Multi-source data integrated dynamic enterprise portrait automatic generation method

Abstract

The invention relates to the technical field of enterprise data analysis, in particular to a multi-source data integrated dynamic enterprise portrait automatic generation method. And multi-dimensional data such as enterprise main business, information, product information, research and development investment and patent information are integrated. And accurately screening innovation keywords by using a preset industry innovation dictionary. By analyzing product and patent information containing innovation keywords, a landing coefficient is calculated, and an actual innovation conversion effect is evaluated. And further fusing the innovation keyword landing coefficient and the occurrence frequency of the innovation keywords in the information message data type, and comprehensively evaluating the enterprise science and technology creative ability value in combination with similar characteristics among the keywords. Based on main business text data, enterprises are classified, innovation ability indexes are comprehensively compared, enterprise innovation ability labels are determined, and personalized comprehensive label information is provided, so that enterprise innovation ability conditions are comprehensively reflected, enterprise innovation ability portraits are generated, and accurate information support is provided for enterprise decision makers, investors and policy formulators.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise data analysis, and in particular to a method for automatically generating dynamic enterprise portraits by integrating multi-source data. Background Art

[0002] In today's information-based and data-driven era, the management and analysis of enterprise data has become one of the key factors in improving enterprise competitiveness. Enterprise data comes from multiple channels, including but not limited to internal business systems, market information, patent databases, etc. For market researchers, investors, policymakers, etc. who want to have a comprehensive and in-depth understanding of the status of enterprises, how to efficiently integrate and utilize these multi-source heterogeneous data to build an accurate and insightful portrait of the enterprise's innovation capabilities has become an urgent problem to be solved.

[0003] Existing technologies mainly rely on quantitative data, such as R&D investment costs, patent information data, etc., when analyzing the innovation capabilities of enterprises and constructing a portrait of their innovation capabilities. Although these data can reflect the innovation capabilities of enterprises to a certain extent, they only show part of the results of innovation and cannot provide an in-depth understanding of the actual operation of enterprises in the innovation process. Therefore, if the connection between various data related to innovation is ignored and a portrait of the enterprise's innovation capabilities is only formulated based on quantitative results, it will not be able to fully and effectively reflect the actual innovation status and innovation potential of the enterprise. Summary of the invention

[0004] In order to solve the technical problem that if the connection between various data related to innovation is ignored and the enterprise innovation capability portrait is formulated only based on quantitative results, it will not be able to fully and effectively reflect the real innovation status and innovation potential of the enterprise. The purpose of the present invention is to provide a method for automatically generating a dynamic enterprise portrait based on multi-source data integration. The technical scheme adopted is as follows: Acquire multi-source heterogeneous data of each enterprise, wherein the multi-source heterogeneous data includes main business text data, news information data, product information data, R&D investment cost and patent information data; In the multi-source heterogeneous data of each enterprise, the similarity between the words in the news information data and the words in the preset industry innovation dictionary is analyzed to screen out innovation keywords; for each innovation keyword, various indicators in the product information data and patent information data containing the innovation keyword are analyzed to calculate the landing coefficient of each innovation keyword; In the multi-source heterogeneous data of each enterprise, the landing coefficient of the innovation keyword is integrated with the frequency of occurrence of the innovation keyword in the information data, and the similar characteristics between different innovation keywords in the information data are combined to determine the technological innovation capability value of each enterprise; Enterprises are classified based on the similarity between the main business text data of different enterprises. In each type of enterprise, the technological innovation capability values, product information data, R&D investment costs and patent information data are comprehensively compared between enterprises to determine the innovation capability label of each enterprise and generate an innovation capability portrait of the enterprise.

[0005] Furthermore, the method for acquiring the innovative keywords includes: Use Jieba word segmentation tool to segment each piece of information data to obtain split words; Based on the Word2vec word embedding model, all split words and innovative words in the preset industry innovation dictionary are analyzed to obtain the word vectors of each split word and each innovative word; Calculate the cosine similarity between the word vectors of each split word and each innovative word to obtain the similarity value between the word vectors; For any split word, if the maximum similarity value of the split word is greater than a preset similarity threshold, the split word is used as an innovative keyword.

[0006] Furthermore, the method for obtaining the landing coefficient includes: The product information data includes the product introduction, average daily sales volume and sales amount of each product; The patent information data includes the number of patent applications, the patent text content of each patent and the number of patent downloads; In the multi-source heterogeneous data of each enterprise, for any innovative keyword, the product conversion index of the innovative keyword is calculated based on the frequency of occurrence of the innovative keyword in the product description of the product information data and the average daily sales volume of the products in the product information data containing the innovative keyword; For any innovative keyword, the patent achievement conversion index of the innovative keyword is calculated based on the frequency of occurrence of the innovative keyword in the patent text content of the patent information data and the number of patent downloads of the patent information data containing the innovative keyword; The normalized value of the product transformation index and the mean of the patent achievement transformation index of each innovation keyword is used as the landing coefficient of each innovation keyword.

[0007] Furthermore, the method for obtaining the product conversion index includes: Arrange the product profiles of all products in the product information data according to the time sequence of product release to obtain a sorted sequence; Use Jieba word segmentation tool to segment the product description in each product information data to obtain product vocabulary; In the product description of each product, each innovative keyword is compared with the product vocabulary, so as to count the number of occurrences of each innovative keyword in each product description as a quantity factor; For any innovative keyword, the sum of the quantity factors of the innovative keyword in all product profiles is used as the first product conversion factor; In the sorting sequence, starting from the second product introduction, the difference between the quantity factor of the innovative keyword in each product introduction and the quantity factor in the adjacent previous product introduction is used as the difference factor, and the sum of all the difference factors corresponding to the sorting sequence is normalized to a value that is used as the second product conversion factor; In the product introductions of all products, the sum of the average daily sales of the products corresponding to the product introductions containing the innovative keyword is used as the third product conversion factor corresponding to the innovative keyword; The normalized value of the sum of the first product conversion factor, the second product conversion factor and the third product conversion factor corresponding to each innovative keyword is used as the product conversion indicator corresponding to each innovative keyword.

