Enterprise multi-dimensional information management system based on artificial intelligence

Abstract

The invention relates to the technical field of enterprise information management, in particular to an enterprise multi-dimensional information management system based on artificial intelligence. Comprising a data acquisition unit used for acquiring enterprise internal data and public data; the data preprocessing unit is used for performing data cleaning on the data acquired by the data acquisition unit and storing the data in the data storage unit; the information evaluation unit is used for extracting the data in the data storage unit, calculating industry trend parameters, market demand parameters and policy orientation indexes and calculating information evaluation parameters M and is used for assisting enterprises to predict and judge market development changes; the information early warning unit is used for evaluating whether the enterprise internal data accords with a preset threshold value or not; and the display unit is used for displaying the data information, the information early warning parameter M and the early warning information in the data storage unit. According to the invention, effective management and analysis of information are realized, enterprise decision making and operation are assisted through identification and mining of enterprise internal information and public information, and management of enterprise information resources is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise information management, and in particular, to an enterprise multi-dimensional information management system based on artificial intelligence. Background Art

[0002] With the continuous development of society and the continuous change of the economic situation, the development of enterprises has gradually changed from extensive development without need to refined management development; especially in China, there are a large number of small, medium and micro enterprises, and the entrepreneurship and development of these enterprises have injected vitality into China's economic development; however, the average survival time of small, medium and micro enterprises is not long. In addition to objective reasons such as the fierce market competition environment, more importantly, small, medium and micro enterprises cannot effectively manage information during the entrepreneurship process, and cannot reasonably analyze and utilize multi-dimensional information to avoid risks; therefore, for a huge number of small, medium and micro enterprises, it is necessary to combine data analysis and artificial intelligence and other tools to assist enterprises in making decisions and operations, and optimize the management of enterprise information resources. At present, the research on enterprise information resource management is still relatively weak, so it is necessary to design an enterprise multi-dimensional information management system. Summary of the Invention

[0003] The purpose of the present invention is to provide an enterprise multi-dimensional information management system based on artificial intelligence to solve the above-mentioned technical defects. The present invention collects enterprise internal data and public data through a data collection unit, and evaluates industry trends, market demands and policy orientations through an information evaluation unit for the collected information data. Through comprehensive evaluation, an information evaluation parameter M is obtained to clarify the guiding significance of a large amount of information for enterprise development, realize the effective management and analysis of information, reduce the lag of enterprise information recognition through the identification and mining of enterprise internal information and public information, assist enterprises in making decisions and operations, and optimize the management of enterprise information resources.

[0004] The purpose of the present invention can be achieved by the following technical solutions: An enterprise multi-dimensional information management system based on artificial intelligence, including a data collection unit for collecting enterprise internal data and public data; a data preprocessing unit for cleaning the data collected by the data collection unit and storing the cleaned data in a data storage unit; an information evaluation unit for extracting the data in the data storage unit and calculating an industry trend parameter T, a market demand parameter K and a policy orientation index P; then calculating an information evaluation parameter M and storing the information evaluation parameter M in the data storage unit for assisting enterprises in predicting and judging market development changes; The information warning unit obtains the information evaluation parameter M, evaluates whether the enterprise internal data meets the preset threshold, and if not, generates a warning message, stores it in the data storage unit and sends it to the display unit; The display unit is used to display the data information, information warning parameter M and warning message in the data storage unit.

[0005] The operation steps of the enterprise multi-dimensional information management system based on artificial intelligence include: S11 Collect enterprise internal data: The enterprise internal data includes enterprise basic information, enterprise innovation and R & D data, enterprise financial data, enterprise human resource data, enterprise internal business data, and further includes R & D data, enterprise cost data, enterprise human resource data, enterprise sales data; The enterprise basic information includes information such as registered capital, industry, establishment time, registered address, and enterprise scale; The enterprise internal data is manually input and maintained; S12 Collect public data: The public data includes industry information, market information, and policy information. The information is collected manually or legal data is obtained by using a crawler, and then the data is screened manually; The industry information includes industry reports, information disclosed by industry associations, information on national official websites, information on industry data websites, etc.; The market information includes user reviews, social media data, online questionnaire feedback, etc.; The policy information includes public policy documents, tax documents, laws and regulations documents, department documents, etc.; S13 Data cleaning: Dedup, handle missing values and outliers for the collected enterprise internal data and public data, and format and store the cleaned data in the data storage unit; S14 Evaluate the industry trend parameter T: Obtain the keywords of the industry information data in the public data through the TF-IDF model, define the keyword distribution weights, calculate the number of documents n_d and the total keyword score t_d, then the industry trend index T = t_d / n_d, T = 0~1, and the total keyword score t_d is obtained by weighted calculation and summation of the TF-IDF values of the keywords in all documents; TF-IDF is a method for measuring the importance of keywords in a text; It determines the uniqueness of a certain word for the current document by calculating the frequency of a word in a certain document (TF) and the universality of the word in all documents (IDF); S15 Evaluate the market demand parameter K: Use the market information data in the public data collected in S12 to construct an LSTM model, score the new market information data after training the LSTM model, calculate the number of data entries i and the sentiment score ki_d, and then calculate the market demand parameter , where ki_d represents the sentiment score of the i-th market information data, Represents the sum of the sentiment scores of i pieces of market information data; The LSTM model can: Input: Text data such as user reviews, social media posts, product feedback, etc.

