solution of the embodiment, the rental features are extracted, including the following steps:

[0024] Based on the site selection information of the enterprise before and after the relocation time point, the change of the site selection decision of the enterprise and the change of the site utilization efficiency are analyzed.

[0025] In the further technical solution of the embodiment, the market promotion and business development features are extracted, including the following steps:

[0026] Based on the enterprise, the consistency of the enterprise's market promotion key area and development direction with the relocation location is analyzed.

[0027] In the further technical solution of the embodiment, the step 3 includes the following steps:

[0028] The multi-modal data is obtained through a crawler algorithm or an API interface, and the multi-modal data is processed through data cleaning and data correction;

[0029] The entity data in the multi-modal data is identified, and the entity data is aligned and mapped with the entities of the knowledge graph;

[0030] The features extracted from the multi-modal data are quantified and vectorized, and the multi-modal feature vectors are fused with the vector representations of the corresponding entities in the knowledge graph;

[0031] Based on the information in the multi-modal data, new relationships are mined and constructed in the entities of the knowledge graph, and the relationships are updated to the knowledge graph.

[0032] In the further technical solution of the embodiment, the step 4 further includes the following steps:

[0033] Data processing of the first site selection feature set and the second site selection feature set;

[0034] Feature selection based on correlation analysis: calculate the correlation between each feature and the possibility of enterprise relocation, select features with high correlation into the model, and eliminate redundant or irrelevant features;

[0035] The weight of different features is determined by using a feature importance evaluation method.

[0036] The first site selection feature set data processing includes the following steps: cleaning and arranging the collected historical site selection decision data, and performing standardization processing.

[0037] The time series data is converted into a machine learning model input format, and the classification data is encoded and converted into a digital form.

[0038] The data processing of the second site selection feature set includes extracting key information from text data through natural language processing technology and converting it into a feature vector; and normalizing the extracted key feature vector of the image data to match the numerical range with other features.

[0039] In the further technical solution of the embodiment, in step 5, when generating the enterprise-oriented investment strategy, the following steps are further included:

[0040] A strategy library is established, and the strategy classification in the strategy library includes enterprise type strategy, enterprise size strategy and industry characteristic strategy.

[0041] The prediction result is analyzed, and the corresponding strategy is extracted from the strategy library to generate an investment strategy for the prediction result.

[0042] The application also provides an enterprise site selection dynamic analysis system based on multi-modal and knowledge base, comprising: a knowledge graph construction and data integration module for collecting enterprise information, industry information, park information and policy information and constructing a structured knowledge graph;

[0043] A historical data analysis module is used to extract a historical site selection data set from the knowledge graph with enterprise development time as the axis, analyze the historical site selection data set and obtain key factors affecting site selection, and integrate the key factors to obtain a first site selection feature set;

[0044] A multi-modal feature extraction module is used to collect multi-modal data of the enterprise, extract a development environment data set from the multi-modal data through text analysis, image analysis and audio analysis, combine the development environment data set with the knowledge graph through a data fusion method, and obtain a second site selection feature set;

[0045] A dynamic analysis and prediction module is used to calculate the correlation of the first site selection feature set and the second site selection feature set, and construct an enterprise site selection prediction model through a machine learning algorithm;

[0046] An investment strategy generation module is used to generate an enterprise-oriented investment strategy according to the prediction result of the enterprise site selection prediction model.

[0047] The embodiment of the present application can clearly understand the key information of the enterprise, such as the industrial chain relationship, the cooperation partner, the competitor, the development strategy and the like of the enterprise, and the first site selection feature set is extracted by the position of the office building park in the industry, the competitive relationship with other parks and the like through mining in the structured knowledge graph and through the enterprise knowledge graph. Meanwhile, the second site selection feature set is extracted by mining features from multi-modal data through various analysis methods, and the multi-modal data covers various forms of data such as text, image, audio, geographic location information and the like, and can reflect the enterprise and the office building park from different angles. The prediction model considering various complex and diversified enterprise site selection driving factors is constructed through the two feature sets, so as to provide a basis for park business attraction. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor.