[0008] Furthermore, the method for obtaining the patent achievement transformation index includes: Jieba word segmentation tool is used to segment the patent text content in each patent information data to obtain patent vocabulary; In the patent text content of each patent, each innovative keyword is compared with the patent vocabulary, so as to count the number of occurrences of each innovative keyword in all patent text contents as the first patent achievement conversion factor; For any innovation keyword, the sum of the patent downloads corresponding to all patent text contents containing the innovation keyword is used as the second patent achievement conversion factor corresponding to the innovation keyword; The normalized value of the sum of the first patent achievement conversion factor and the second patent achievement conversion factor corresponding to the innovation keyword is used as the patent achievement conversion index corresponding to the innovation keyword.

[0009] Furthermore, the method for obtaining the technological innovation capability value includes: In each piece of information data, analyze the similar characteristics between innovative keywords and determine the limiting factors of each piece of information data; For the innovative keywords in each piece of information data, the number of occurrences of the innovative keywords in each piece of information data is weighted using the landing coefficient of the innovative keywords, thereby obtaining the technological innovation factor of each piece of information data; According to the technological innovation factor and the limitation factor of each piece of information data, the technological innovation coefficient of each piece of information data is determined, and the technological innovation coefficient is positively correlated with the technological innovation factor, and the technological innovation coefficient is negatively correlated with the limitation factor; The normalized value of the mean of the technological innovation coefficient of all information data of each enterprise is used as the technological innovation capability value of each enterprise.

[0010] Furthermore, the method for obtaining the limitation factor includes: Among all the innovative keywords in each piece of information data, any two innovative keywords are combined to obtain all non-repeating keyword combinations; In each keyword combination, the cosine similarity between the word vectors of two innovative keywords is calculated as the similarity coefficient; The cumulative sum of the similarity coefficients of all keyword combinations is used as the limiting factor for each piece of information data.

[0011] Furthermore, the classification of enterprises based on the similarity between the main business text data of different enterprises includes: For the main business text data of each enterprise, use Jieba word segmentation tool to process it and obtain the word segmentation set corresponding to the main business text data of each enterprise; For any two companies, calculate the Jaccard similarity coefficient between the word sets of these two companies; The optimal K value is preset, and all enterprises are clustered based on the optimal K value and the K-means clustering algorithm to obtain clusters; wherein the distance metric is the value after the negative correlation mapping of the Jaccard similarity coefficient between two enterprises.

[0012] Furthermore, the method for obtaining the innovation capability label includes: For any enterprise, the number of patent applications, sales and value of all products, R&D investment costs, and technological innovation capability of the enterprise are used as innovation indicator data; In each type of enterprise, the entropy weight method is used to calculate the weight factor of each innovation indicator data of all enterprises; For any enterprise, the value of the innovation index data is weighted and averaged using the weight factor of the enterprise's innovation index data to obtain the innovation capability value of the enterprise; In each type of enterprise, the enterprises are sorted in ascending order according to their innovation capability values ​​to obtain a ranking sequence, and the sequence number value of each enterprise in the ranking sequence is used as the innovation capability label of each enterprise.

[0013] Furthermore, the method for obtaining the enterprise innovation capability portrait includes: Generate corresponding enterprise innovation capability portraits based on visualization tools and the enterprise's innovation capability labels.

[0014] The present invention has the following beneficial effects: First, the main business text data, information data, product information data, R&D investment cost and patent information data of each enterprise are obtained, and the multi-dimensional information of the enterprise is fully integrated to more accurately reflect the real situation of the enterprise, and provide a rich and comprehensive data foundation for the construction of the enterprise innovation capability portrait. Using the preset industry innovation dictionary, the similarity between the vocabulary and the vocabulary in the information information data is analyzed, and the innovation keywords are accurately screened out. This step effectively improves the accuracy of innovation identification. Then, by analyzing the various indicators in the product information data and patent information data where the innovation keywords exist, the landing coefficient of each innovation keyword is calculated. This process evaluates the actual transformation effect of innovation and helps to identify the actual value and market potential of enterprise innovation. Further, in order to more comprehensively quantify the innovation ability of the enterprise, the present invention integrates the landing coefficient of the innovation keyword with the frequency of occurrence of the innovation keyword in the information information data, and combines the similar characteristics between different innovation keywords to comprehensively evaluate the value of the enterprise's scientific and technological innovation ability. This step comprehensively considers the quantity, quality and transformation effect of innovation, and provides a scientific basis for the quantitative evaluation of the enterprise's innovation ability. Finally, the enterprises are classified based on the similarity of the main business text data, and the various innovation capability indicators between enterprises in each category are comprehensively compared to determine the innovation capability label of the enterprise. This process provides personalized and comprehensive label information for the generation of enterprise innovation capability portraits. Combined with the above steps, the present invention can dynamically generate an enterprise's innovation capability portrait, which not only fully reflects the enterprise's innovation capability status, but also can be adjusted in real time as the enterprise data is updated, providing timely and accurate information support for enterprise decision makers, investors and policy makers. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A method flow chart of a method for automatically generating a dynamic enterprise portrait by integrating multi-source data provided by an embodiment of the present invention; Figure 2 A method flow chart of a method for obtaining a landing coefficient provided by an embodiment of the present invention; Figure 3A method flow chart of a method for obtaining technological innovation capability value provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a method for automatically generating a dynamic enterprise portrait based on multi-source data integration proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The specific scheme of the method for automatically generating a dynamic enterprise portrait by integrating multi-source data provided by the present invention is described in detail below with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a method flow chart of a method for automatically generating a dynamic enterprise portrait by integrating multi-source data provided by an embodiment of the present invention, the method comprising the following steps: Step S1: Obtain multi-source heterogeneous data of each enterprise, where the multi-source heterogeneous data includes main business text data, news information data, product information data, R&D investment costs, and patent information data.