[0006] Output: The sentiment classification of each piece of text, such as positive / negative sentiment, and / or sentiment score, such as the sentiment intensity between 0 and 1. By classifying the sentiment of a large amount of text, we can summarize the overall market demand trend for a certain product or service and quantify it with the market demand parameter K; K approaches 1: The market feedback is positive and the demand is strong; K approaches 0: The market feedback is negative and the demand is insufficient; S16 evaluates the policy orientation index P: Define supportive keywords and restrictive keywords, and assign weights to the supportive keywords and restrictive keywords. Calculate the supportive score S_positive of the supportive keywords and the negative score S_negative of the restrictive keywords in the policy information data in the public data through the TF-IDF model to obtain the support strength parameter , where Si represents the numerical value of the support strength parameter of the i-th policy information data; among them, the supportive score (S_positive) = ∑(supportive keyword weight × TF-IDF (supportive keyword)), and the negative score (S_negative) = ∑(restrictive keyword weight × TF-IDF (restrictive keyword)); Si is normalized to ensure Si = 0 to 1; Define the policy coverage parameter Ci, which represents the coverage of the i-th policy information data, and Ci = 0 to 1; Define the implementation intensity parameter Ei, which represents the implementation intensity of the i-th policy information data, and Ei = 0 to 1; Calculate the policy orientation index , where i represents the total number of policy information data obtained, and Pi represents the numerical value of the policy orientation index of the i-th policy information data, where Pi = p1*Si + p2*Ci + p3*Ei; where p1, p2, and p3 are the weight parameters of the policy orientation index, p1, p2, and p3 are all greater than 0, and p1 + p2 + p3 = 1; S17 obtains the information warning parameter M: Dynamically store the evaluated industry trend parameter T, market demand parameter K, and policy orientation index P, and then calculate the information warning parameter , where , are the exponential factors of the industry trend parameter T and the market demand parameter K respectively and are both greater than 0; is the influencing factor of the policy orientation index P; S18 Determine whether a warning message is generated: Set the internal data threshold of the enterprise, then set the threshold of the information evaluation parameter M according to the enterprise development plan, and combine the internal data of the enterprise, the internal data threshold of the enterprise, the information evaluation parameter M, and the threshold of the information evaluation parameter M to determine whether a warning message is generated through evaluation; S19 Information display: The display unit obtains the data information, the information warning parameter M, and the warning message in the data storage unit and performs data visualization display, which is used to reflect the internal data situation of the enterprise, facilitate the monitoring of the enterprise operation situation, and at the same time reflect the changes in industry trends, market demands, and policy orientations. The value of M is displayed in real time through the dashboard, and the multi-dimensional information of the enterprise is displayed through the display unit.

[0007] In the present invention, the industry trend parameter T is calculated by obtaining the public data in the data storage unit through the TF-IDF model, extracting the keywords of the industry information data in the public data, defining the weights assigned to the keywords, calculating the number of documents n_d and the total keyword score t_d, then the industry trend index T = t_d / n_d, T = 0~1, and the total keyword score t_d is obtained by weighted calculation and summation of the TF-IDF values of the keywords in all documents.

[0008] Among them, in the TF-IDF model, the term frequency TF = n(w) / t(w), where n(w) represents the number of times the keyword w appears in the document, and t(w) represents the total number of words in the document; the inverse document frequency IDF is calculated as IDF = log(n_d / w_d), where n_d represents the number of documents, and w_d represents the number of documents containing the keyword w; the TF-IDF value is calculated as TF-IDF = TF * IDF; calculate the total keyword score ; TF-IDF(w) represents the TF-IDF value of the keyword w, and q_w represents the weight of the keyword w.

[0009] It should be further noted that: The market demand parameter K is evaluated for the market information data through the LSTM model, the number of data items i and the sentiment score ki_d are calculated, and then the market demand parameter is calculated where ki_d represents the sentiment score of the i-th market information data item, represents the sum of the sentiment scores of i market information data items.

[0010] Among them, the construction steps of the LSTM model include: S151 Obtain market information data: The market information data includes user comments, social media data, and online questionnaire feedback; the obtained market information data is sentiment-annotated to obtain a set of sequences; S152 Text preprocessing: Convert the text data into numerical word vectors. First, clean the text, including removing punctuation marks, converting to lowercase, and removing stop words; then perform word vector conversion; S153 Construct an LSTM model: It includes a word embedding layer for converting text data into word vectors; a long short-term memory layer for extracting the time series features of the text and learning the emotional expression pattern of the text; and a fully connected layer using the sigmoid activation function for finally outputting an emotional score between 0 and 1. S154 Train the LSTM model: Train it with the prepared market information data and emotional labels. S155 Perform emotional prediction using the trained LSTM model: Input the market information data into the trained LSTM model to obtain the emotional score ki_d, where the emotional score ki_d is greater than 0 and less than 1. S156 Calculate the market demand parameter K: Sum up the emotional scores ki_d obtained in S155, and then calculate the average value of the emotional scores, which is the market demand parameter K. The formula is 。

[0011] It should be further noted that: The policy orientation index P evaluates the policy information data through the TF-IDF model, calculates the number of data entries i and the policy orientation index value Pi, and then calculates the policy orientation index where Pi represents the policy orientation index value of the i-th policy information data, represents the sum of the policy orientation index values of i policy information data entries.