[0049] Figure 1 The step flow chart of the enterprise site selection driving force analysis method based on multi-modal and knowledge base of the present application;

[0050] Figure 2 The structural schematic diagram of the enterprise site selection driving force analysis system based on multi-modal and knowledge base of the present application. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. The well-known modules, units and their mutual connection, link, communication or operation are not shown or not described in detail. And the described features, architectures or functions can be combined in any way in one or more embodiments. Those skilled in the art should understand that the following various embodiments are only used for example, not for limiting the protection scope of the present application. It can also be easily understood that the modules or units or processing methods in each embodiment described herein and shown in the drawings can be combined and designed in various different configurations. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] The traditional business attraction method usually has the following deficiencies:

[0053] Lack of systematic market research. Traditional marketing methods may rely solely on simple questionnaires or limited industry reports, failing to fully and accurately grasp the various factors considered by different types of enterprises when selecting locations. For example, emerging technology companies may be more concerned about the surrounding innovation ecosystem, the distribution of universities and research institutions, and talent reserves, but traditional methods often fail to delve into these specific needs.

[0054] Dependence on subjective experience. Marketing personnel often rely on personal experience and intuition to determine potential target enterprises, lacking objective data support. This subjective judgment is prone to bias, leading to inaccurate marketing direction and wasting a lot of time and resources on enterprises that do not have actual moving possibilities, resulting in extremely low efficiency.

[0055] Inability to track market dynamics in a timely manner. As the market continues to change, the location selection needs of enterprises are also dynamically adjusting. Traditional marketing methods are unable to track these changes in real time and adjust marketing strategies in a timely manner, leaving marketing work in a passive state.

[0056] Differences in enterprise types. Enterprises in different industries have vastly different requirements for office environments. For example, financial enterprises may place greater emphasis on the high-end image of office buildings, security facilities, and proximity to financial regulatory agencies, while creative industry enterprises prefer office spaces with a creative atmosphere and open layout. Traditional marketing methods are unable to provide personalized solutions tailored to the characteristics of different industries.

[0057] Different development stages of enterprises. Enterprises at different stages of development have different location selection needs. Start-ups may focus on cost control, entrepreneurship support policies, and proximity to potential partners, while mature enterprises may prioritize brand image, office space expansion, and surrounding commercial support. Existing marketing methods often fail to accurately target and serve enterprises based on their development stage.

[0058] Diversification of employee needs. The needs of enterprise employees also influence enterprise location decisions. For example, young employees may prioritize surrounding entertainment facilities, convenient transportation, and housing costs, while employees with families may value surrounding educational resources and medical facilities. Traditional marketing solutions often overlook the impact of employee needs on enterprise location decisions, failing to meet the individual needs of enterprises in this regard.

[0059] Single data source. Traditional analysis tools rely primarily on statistical data, questionnaires, and other single data sources, failing to fully utilize the rich information contained in multi-modal data. For example, image data can visually demonstrate the appearance and internal facilities of office building parks, and audio data can reflect the noise level of the surrounding environment, but these data are often overlooked in existing tools.

[0060] Lack of knowledge graph integration capability. Enterprise knowledge graph can integrate various knowledge and information inside and outside the enterprise, providing strong support for in-depth analysis of enterprise site selection motivation. However, most of the existing investment analysis tools lack effective use of knowledge graph, and cannot build comprehensive and accurate enterprise relationship network and industry ecosystem, making it difficult to understand the site selection decision-making process of enterprises from a deep level.

[0061] Lack of knowledge graph integration capability. Enterprise knowledge graph can integrate various knowledge and information inside and outside the enterprise, providing strong support for in-depth analysis of enterprise site selection motivation. However, most of the existing investment analysis tools lack effective use of knowledge graph, and cannot build comprehensive and accurate enterprise relationship network and industry ecosystem, making it difficult to understand the site selection decision-making process of enterprises from a deep level.

[0062] In order to solve the above problems, the present application uses multi-modal data and enterprise knowledge graph to build a prediction model, which can comprehensively and deeply understand the site selection needs and preferences of enterprises, and develop personalized investment strategies for office park, so as to attract potential customers to settle down, improve the success rate of investment, and promote the sustainable development of office park, as described below.

[0063] First embodiment:

[0064] As shown in Figure 1 , the present embodiment discloses a method for analyzing the site selection motivation of enterprises based on multi-modal and knowledge base, which includes

[0065] Step 1: Collect enterprise information, industry information, park information and policy information and build a structured knowledge graph;

[0066] The enterprise knowledge graph covers the basic information of the enterprise, such as enterprise name, establishment time, industry to which it belongs, organizational structure, etc.; the business information of the enterprise, including main products or services, market share, competitors, etc.; the relationship network of the enterprise, such as upstream and downstream enterprises in the industry chain, partners, etc. At the same time, integrate multi-source data, including information of office park, such as geographic location, facility condition, rent price, etc., and external data, such as policies and regulations, industry trends, etc. Through ontology construction, entity recognition and relationship extraction, etc., these data are organized into a structured knowledge graph, providing a basis for subsequent analysis.