[0021] Corporate portraits can help investors evaluate a company's innovation potential and investment value. For example, when screening investment projects, venture capital institutions can select companies with high innovation potential for investment based on corporate innovation capability portraits. In today's era of information explosion, the sources of corporate data are increasingly diverse. For market researchers, investors, policymakers, etc. who want to have a comprehensive and in-depth understanding of corporate conditions, it is necessary to efficiently integrate and utilize these multi-source heterogeneous data to build an accurate and insightful corporate innovation capability portrait.

[0022] Therefore, it is necessary to comprehensively collect and integrate the multi-source heterogeneous data of each enterprise to build a rich and comprehensive data foundation to provide strong support for the subsequent generation of enterprise innovation capability portraits. Specifically, the website data of each enterprise in the past five years is obtained by crawlers, and the multi-source heterogeneous data required in the embodiment of the present invention is extracted from it, which includes the main business text data, information information data, product information data, R&D investment cost and patent information data of each enterprise. Among them, the main business text data mainly describes the key texts such as the business field and market positioning of the enterprise, which helps to understand the core business and development direction of the enterprise; the information information data mainly describes the forums and news text information of the enterprise, which helps to reflect the performance of the enterprise in the market and is an important source for identifying the innovation dynamics of the enterprise; the product information data mainly includes the products produced by the enterprise, the product introduction of each product, the average daily sales volume of the product and the product sales, which can reflect the product innovation ability of the enterprise; the patent information data mainly includes the number of patent applications of the enterprise, the patent text content of each patent and the number of patent downloads, which reflects the technological innovation ability of the enterprise; the R&D investment cost can measure the investment intensity of the enterprise innovation.

[0023] In the embodiments of the present invention, the collection and acquisition of various information data of the enterprise are all authorized by the relevant users, and the process does not violate relevant laws and regulations and does not violate public order and good customs.

[0024] It should be noted that the number of enterprises analyzed in the embodiment of the present invention is set to 50, and the specific number can be adjusted according to the implementation scenario and is not limited here.

[0025] Step S2: In the multi-source heterogeneous data of each enterprise, analyze the similarity between the words in the news information data and the words in the preset industry innovation dictionary, so as to screen out innovative keywords; for each innovative keyword, analyze the various indicators in the product information data and patent information data containing the innovative keyword, and calculate the landing coefficient of each innovative keyword.

[0026] The information data of enterprises may involve more cutting-edge technologies and innovative concepts, which can reflect the innovative elements of enterprises, show the technical reserves and innovation direction of enterprises. Therefore, by analyzing the similarity of the words in each piece of information data of each enterprise with the words in the preset industry innovation dictionary, the innovative keywords in the information data can be screened out. The innovative keywords represent the innovative attempts or breakthroughs of enterprises in their fields, which can reflect the information elements of technological innovation of enterprises. However, in order to create a good image, enterprises may deliberately pile up popular innovative keywords in information data, such as "big data" and "blockchain", but they have not actually applied these technologies to innovation activities. For example, a traditional manufacturing company mentioned "big data" many times in its report, but simply described it as a possible future development direction, and there were no actual big data application projects or innovative results. Therefore, some innovation keywords will mislead the judgment of the company's innovation ability; and the development of patents and new products is a direct reflection of innovation results, and also reflects the conversion ability of the innovation keywords mentioned in the company's information data. Therefore, the innovation keywords are linked to the product information data and patent information data to calculate the landing coefficient of each innovation keyword. The landing coefficient can be used as a weight to measure the company's technological innovation ability in the subsequent process, which helps to more accurately measure the company's technological innovation ability.

[0027] First, for each piece of information data of each enterprise, the similarity between the words therein and the words in the preset industry innovation dictionary is analyzed to screen out innovation keywords.

[0028] Preferably, in one embodiment of the present invention, the method for acquiring innovative keywords includes:

[0029] Jieba word segmentation tool is used to segment each piece of information data to obtain split words. Through this process, long text content can be split into separate words, which helps to understand and analyze the text content of information data more accurately.

[0030] Based on the Word2vec word embedding model, all split words and innovative words in the preset industry innovation dictionary are analyzed to obtain the word vector of each split word and each innovative word. The Word2vec word embedding model can map words to a high-dimensional vector space, so that semantically similar words are closer in the vector space. This method can capture the semantic relationship between words, making the subsequent similar situation analysis results more accurate.

[0031] Then calculate the cosine similarity between the word vectors of each split word and each innovative word to get the similarity value between the word vectors. The value range of cosine similarity is between -1 and 1, and the closer it is to 1, the closer the two word vectors are, and the higher the similarity between the split word and the innovative word will be.