[0067] Step 2: Extract historical site selection data set from the knowledge graph with enterprise development time as the axis, analyze the historical site selection data set and obtain key factors affecting site selection, and integrate the key factors to obtain a first site selection feature set; In this embodiment, the migration of the enterprise is analyzed with historical time as the anchor point. The past site selection decision data of the enterprise is collected, and the historical site selection data set includes the time, reason, target location, etc. of the migration. Through in-depth analysis of these historical data, important factors affecting the site selection of the enterprise are mined. Specifically, in step 2, the historical site selection data set includes enterprise development information and enterprise address information, and when analyzing the historical site selection data set and obtaining key factors affecting site selection, the following steps are further included:

[0068] Step 21: Collect the historical migration records of the enterprise;

[0069] Step 22: Analyze the migration reasons and time regularity in the historical migration records, and mine the key factors;

[0070] Step 23: Extract development features, recruitment features, leasing features, market promotion and business development features from the key factors through data analysis and processing.

[0071] Extracting the development features includes the following steps: drawing an enterprise revenue growth curve based on time series, marking the migration time point on the enterprise revenue growth curve, analyzing the change trend of the enterprise revenue growth curve before and after the migration time point, analyzing the change trend of the revenue growth, and classifying the cities with higher revenue growth change trend; Compare the revenue growth amplitude of different migration locations to analyze the influence of different regional markets on enterprise revenue. Classify the target locations according to economic regions, city sizes, etc. to further determine which types of regions are more conducive to enterprise revenue growth; Some emerging cities may have policy support and emerging market demand, and have a unique role in promoting the revenue of specific enterprises.

[0072] Draw a market share change curve based on time series, mark the migration time point on the market share change curve, analyze the change trend of the market share before and after the migration time point, and summarize the regional competitors before and after the migration. Market share expansion and site selection, research the change of market share data before and after the migration. When the market share of the enterprise expands rapidly in the new location, it is necessary to analyze the distribution of competitors in the region where the new location is located. The enterprise migrated from a competitive first-tier city to a second-tier city with development potential in a science and technology park, taking advantage of the relatively relaxed competitive environment and policy support in the local area, and rapidly expanding the market share.

[0073] Extracting the recruitment features, including the following steps: drawing a recruitment number change curve based on a time series, marking the relocation time point on the recruitment number change curve, and analyzing the recruitment number change of the enterprise before and after the relocation time point; recruitment number and site selection, and statistics of the change of the recruitment number of the enterprise before and after each relocation. It is found that the recruitment number of the enterprise before relocation increases significantly, which indicates that the possibility of subsequent relocation to a new site is greater.

[0074] Based on the enterprise's job demand structure, the relevance of the area after the relocation time point to the job demand structure is analyzed. Job demand and site selection: according to the job demand structure of the enterprise in different periods, the relevance of the relocation decision is analyzed. If the enterprise greatly increases the demand for R&D positions in a certain stage, and the new site is selected in an area with a strong scientific and technological research atmosphere and relevant industrial clusters, it indicates that the satisfaction of the industrial environment to the enterprise's job demand is an important factor in site selection.

[0075] Extracting the lease features, including the following steps: based on the site selection information of the enterprise before and after the relocation time point, analyzing the change of the site selection decision of the enterprise and the change of the site utilization efficiency. Through the enterprise knowledge graph, it is found that there is a positive correlation between the lease expiration time of the enterprise and the relocation action, and the enterprise generally relocates before the lease expires. If the rent of the original lease location rises too high, the enterprise may look for a new site with a higher cost performance; but if the enterprise plans to relocate before the lease expires, and the new site has obvious advantages in resources, market, etc., it may mean that the enterprise has a long-term strategic layout, not just to reduce the lease cost. Compare the relocation selection under different lease expiration conditions. For relocation after long-term lease (such as more than 5 years), the enterprise may be more cautious in considering the stability and development potential of the new site; while relocation after short-term lease (such as 1-2 years) may focus more on flexibility and short-term cost control. At the same time, analyze the utilization efficiency of the site during the lease period. If the original site utilization rate is low and cannot be improved through modification, it may prompt the enterprise to choose a more suitable new site after the lease expires, and the site utilization efficiency is a potential key feature.