[0032] Based on the above process, for any split word, there is a similarity value between the split word and each innovative word, so if the maximum similarity value of the split word is greater than the preset similarity threshold, the split word is considered to be an innovative keyword. Conversely, if the maximum similarity value of the split word is still less than or equal to the preset similarity threshold, the split word is considered not to be an innovative keyword.

[0033] It should be noted that the Jieba word segmentation tool and the Word2vec word embedding model are both well-known technologies, and the specific process will not be elaborated here; the preset similarity threshold is set to 0.5, and the specific value can be adjusted according to the implementation scenario and is not limited here; the method for obtaining the preset industry innovation dictionary can be formulated by capturing professional terms in authoritative industry materials, academic literature, and technical reports. For example, in the field of science and technology, the innovative words in the industry innovation dictionary can be: artificial intelligence, big data analysis, blockchain technology, 5G communication, etc. In the manufacturing field, the innovative words in the industry innovation dictionary can be: intelligent manufacturing, Industry 4.0, 3D printing technology, etc.

[0034] So far, for any enterprise, by processing each piece of information data of the enterprise, the innovation keywords can be screened out. In view of the fact that some enterprises may use innovation keywords in their information data in order to create a good image, but have not actually implemented relevant technologies or projects, this behavior may lead to the outside world's misjudgment of the enterprise's innovation ability; and the development of patents and new products is a direct reflection of the enterprise's innovation achievements, and is also an important indicator for measuring the enterprise's innovation ability. Therefore, in the embodiment of the present invention, in order to measure the actual application degree of each enterprise's innovation keywords, the various indicators in the product information data and patent information data with innovation keywords are analyzed, and the concept of landing coefficient is introduced to characterize the correlation between the innovation keywords mentioned in the enterprise's information data and the actual innovation ability of the enterprise.

[0035] Preferably, in one embodiment of the present invention, the method for obtaining the landing coefficient includes: See also Figure 2 , which shows a method flow chart of a method for obtaining a landing coefficient in one embodiment of the present invention, the method comprising the following steps: Step S201: Acquire various indicators in product information data and patent information data of each enterprise.

[0036] Based on the data acquisition process in step S1, it can be seen that the product information data includes the product introduction of each product, the average daily sales volume of the product, and the product sales; the patent information data includes the number of patent applications, the patent text content of each patent, and the number of patent downloads.

[0037] Step S202: For each innovative keyword, the product conversion index of each innovative keyword is calculated according to the appearance frequency of the innovative keyword in the product introduction of the product information data and the average daily sales volume of the products in the product information data containing the innovative keyword.

[0038] The frequency of appearance of innovative keywords in product descriptions reflects the degree of market attention and acceptance. Combined with the average daily sales of the product, it is possible to further evaluate whether the innovative keywords have truly been transformed into market demand.

[0039] Therefore, in the multi-source heterogeneous data of each enterprise, the Jieba word segmentation tool is first used to segment the product description in each product information data to obtain the product vocabulary.

[0040] It should be noted that the Jieba word segmentation tool obtaining a single word is a well-known technology, and the specific process will not be described here.

[0041] In the product introduction of each product, each innovative keyword is compared with the product vocabulary to count the number of times each innovative keyword appears in each product introduction as a quantity factor. The larger the quantity factor, the higher the attention paid to the innovative keyword in the product introduction, and the higher the conversion rate.

[0042] Therefore, for any innovative keyword, the sum of the quantity factors of the innovative keyword in all product introductions is used as the first product conversion factor. The larger the first product conversion factor is, the more extensive the application of the innovative keyword in the product, and thus the higher the product conversion effect.

[0043] As products are continuously released, if the frequency of an innovative keyword appearing in the product introduction gradually increases, that is, when the quantity factor becomes larger and larger, it can be considered that the correlation between the innovative keyword and the actual R&D capability of the enterprise is stronger, that is, the enterprise's product conversion capability is stronger.

[0044] Therefore, the product profiles of all products in the product information data are arranged according to the chronological order of product release to obtain a sorted sequence.

[0045] Then, in the sorting sequence, starting from the second product introduction, the difference between the quantity factor of each innovative keyword in each product introduction and the adjacent previous product introduction is used as the difference factor. Based on the above analysis, the difference factor is a positive number, and the larger it is, the higher the company's technology transformation capability is. So far, in the sorting sequence, there is a difference factor corresponding to each two adjacent product introductions. The sum of all the difference factors corresponding to the sorting sequence can be normalized as the second product transformation factor. The larger the second product transformation factor is, the higher the company's product transformation effect is and the stronger its technology transformation capability is. Since the difference factor may be positive or negative, the normalization method here can be used. .

[0046] Given that product sales can directly reflect the degree of market acceptance after technology is transformed into products, for any innovative keyword, in the product introductions of all products, the sum of the average daily sales of products corresponding to the product introductions containing the innovative keyword will be used as the third product conversion factor corresponding to the innovative keyword. The larger the third product conversion factor, the better the sales of products containing the innovative keyword, that is, the better the technology conversion effect.

[0047] So far, in each enterprise, for each innovative keyword extracted, it has three indicators that characterize product conversion capabilities, and these three indicators can be integrated: the sum of the first product conversion factor, the second product conversion factor, and the third product conversion factor corresponding to each innovative keyword is normalized as the product conversion index corresponding to each innovative keyword. The larger the product conversion index, the stronger the actual innovation capability of the enterprise under the technology corresponding to the innovative keyword. Normalization is a technical means well known to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0048] Step S203: For each innovative keyword, the patent achievement conversion index of each innovative keyword is calculated according to the frequency of occurrence of the innovative keyword in the patent text content of the patent information data and the patent download volume of the patent information data containing the innovative keyword.

[0049] The frequency of occurrence of innovative keywords in patent text content reflects the activity and innovation of its technological research, while the number of patent downloads reflects the market's attention to technological innovation and its practical application potential.

[0050] Therefore, in the multi-source heterogeneous data of each enterprise, the Jieba word segmentation tool is first used to segment the patent text content in each patent information data to obtain patent vocabulary. It should be noted that the Jieba word segmentation tool is a well-known technology for obtaining a single vocabulary, and the specific process is not repeated here.

[0051] Then, in the patent text content of each patent, each innovative keyword is compared with the patent vocabulary, so as to count the number of occurrences of each innovative keyword in all patent text contents as the first patent achievement conversion factor. The larger the first patent achievement conversion factor, the more times the innovative keyword is mentioned in the patent text content, which is considered to be a higher level of attention and a better patent achievement conversion effect.

[0052] Then, for any innovative keyword, the sum of the patent download volumes corresponding to all patent text contents containing the innovative keyword is used as the second patent achievement conversion factor corresponding to the innovative keyword. The larger the second patent achievement conversion factor is, the more attractive the technology in the patent document mentioning the innovative keyword is to the market, which is also regarded as a better patent achievement conversion effect.

[0053] Finally, the sum of the first patent achievement conversion factor and the second patent achievement conversion factor corresponding to the innovation keyword is normalized to the value obtained as the patent achievement conversion index corresponding to the innovation keyword. The larger the patent achievement conversion index, the better the effect of the innovation keyword in the enterprise's patent achievement conversion, indicating that the enterprise has a stronger innovation ability under the technology corresponding to the innovation keyword. Normalization is a technical means well known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0054] Step S204: The product transformation index corresponding to each innovation keyword is integrated with the patent achievement transformation index to obtain the landing coefficient corresponding to each innovation keyword.

[0055] Based on the calculation process in the above steps, the product transformation index and patent achievement transformation index corresponding to each innovative keyword can be obtained, and the larger the product transformation index corresponding to a certain innovative keyword, the stronger the actual innovation ability of the enterprise under the technology corresponding to the innovative keyword; similarly, the larger the patent achievement transformation index corresponding to a certain innovative keyword, the stronger the innovation ability of the enterprise under the technology corresponding to the innovative keyword. Therefore, for any innovative keyword, the value after normalizing the mean of the product transformation index and the patent achievement transformation index of the innovative keyword is used as the landing coefficient of the innovative keyword. The larger the landing coefficient of the innovative keyword, the stronger the technology transformation ability of the enterprise under the innovative keyword. Normalization is a technical means well known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0056] Step S3: In the multi-source heterogeneous data of each enterprise, the landing coefficient of the innovation keyword is integrated with the frequency of occurrence of the innovation keyword in the information data, and the similar characteristics between different innovation keywords in the information data are combined to determine the technological innovation capability value of each enterprise.

[0057] In step S2, the landing coefficient of each innovative keyword contained in the information data of each enterprise can be calculated, which analyzes the innovation transformation capability of the enterprise. In this step, the landing coefficient of each innovative keyword can be used as a weight to perform weighted fusion on the frequency of occurrence of the enterprise's innovative keywords in the information data. In addition, in the information data, since different innovative keywords may have different or similar technical connotations or application scenarios, new indicator data for measuring the innovation capability of each enterprise is constructed in combination with the similar characteristics of different keywords in the information data, that is, the technological innovation capability value. This indicator integrates the relationship between multiple data, so as to more accurately reflect the technological innovation capability of the enterprise.

[0058] Preferably, in one embodiment of the present invention, the method for obtaining the technological innovation capability value includes: See also Figure 3 , which shows a method flow chart of a method for obtaining a technological innovation capability value in one embodiment of the present invention, the method comprising the following steps: Step S301: In each piece of information data of each enterprise, similar features between innovative keywords are analyzed to determine the limiting factors of each piece of information data.

[0059] Among all the innovative keywords in each piece of news information data, any two innovative keywords are combined to obtain all non-repeating keyword combinations.

[0060] Then, in each keyword combination, the cosine similarity between the word vectors of the two innovative keywords is calculated as the similarity coefficient. The larger the similarity coefficient, the closer the word vectors of the two innovative keywords are, and the higher the similarity between the two innovative keywords. It should be noted that the method for obtaining the word vector here is recorded in step S2 and will not be repeated here.

[0061] Finally, in each piece of information data, the cumulative sum of the similarity coefficients of all keyword combinations is used as the limitation factor of each piece of information data. The larger the limitation factor, the higher the similarity of the innovative keywords in the information, which means that the scope covered by the company's information data is smaller, which means that the company's innovation ability is relatively weak; conversely, the smaller the limitation factor, the larger the scope covered by the company's information data, which means that the company's innovation ability is stronger.

[0062] Step S302: The landing coefficient of each innovative keyword is integrated with the number of occurrences in each piece of information data to obtain the technological innovation factor of each piece of information data.

[0063] The landing coefficient of innovative keywords can reflect the technology transformation capability of enterprises under the innovative keywords, that is, the actual value of innovative keywords; and the number of occurrences of innovative keywords in each piece of information data represents the scientific research situation of enterprises in the corresponding fields of innovative keywords, etc. Therefore, the landing coefficient of innovative keywords is used to weight the number of occurrences of innovative keywords in each piece of information data. Specifically, for the innovative keywords in each piece of consulting information data, the product of the landing coefficient of each innovative keyword and the number of occurrences of each innovative keyword in each piece of information data is used as the weighting factor. The larger the weighting factor, the stronger the scientific research capability of the enterprise in the corresponding fields of innovative keywords. Subsequently, the sum of the weighting factors corresponding to all innovative keywords in each piece of information data is used as the technological innovation factor of each piece of information data. The larger the technological innovation factor of a piece of information data, the stronger the technological innovation capability of the enterprise reflected by this piece of information data.