[0076] Extracting the market promotion and business development features, including the following steps:

[0077] Based on the enterprise, the consistency of the enterprise's market promotion key area and development direction with the relocation location is analyzed. If the enterprise's market promotion plan lists a certain area as a key market to expand, and then relocates the enterprise to that area or nearby, it indicates that the target market of market promotion has a direct impact on site selection. Research the relationship between market promotion indicators such as ad placement effect and marketing activity response rate in different areas and site selection. If the conversion rate of ad placement in a certain area is high and the participation rate of marketing activities is high, the enterprise may consider shifting the business focus to that area, including relocating some departments or businesses to that area to better utilize the good market promotion environment in that area.

[0078] Business development direction and site selection. When the business of an enterprise expands into new areas, analyze the requirements of the new business for site selection. For example, when an enterprise transforms from traditional manufacturing to intelligent manufacturing, it may need to move to an area with an advanced manufacturing industry cluster that can provide technical support and supporting services related to intelligent manufacturing. By analyzing the direction of business transformation of the enterprise and the industrial environment, technical resources, and other factors of the new site, determine the key relationship between the new business development needs and site selection.

[0079] Step 3: Collect multi-modal data of the enterprise, extract development environment data set from the multi-modal data through text analysis, image analysis and audio analysis, and combine the development environment data set with the knowledge graph through data fusion method to obtain the second site selection feature set. Those skilled in the art can understand that the multi-modal data of the enterprise includes text, image, audio, etc. The development dynamics, reputation and other features of the enterprise are extracted from the text data such as enterprise news, social media and industry reports by using text analysis technology. Through image analysis, the appearance and internal facilities of the office park are processed, and the features such as environment aesthetics and facility modernization degree are extracted. Combined with audio data, the noise situation of the surrounding environment is analyzed, and the features extracted from the multi-modal data are summarized to form a development environment data set. The development environment data set is combined with the knowledge graph to obtain a second site selection feature set with more comprehensive and accurate representation of enterprise and office park features.

[0080] In the further technical solution of the embodiment, the step 3 includes the following steps:

[0081] The multi-modal data is obtained through a crawler algorithm or an API interface, and the multi-modal data is processed through data cleaning and data correction. Most of the original data is obtained through crawling, but specific APIs or authorization may be required to obtain some high-quality image data, audio data, and manually collected and corrected local data. After data cleaning and data processing of the collected data, the required data information is extracted.

[0082] aligning and mapping the entity data in the multi-modal data with entities in the knowledge graph; first, identifying entities in the multi-modal data and existing entities in the knowledge graph. For example, entity such as enterprise name, office park name mentioned in the text data is extracted through named entity recognition technology and aligned with the corresponding enterprise and park entity in the knowledge graph. For the office park shown in the image data, the landmark building or feature of the park is identified through image recognition technology and matched with the description of the park in the knowledge graph. For the surrounding environment information involved in the audio data, such as specific noise sources (such as near the airport, factory, etc.), it is associated with the surrounding environment entity related to the geographical location of the park in the knowledge graph.

[0083] quantifying and vectorizing the features extracted from the multi-modal data, and fusing the multi-modal feature vectors with the vector representation of the corresponding entity in the knowledge graph; quantifying and vectorizing the features extracted from the multi-modal data. For text features, word vector models (such as Word2Vec, Bert, etc.) can be used to convert text describing enterprise development dynamics, reputation into vector form. Image features can be extracted through convolutional neural network (CNN) to obtain feature vectors such as environment aesthetics, facility modernization degree, etc. Audio features can be converted into feature vectors through audio processing algorithms. Then, these multi-modal feature vectors are fused with the vector representation of the corresponding entity in the knowledge graph. Weighted average, splicing, etc. can be used to integrate multi-modal feature vectors into the vector representation of enterprise and office park entities in the knowledge graph, thereby updating and enriching the feature information of the entity.

[0084] Based on the information in the multi-modal data, new relationships are mined and constructed in the entities of the knowledge graph, and the relationships are updated to the knowledge graph. For example, from the text data, it is found that an enterprise and a certain office park have a potential leasing relationship because of the specific business development needs; from the image data, it is seen that the internal facilities of the office park are adapted to the business type of the enterprise; from the audio data, it is inferred that the surrounding environmental noise has an impact on the office environment of the enterprise. These new relationships are added to the knowledge graph, and the existing relationships are updated and improved, such as the causal relationship between enterprise development dynamics and site selection decision, so that the knowledge graph more accurately reflects the complex relationship between enterprises and office parks.

[0085] Step 4: Based on the first and second sets of site selection features, an enterprise site selection prediction model is constructed by a machine learning algorithm; the machine learning algorithm can be a deep learning neural network, random forest, etc. By the first and second sets of site selection features, the enterprise site selection behavior can be modeled to consider multiple factors, such as enterprise development potential, recruitment, lease expiration, market promotion, business development, etc., as well as the characteristics of the office park, to predict the possibility of relocation and target location of the enterprise in the next one, two, or three years. The present application uses a knowledge graph containing historical data and multi-modal data to construct a model to predict the possibility of relocation and target location, fully considering the internal development factors (such as development potential, recruitment, etc.) and external environmental factors (such as office park characteristics), to provide a data basis for targeted business attraction of the park.

[0086] In a further technical solution of the present embodiment, step 4 further includes the following steps:

[0087] Data processing of the first and second sets of site selection features;

[0088] Data processing of the first set of site selection features includes the following steps: cleaning and organizing the collected historical site selection decision data. For example, for enterprise revenue growth data, remove outliers (such as large fluctuations in revenue due to special one-time events), and perform standardization processing to make it comparable. Convert time series data (such as revenue growth curve, market share change, etc.) into a format suitable for machine learning model input, such as extracting growth rate, change slope, etc. Encode classification data (such as relocation reason categories) into numerical form.

[0089] Data processing of the second set of site selection features includes the following steps: for text data, extract key information and convert it into a feature vector through natural language processing (NLP) technology. For example, use a topic model (such as LDA) to extract features related to enterprise strategy topics from enterprise news and industry reports, or use sentiment analysis tools to obtain sentiment polarity scores related to enterprise reputation. Normalize the environmental aesthetics, facility modernization degree, etc. feature vectors extracted from image data to match the numerical range with other features.

[0090] Feature selection based on correlation analysis: calculate the correlation between each feature and the possibility of enterprise relocation, for example, by calculating the Pearson correlation coefficient, find the correlation between the enterprise development potential indicator (such as revenue growth rate) and the possibility of relocation, and the correlation between the facility modernization degree of the office park and the willingness of the enterprise to move in. Select features with high correlation into the model, eliminate redundant or irrelevant features to reduce the complexity of the model and improve the generalization ability;

[0091] The weight of different features is determined by using feature importance evaluation methods such as feature importance evaluation in random forest, and higher weight is given to factors that play a key role in enterprise site selection decision, such as the matching degree of market promotion key area and office park location, so that the model pays more attention to these important features during training.

[0092] Step 5: According to the prediction results of the enterprise site selection prediction model, generate enterprise-oriented investment strategy. In this embodiment, for different types of enterprises and predicted relocation possibilities, corresponding preferential policies, service schemes, etc. are formulated. For example, for enterprises with higher relocation possibility, customized leasing schemes, decoration subsidies, business support, etc. are provided; for enterprises with great development potential, talent recruitment services, technology cooperation opportunities, etc. are provided. Through precise investment strategy, the investment success rate of office park is improved, and potential enterprises are attracted to settle.

[0093] Further, in step 5, when generating enterprise-oriented investment strategy, the following steps are further included:

[0094] Establish a strategy library, and the strategy classification in the strategy library includes enterprise type strategy, enterprise size strategy and industry characteristic strategy; analyze the prediction results and extract the corresponding strategy from the strategy library to generate the investment strategy for the prediction results.

[0095] Specifically, the strategies for different industry characteristics are as follows:

[0096] Manufacturing enterprises: For manufacturing enterprises with relocation intention, if their production process has high requirements for logistics transportation, the park can provide sites close to transportation hubs (such as ports, railway freight stations, highway entrances) and provide efficient logistics management services, including logistics information platform docking, optimization of cargo handling facilities, etc. If the enterprise has special requirements for production environment (such as temperature and humidity control), the park can invest in the construction of standard factory buildings that meet the conditions and provide related environmental maintenance equipment and technical support.

[0097] Technology research and development enterprises: For technology-based enterprises, especially research and development-intensive enterprises, shared laboratories and research and development platforms can be created, equipped with high-end experimental instruments, and preferential packages or sharing agreements for instrument use can be provided. Establish a technology achievement transformation service center to assist enterprises in commercializing research and development achievements, including patent application, technology transaction matchmaking services, etc. Provide network security protection services to build a high-speed and secure network environment within the park to meet the security needs of enterprise data transmission and storage.