[0064] It should be noted that when calculating the technological innovation factor, the same innovation keyword in each piece of information data is regarded as one innovation keyword.

[0065] Step S303: Among all the information data of each enterprise, the limitation factors and technological innovation factors of the information data are integrated to obtain the technological innovation capability value of each enterprise.

[0066] The technological innovation factor and limitation factor of each piece of information data are combined to determine the technological innovation coefficient of each piece of information data. Based on the above analysis, it can be seen that the smaller the limitation factor, the larger the scope covered by the enterprise's information data, indicating that the enterprise has stronger innovation capabilities; the larger the technological innovation factor, the stronger the technological innovation capability of the enterprise reflected by this piece of information data. Therefore, the technological innovation coefficient is positively correlated with the technological innovation factor, and the technological innovation coefficient is negatively correlated with the limitation factor. The formula model of the technological innovation coefficient includes: in, Indicates the technological innovation coefficient of each piece of information data of each enterprise; Represents the technological innovation factor of each piece of information data of each enterprise; Indicates the limitation factor of each piece of information data of each enterprise; Indicates the preset first parameter; Represents the normalization function.

[0067] In the formula model of the technological innovation coefficient, based on the calculation in step S301, it can be known that the value of the limitation factor may be positive or negative, so in this embodiment of the present invention, the limitation factor is used. The function normalizes the limitation factors of each piece of information data so that its value after processing is between 0 and 1. The closer it is to 0, the larger the scope covered by the information data of the enterprise, and the stronger the innovation ability of the enterprise. In order to realize that the technological innovation coefficient is positively correlated with the technological innovation factor, and the technological innovation coefficient is negatively correlated with the limitation factor, the above formula model is constructed to obtain the technological innovation coefficient. At this time, the larger the technological innovation coefficient, the higher the technological innovation ability of the enterprise shown by a piece of information data.

[0068] Finally, the average value of the technological innovation coefficient of all information data of each enterprise is normalized and used as the technological innovation capability value of each enterprise. Normalization is a technical means well known to those skilled in the art, and the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0069] It should be noted that the first parameter is preset The function of is to prevent the denominator from being 0. The value can be 0.001. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0070] Step S4: Classify enterprises based on the similarity between the main business text data of different enterprises; in each type of enterprise, comprehensively compare the technological innovation capability values, product information data, R&D investment costs and patent information data between enterprises, so as to determine the innovation capability label of each enterprise and generate an enterprise innovation capability portrait.

[0071] Based on the above steps, by analyzing and processing the multi-source heterogeneous data of each enterprise in recent years, the scientific and technological innovation capability value of each enterprise in recent years can be determined to reflect the scientific and technological innovation capability of the enterprise and provide a reference for investors, etc.; however, since the driving factors of innovation in various industries are different, when constructing a portrait of the innovation capability of an enterprise, the embodiment of the present invention may classify enterprises in the same field into one category of enterprises, and then comprehensively compare the scientific and technological innovation capability values, product information data, R&D investment costs, and patent information data between enterprises in each category, so as to determine the innovation capability label of each enterprise and generate a portrait of the innovation capability of the enterprise.

[0072] First, enterprises need to be clustered. The main business text data of an enterprise mainly describes key texts such as the enterprise's business field and market positioning. Therefore, in order to classify enterprises that may be in the same field into one category, the similarity between the main business text data of different enterprises can be analyzed to classify the enterprises.

[0073] Preferably, in one embodiment of the present invention, enterprises are classified based on the similarity between the main business text data of different enterprises, including: It is difficult to directly compare the similarities between text data, so for the main business text data of each enterprise, we first use the Jieba word segmentation tool to process it, so as to obtain the word segmentation set corresponding to the main business text data of each enterprise, so that in the subsequent process, we can compare the field similarities between enterprises at the lexical level.

[0074] The Jaccard similarity coefficient is a commonly used set similarity measurement method. In the context of text data, it can reflect the proportion of common words in the word sets corresponding to the main business text data of two companies, thereby measuring the similarity between the two companies. Therefore, for any two companies, the Jaccard similarity coefficient between the word sets of the two companies is calculated. The larger the Jaccard similarity coefficient, the higher the similarity between the two companies.

[0075] Finally, the optimal K value is preset to obtain the clustering clusters of the specified data. All enterprises are clustered based on the optimal K value and K-means clustering algorithm to obtain clusters. Among them, the distance metric is the value after the negative correlation mapping of the Jaccard similarity coefficient between two enterprises. The negative correlation mapping here can be adopted by the formula ,in, represents an exponential function with the natural constant e as the base, and x represents the independent variable. At this point, the enterprises in each cluster have similar main businesses, which helps to further compare the innovation capabilities of enterprises with similar business areas, making the subsequent enterprise innovation capability portrait more accurate.

[0076] It should be noted that the use of Jieba word segmentation tool, the calculation method of Jaccard similarity coefficient and K-means clustering algorithm are all well-known technologies, and the specific process will not be repeated here; the preset optimal K value is set to 5, and the specific value can be adjusted according to the implementation scenario and is not limited here.

[0077] After clustering the enterprises, for each type of enterprise, we can integrate various data representing the enterprise's innovation situation to determine a more comprehensive indicator that can fully characterize the enterprise's innovation capability, and then formulate an innovation capability label for each enterprise in each type of enterprise.

[0078] Preferably, in one embodiment of the present invention, the method for obtaining the innovation capability label includes: Innovation capability is a multi-dimensional concept that should involve many aspects of an enterprise. Therefore, for any enterprise, the number of patent applications, sales and value of all products, R&D investment costs, and technological innovation capability values ​​of the enterprise are used as innovation indicator data. The sales and value of all products can be obtained by adding up the sales of all types of products.

[0079] Since different innovation indicator data may have different importance in evaluating the innovation capability of an enterprise, the entropy weight method is used to calculate the weight factor of each innovation indicator data of all enterprises in each type of enterprise. The entropy weight method calculates the weight based on the concept of information entropy. Therefore, the larger the weight factor of a certain innovation indicator data, the higher the information content of the innovation indicator data and the greater its influence.

[0080] Then, for any enterprise, the weight factor of the enterprise's innovation index data is used to weighted average the values ​​of the innovation index data to obtain the enterprise's innovation capability value. At this time, the larger the innovation capability value of an enterprise, the stronger the comprehensive innovation capability of the enterprise.

[0081] Finally, in each type of enterprise, the enterprises are sorted in ascending order according to their innovation capability values ​​to obtain a ranking sequence. The serial number value of each enterprise in the ranking sequence is used as the innovation capability label of each enterprise. The larger the value of the innovation capability label, the stronger the comprehensive innovation capability of the enterprise.

[0082] Finally, for each type of enterprise, a portrait of the enterprise’s innovation capability is generated based on each enterprise’s innovation capability label.

[0083] Preferably, in one embodiment of the present invention, the method for obtaining the enterprise innovation capability portrait includes: Generate corresponding enterprise innovation capability portraits based on visualization tools and the enterprise's innovation capability labels.

[0084] It should be noted that visualization tools can adopt the structure of knowledge graphs or use Tableau to create dashboards for intuitive display. The choice of visualization tools is not limited here.

[0085] In order to facilitate calculation, all indicator data involved in the calculation in the embodiment of the present invention are preprocessed to eliminate the dimension effect. The specific means of eliminating the dimension effect are technical means well known to those skilled in the art and are not limited here.

[0086] In summary, firstly, the main business text data, information data, product information data, R&D investment cost and patent information data of each enterprise are obtained, and the multi-dimensional information of the enterprise is fully integrated, so as to more accurately reflect the real situation of the enterprise and provide a rich and comprehensive data basis for the construction of the enterprise innovation capability portrait. The similarity of the vocabulary in the preset industry innovation dictionary analysis and information data is used to accurately screen out the innovation keywords. This step effectively improves the accuracy of innovation identification. Then, by analyzing the various indicators in the product information data and patent information data with innovation keywords, the landing coefficient of each innovation keyword is calculated. This process evaluates the actual transformation effect of innovation and helps to identify the actual value and market potential of enterprise innovation. Further, in order to more comprehensively quantify the innovation ability of the enterprise, the embodiment of the present invention integrates the landing coefficient of the innovation keyword with the frequency of occurrence of the innovation keyword in the information data, and combines the similar characteristics between different innovation keywords to comprehensively evaluate the value of the enterprise's scientific and technological innovation ability. This step comprehensively considers the quantity, quality and transformation effect of innovation, and provides a scientific basis for the quantitative evaluation of the enterprise's innovation ability. Finally, the enterprises are classified based on the similarity of the main business text data, and the various innovation capability indicators between enterprises in each category are comprehensively compared to determine the innovation capability label of the enterprise. This process provides personalized and comprehensive label information for the generation of enterprise innovation capability portraits. Combined with the above steps, the embodiment of the present invention can dynamically generate an innovation capability portrait of an enterprise. This portrait not only fully reflects the innovation capability status of the enterprise, but also can be adjusted in real time as the enterprise data is updated, providing timely and accurate information support for enterprise decision makers, investors and policy makers.

[0087] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0088] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for automatically generating dynamic enterprise portraits based on multi-source data integration, characterized in that: The method comprises: Acquire multi-source heterogeneous data of each enterprise, wherein the multi-source heterogeneous data includes main business text data, news information data, product information data, R&D investment cost and patent information data; In the multi-source heterogeneous data of each enterprise, the similarity between the words in the news information data and the words in the preset industry innovation dictionary is analyzed to screen out innovation keywords; for each innovation keyword, various indicators in the product information data and patent information data containing the innovation keyword are analyzed to calculate the landing coefficient of each innovation keyword; In the multi-source heterogeneous data of each enterprise, the landing coefficient of the innovation keyword is integrated with the frequency of occurrence of the innovation keyword in the information data, and the similar characteristics between different innovation keywords in the information data are combined to determine the technological innovation capability value of each enterprise; Enterprises are classified based on the similarity between the main business text data of different enterprises. In each type of enterprise, the technological innovation capability values, product information data, R&D investment costs and patent information data are comprehensively compared between enterprises to determine the innovation capability label of each enterprise and generate an innovation capability portrait of the enterprise.

2. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 1, characterized in that: The method for obtaining the innovative keywords includes: Use Jieba word segmentation tool to segment each piece of information data to obtain split words; Based on the Word2vec word embedding model, all split words and innovative words in the preset industry innovation dictionary are analyzed to obtain the word vectors of each split word and each innovative word; Calculate the cosine similarity between the word vectors of each split word and each innovative word to obtain the similarity value between the word vectors; For any split word, if the maximum similarity value of the split word is greater than a preset similarity threshold, the split word is used as an innovative keyword.

3. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 1 is characterized in that: The method for obtaining the landing coefficient includes: The product information data includes the product introduction, average daily sales volume and sales amount of each product; The patent information data includes the number of patent applications, the patent text content of each patent and the number of patent downloads; In the multi-source heterogeneous data of each enterprise, for any innovative keyword, the product conversion index of the innovative keyword is calculated based on the frequency of occurrence of the innovative keyword in the product description of the product information data and the average daily sales volume of the products in the product information data containing the innovative keyword; For any innovative keyword, the patent achievement conversion index of the innovative keyword is calculated based on the frequency of occurrence of the innovative keyword in the patent text content of the patent information data and the number of patent downloads of the patent information data containing the innovative keyword; The normalized value of the product transformation index and the mean of the patent achievement transformation index of each innovation keyword is used as the landing coefficient of each innovation keyword.

4. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 3 is characterized in that: The method for obtaining the product conversion index includes: Arrange the product profiles of all products in the product information data according to the time sequence of product release to obtain a sorted sequence; Use Jieba word segmentation tool to segment the product description in each product information data to obtain product vocabulary; In the product description of each product, each innovative keyword is compared with the product vocabulary, so as to count the number of occurrences of each innovative keyword in each product description as a quantity factor; For any innovative keyword, the sum of the quantity factors of the innovative keyword in all product profiles is used as the first product conversion factor; In the sorting sequence, starting from the second product introduction, the difference between the quantity factor of the innovative keyword in each product introduction and the quantity factor in the adjacent previous product introduction is used as the difference factor, and the sum of all the difference factors corresponding to the sorting sequence is normalized to a value that is used as the second product conversion factor; In the product introductions of all products, the sum of the average daily sales of the products corresponding to the product introductions containing the innovative keyword is used as the third product conversion factor corresponding to the innovative keyword; The normalized sum of the first product conversion factor, the second product conversion factor, and the third product conversion factor corresponding to each innovative keyword is used as the product conversion indicator corresponding to each innovative keyword.

5. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 3 is characterized in that: The method for obtaining the patent achievement transformation indicator includes: Jieba word segmentation tool is used to segment the patent text content in each patent information data to obtain patent vocabulary; In the patent text content of each patent, each innovative keyword is compared with the patent vocabulary, so as to count the number of occurrences of each innovative keyword in all patent text contents as the first patent achievement conversion factor; For any innovation keyword, the sum of the patent downloads corresponding to all patent text contents containing the innovation keyword is used as the second patent achievement conversion factor corresponding to the innovation keyword; The normalized value of the sum of the first patent achievement conversion factor and the second patent achievement conversion factor corresponding to the innovation keyword is used as the patent achievement conversion index corresponding to the innovation keyword.

6. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 2 is characterized in that: The method for obtaining the technological innovation capability value includes: In each piece of information data, analyze the similar characteristics between innovative keywords and determine the limiting factors of each piece of information data; For the innovative keywords in each piece of information data, the number of occurrences of the innovative keywords in each piece of information data is weighted using the landing coefficient of the innovative keywords, thereby obtaining the technological innovation factor of each piece of information data; According to the technological innovation factor and the limitation factor of each piece of information data, the technological innovation coefficient of each piece of information data is determined, and the technological innovation coefficient is positively correlated with the technological innovation factor, and the technological innovation coefficient is negatively correlated with the limitation factor; The normalized value of the mean of the technological innovation coefficient of all information data of each enterprise is used as the technological innovation capability value of each enterprise.

7. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 6 is characterized in that: The method for obtaining the limitation factor includes: Among all the innovative keywords in each piece of information data, any two innovative keywords are combined to obtain all non-repeating keyword combinations; In each keyword combination, the cosine similarity between the word vectors of two innovative keywords is calculated as the similarity coefficient; The cumulative sum of the similarity coefficients of all keyword combinations is used as the limiting factor for each piece of information data.

8. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 1, characterized in that: The enterprise classification based on the similarity between the main business text data of different enterprises includes: For the main business text data of each enterprise, use Jieba word segmentation tool to process it and obtain the word segmentation set corresponding to the main business text data of each enterprise; For any two companies, calculate the Jaccard similarity coefficient between the word sets of these two companies; The optimal K value is preset, and all enterprises are clustered based on the optimal K value and the K-means clustering algorithm to obtain clusters; wherein the distance metric is the value after the negative correlation mapping of the Jaccard similarity coefficient between two enterprises.

9. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 3 is characterized in that: The method for obtaining the innovation capability label includes: For any enterprise, the number of patent applications, sales and value of all products, R&D investment costs, and technological innovation capability of the enterprise are used as innovation indicator data; In each type of enterprise, the entropy weight method is used to calculate the weight factor of each innovation indicator data of all enterprises; For any enterprise, the value of the innovation index data is weighted and averaged using the weight factor of the enterprise's innovation index data to obtain the innovation capability value of the enterprise; In each type of enterprise, the enterprises are sorted in ascending order according to their innovation capability values ​​to obtain a ranking sequence, and the sequence number value of each enterprise in the ranking sequence is used as the innovation capability label of each enterprise.

10. The method for automatically generating dynamic enterprise portraits based on multi-source data integration according to claim 1, characterized in that: The method for obtaining the enterprise innovation capability portrait includes: Generate corresponding enterprise innovation capability portraits based on visualization tools and the enterprise's innovation capability labels.

Images