[0098] Financial Services Enterprises: For financial enterprises, the park can set up a financial data center, providing high-speed and stable financial network lines to ensure the real-time and security of financial transactions. Create a financial business district with facilities such as high-end business conference rooms and reception rooms to meet the needs of business negotiations and customer reception. Provide financial regulatory policy consulting services to help enterprises communicate and coordinate with local financial regulatory departments to ensure compliance with the operation of enterprises.

[0099] Strategies for different enterprise sizes are as follows:

[0100] Large Enterprises: For large enterprises, the park can provide whole large areas of land or customized multi-story office buildings to meet their large-scale office and business expansion space needs. Assist enterprises in negotiating tax incentives and special industry support policies with local governments, such as setting up special funds to support major project construction in the park. Provide high-end living support services for senior executives and employees, such as building exclusive executive apartments, employee dormitories, and high-quality schools and hospitals.

[0101] Small and Medium-sized Enterprises: For small and medium-sized enterprises, flexible rental options such as shared office space and small independent office rentals are provided, and rental areas can be adjusted flexibly according to enterprise development. Establish a small and medium-sized enterprise incubation center to provide one-stop entrepreneurship services such as entrepreneurship guidance, financial consulting, legal consulting, and reduce enterprise operating costs. Organize exchange and cooperation activities between small and medium-sized enterprises, such as industry salons and project matching meetings, to promote resource sharing and cooperative development among enterprises.

[0102] Strategies for different industry characteristics are as follows:

[0103] In-depth enterprise research. Communicate face-to-face with enterprises to understand their core business, development strategy, future planning, etc. For example, ask if the enterprise has new product line plans and the requirements of these plans for production facilities and space. Understand the types and quantities of talent the enterprise needs, especially for professional skill talent and high-end management talent recruitment plans. Analyze the financial situation of the enterprise, including revenue structure, cost composition, etc. If the enterprise is sensitive to cost, measures to reduce its operating costs, such as rent discounts and energy cost control, can be considered in the recruitment strategy; if the enterprise is in good financial condition and focuses on quality, high-end park facilities and value-added services can be provided.

[0104] Analyze industry trends and competitive environment. Research the development trend of the industry in which the enterprise is located, such as industry growth rate, technological innovation direction, etc. For enterprises in rapidly developing industries, the park should provide facilities and services that can keep up with the pace of industry development, such as for emerging artificial intelligence industry enterprises, the park should have the ability to support large-scale data operation and related technical talent reserve. Understand the enterprise's competitive position and competitor situation in the industry. If the enterprise hopes to improve its competitiveness by relocation, formulate corresponding strategies based on the advantages of competitors and the differentiated characteristics of the park.

[0105] Combine the park's own resources and advantages. Sort out the park's existing resources, including land, buildings, facilities, service team, etc. If the park has idle land resources, it can provide preferential conditions for land development and construction for enterprises that need to expand production scale; if the park has a professional property service team, it can use this as a selling point for enterprises that pay attention to the quality of park management, and introduce the content and quality guarantee measures of property service in detail. Dig out the advantages of the park, such as geographical location advantage (close to scientific research institutions, business centers, etc.), policy advantage (special park preferential policy, local government support policy). For enterprises that rely on the transformation of scientific research achievements, emphasize the advantage of the park close to scientific research institutions, which can promote industry-university-research cooperation; for policy-sensitive enterprises, interpret the policy preferential and support measures that the park can enjoy, such as tax reduction and exemption, government subsidies, etc.

[0106] The embodiment of the present application can clearly understand the key information of the enterprise, such as the industrial chain relationship, cooperation partner, competitor, development strategy, and the position of the office building park in the industry, the competitive relationship with other parks, etc. through the enterprise knowledge graph, and extract the first site selection feature set. At the same time, through various analysis methods, the features are extracted from multi-modal data, and the multi-modal data covers various forms of data such as text, image, audio, geographic location information, etc., which can reflect the second site selection feature set of the enterprise and the office building park from different angles. Through the two feature sets, a prediction model that can comprehensively consider various complex and diversified enterprise site selection driving factors is constructed, thereby providing a basis for park business attraction.

[0107] Second embodiment:

[0108] As shown in Figure 2 , the embodiment discloses an enterprise site selection driving force analysis system based on multi-modal and knowledge base, comprising: a knowledge graph construction and data integration module 201, which is used for collecting enterprise information, industry information, park information, policy information and constructing a structured knowledge graph;

[0109] a historical data analysis module 202 configured to extract a historical site selection dataset from the knowledge graph with the enterprise development time as the axis, analyze the historical site selection dataset, and obtain key factors affecting site selection, and integrate the key factors to obtain a first site selection feature set;

[0110] a multi-modal feature extraction module 203 configured to collect multi-modal data of the enterprise, extract a development environment dataset from the multi-modal data through text analysis, image analysis, and audio analysis, combine the development environment dataset with the knowledge graph through a data fusion method, and obtain a second site selection feature set;

[0111] a dynamic analysis and prediction module 204 configured to calculate the correlation of the first site selection feature set and the second site selection feature set, and construct an enterprise site selection prediction model through a machine learning algorithm;

[0112] a business solicitation strategy generation module 205 configured to generate a business solicitation strategy for the enterprise according to the prediction result of the enterprise site selection prediction model.

[0113] The application has the following advantages:

[0114] The multi-source multi-modal data and the knowledge graph are fully utilized. The prior art often relies on limited data sources, such as simple questionnaire surveys or fixed databases, and the data dimension is single and slow to update. The present application widely collects multi-modal data, from enterprise internal management information to real-time dynamic data of the surrounding environment of the office building park, and also covers macro information such as policies and regulations and industry trends. For example, not only basic data such as enterprise financial statements and employee numbers are obtained, but also the dynamics of enterprises on social media, job information on recruitment websites, etc. For office building parks, in addition to basic facility data, real-time traffic flow data, commercial facility update information, etc. are also included, providing rich materials for analysis.

[0115] The historical data is deeply mined. The traditional technology does not make full use of the historical data of enterprise site selection, and only records the relocation events. The present application takes the historical time as the axis, deeply analyzes the factors behind the relocation of enterprises, such as enterprise development potential measured by revenue growth and market share change, recruitment situation analyzed from recruitment frequency and job type change, lease expiration rules, market promotion, and business development direction change, etc. Through big data analysis, the key factors affecting site selection decision at different stages are found out, providing a strong basis for prediction.

[0116] The analysis accuracy of data is improved, and multi-modal information features are more accurately extracted. Existing analysis is difficult to handle multi-modal data. The present application uses text analysis to mine information such as development strategy and reputation of enterprises in news and reports; obtains features such as appearance and internal layout of office building parks through image analysis; and evaluates surrounding noise conditions through audio processing. Multi-modal features complement each other to provide a comprehensive perspective for site selection analysis. For example, the expansion plan of an enterprise in the text combined with the crowded situation of the existing office space in the image can more accurately judge the space demand of the enterprise for a new office.

[0117] Advanced prediction models are constructed and optimized. Traditional models mostly use simple statistics or rule-based reasoning, which cannot adapt to complex situations. The present application uses algorithms such as neural networks and random forests to construct models, which can automatically learn complex relationships in data and improve stability. And establish a dynamic optimization mechanism, according to the feedback of investment and new data, real-time update model parameters, ensure the accuracy of prediction and the effectiveness of investment strategy.

[0118] Individualized strategy formulation is realized. Traditional investment strategies are general and lack of pertinence. The present application is based on multi-dimensional data of enterprises to customize investment strategies for different enterprises. For example, provide R&D subsidies and high-speed networks for technology enterprises; provide high-end office facilities and security services for financial enterprises, etc. According to the development stage and demand changes of enterprises, real-time adjustment of strategies, such as providing larger office space selection or cooperation expansion opportunities when the enterprise expands.

[0119] The efficiency and success rate of investment are improved. Traditional investment goals are not clear, and resources are wasted. The present application accurately locates potential enterprises, concentrates resources, and improves investment efficiency. At the same time, individualized strategies meet the needs of enterprises, improve the attractiveness and competitiveness of the park, attract high-quality enterprises to settle down, promote the sustainable development of the park, and realize the win-win of the park and the enterprise.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional modules is taken as an example for description, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0121] The technical features of the above embodiments can be combined arbitrarily, and for the sake of a brief description, not all possible combinations of the technical features in the above embodiments are described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0122] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the application. It should be noted that for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for analyzing enterprise location dynamics based on multimodal and knowledge base methods, characterized in that, include Step 1: Collect enterprise information, industry information, park information, and policy information, and construct a structured knowledge graph; Step 2: Extract historical site selection datasets from the knowledge graph using the enterprise development time as the axis, analyze the historical site selection datasets and obtain key factors affecting site selection, and integrate the key factors to obtain the first site selection feature set; Step 3: Collect multimodal data of enterprises, extract development environment dataset from the multimodal data through text analysis, image analysis and audio analysis respectively, and combine the development environment dataset with the knowledge graph through data fusion method to obtain the second location feature set; Step 4: Based on the first and second location feature sets, construct an enterprise location prediction model using machine learning algorithms; Step 5: Based on the prediction results of the enterprise site selection prediction model, generate an investment promotion strategy for enterprises; In step 2, the historical site selection dataset includes enterprise development information and enterprise address information. When analyzing the historical site selection dataset and obtaining the key factors affecting site selection, the following steps are also included: Step 21: Collect the company's historical relocation records; Step 22: Analyze the reasons for relocation and the time patterns in historical relocation records to uncover the key factors mentioned above; Step 23: Extract development characteristics, recruitment characteristics, leasing characteristics, marketing and business development characteristics from the key factors through data analysis and processing; Step 3 includes the following steps: The multimodal data is obtained through web crawling algorithms or API interfaces, and then processed through data cleaning and data correction. Identify entity data in the multimodal data, and align and map the entity data with the entities in the knowledge graph; The features extracted from the multimodal data are quantized and vectorized, and the multimodal feature vectors are fused with the vector representations of the corresponding entities in the knowledge graph. New relationships are mined and constructed from the entities in the knowledge graph based on information from multimodal data, and the relationships are then updated in the knowledge graph.

2. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Extracting the developmental features includes the following steps: Based on time series, the revenue growth curve of the enterprise is plotted, the relocation time point is marked on the revenue growth curve, the change trend of the enterprise revenue growth curve before and after the relocation time point is analyzed, the change trend of revenue growth before and after the relocation time point is analyzed, and the cities with higher revenue growth trends are summarized and classified. A market share change curve is plotted based on time series data. The relocation time point is marked on the market share change curve, and the trend of market share change before and after the relocation time point is analyzed. Regional competitors before and after the relocation are summarized.

3. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Extracting the recruitment features includes the following steps: A time series plot of the number of hires change curve is generated, and the relocation time point is marked on the curve to analyze the change in the number of hires before and after the relocation time point. This analysis examines the correlation between the region and the job demand structure after the company's relocation.

4. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Extracting the rental features includes the following steps: Based on the company's site selection information before and after the relocation time, we analyze the changes in the company's site selection decisions and site utilization efficiency.

5. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Extracting the aforementioned market promotion and business development characteristics includes the following steps: Analyze the consistency between the company's key marketing regions and development direction and the relocation location.

6. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Step 4 also includes the following steps: Data processing of the first and second location feature sets; Feature selection based on correlation analysis: Calculate the correlation between each feature and the probability of enterprise relocation, select features with high correlation to enter the model, and remove redundant or irrelevant features. The weights of different features are determined using a feature importance assessment method.

7. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 6, characterized in that, The data processing of the first site selection feature set includes the following steps: cleaning and organizing the collected historical site selection decision data, and performing standardization processing; Convert time series data into the input format for machine learning models, and encode categorical data into numerical form; The data processing of the second location feature set includes extracting key information from text data using natural language processing techniques and converting it into feature vectors; and normalizing the key feature vectors extracted from image data so that their numerical range matches that of other features.

8. The enterprise location dynamics analysis method based on multimodal and knowledge base as described in claim 1, characterized in that, Step 5, when generating an investment promotion strategy for enterprises, also includes the following steps: Establish a strategy library, in which strategies are categorized by enterprise type, enterprise size, and industry characteristics. The prediction results are analyzed, and corresponding strategies are extracted from the strategy library to generate an investment promotion strategy based on the prediction results.

9. A system applied to the enterprise location dynamics analysis method based on multimodal and knowledge base as described in any one of claims 1-8, characterized in that, include The knowledge graph construction and data integration module is used to collect enterprise information, industry information, park information, and policy information and construct a structured knowledge graph. The historical data analysis module is used to extract historical site selection datasets from the knowledge graph with the enterprise development time as the axis, analyze the historical site selection datasets and obtain key factors affecting site selection, and integrate the key factors to obtain a first site selection feature set. The multimodal feature extraction module is used to collect multimodal data of enterprises, extract development environment dataset from the multimodal data through text analysis, image analysis and audio analysis, and combine the development environment dataset with the knowledge graph through data fusion method to obtain a second location feature set; The dynamic analysis and prediction module is used to calculate the correlation between the first location feature set and the second location feature set, and to build an enterprise location prediction model through machine learning algorithms. The investment promotion strategy generation module is used to generate investment promotion strategies for enterprises based on the prediction results of the enterprise site selection prediction model.

Images