[0012] Among them, the calculation steps of the policy orientation index value Pi include: S161 Obtain policy information data: The policy information data includes public policy documents, tax documents, laws and regulations documents, and departmental documents. S162 Define keywords: Define supportive keywords and restrictive keywords, and assign weights to the supportive keywords and restrictive keywords. S163 Calculate the supportive score S_positive: Calculate the supportive score S_positive of the supportive keywords in the policy information data in the public data through the TF-IDF model. The supportive score (S_positive) = ∑(weight of supportive keyword × TF-IDF (supportive keyword)). S164 Calculate the negativity score S_negative: Calculate the negativity score S_negative of the restrictive keywords in the policy information data in the public data through the TF-IDF model. The negativity score (S_negative) = ∑(weight of restrictive keyword × TF-IDF (restrictive keyword)). S165 Calculate the support strength parameter Si: The support strength parameter , Si represents the numerical value of the support strength parameter of the i-th policy information data; S166 defines the policy coverage parameter Ci, representing the coverage of the i-th policy information data, where Ci = 0~1; S167 defines the execution intensity parameter Ei, representing the execution intensity of the i-th policy information data, where Ei = 0~1; S168 calculates the policy orientation index value Pi: The policy orientation index value Pi is calculated through the formula Pi = p1*Si + p2*Ci + p3*Ei, where p1, p2, and p3 are the weight parameters of the policy orientation index, and p1 + p2 + p3 = 1. The operation steps of the information warning unit include: S181 Obtain enterprise internal data: Enterprise internal data includes R & D data, enterprise cost data, enterprise human resource data, and enterprise sales data; S182 Set internal thresholds: Set the upper and lower thresholds for R & D data, set the upper and lower thresholds for enterprise cost data, set the upper and lower thresholds for enterprise human resource data, and set the upper and lower thresholds for enterprise sales data; S183 Set the threshold of the information evaluation parameter M: Set the threshold of the information evaluation parameter M according to the enterprise development plan; S184 Evaluate enterprise internal data: Obtain the information evaluation parameter M and compare it with the set threshold of the information evaluation parameter M. If M is less than the set threshold of the information evaluation parameter M, respectively compare whether the enterprise's innovation R & D data meets the lower threshold and whether the enterprise's sales data meets the lower threshold. If the determination is negative, generate a warning message and send it to the display unit; otherwise, no warning message is generated; If M is less than the set threshold of the information evaluation parameter M, respectively compare whether the enterprise's cost data exceeds the lower threshold and whether the enterprise's human resource data exceeds the lower threshold. If the determination is positive, generate a warning message and send it to the display unit; otherwise, no warning message is generated; If M is greater than the set threshold of the information evaluation parameter M, respectively compare whether the enterprise's innovation R & D data meets the upper threshold and whether the enterprise's sales data meets the upper threshold. If the determination is negative, generate a warning message and send it to the display unit; otherwise, no warning message is generated; If M is greater than the set threshold of the information evaluation parameter M, respectively compare whether the enterprise's cost data exceeds the upper threshold and whether the enterprise's human resource data exceeds the upper threshold. If the determination is positive, generate a warning message and send it to the display unit; otherwise, no warning message is generated.

[0013] Calculate the information evaluation parameter M according to the obtained industry trend parameter T, market demand parameter K, and policy orientation index P. The calculation formula is: , where , are respectively the industry trend parameter T and the market demand parameter K index factors and are both greater than 0; is the influencing factor of the policy orientation index P.

[0014] In the process of enterprise entrepreneurship and development, the collection and analysis of information are very important. Without understanding the market changes and industry development laws and blindly carrying out development and investment, it is easy to lead to entrepreneurial failure due to information asymmetry. Therefore, through the enterprise multi-dimensional information management system, necessary data collection is carried out. On the one hand, it is beneficial for the enterprise to keep up with the industry dynamics in real time. On the other hand, through data analysis, it can guide the market development layout. When the enterprise development information is not equal to the industry development status, the information evaluation parameter M is used to evaluate the difference, assisting the enterprise to timely evaluate the industry development situation and timely adjust the enterprise strategic layout.

[0015] The beneficial effects of the present invention are as follows: (1) The present invention collects, analyzes and displays the information required for enterprise development through collecting internal data and public data of the enterprise, which is beneficial to the management of multi-dimensional information of the enterprise; (2) The present invention collects and manages multi-dimensional information, excavates the internal connection of information data, and obtains the information evaluation parameter through evaluation and analysis, which can predict the impact of industry development, market demand and policy orientation on the future development of the enterprise, and assist the enterprise in strategic layout and resource allocation; (3) According to the collected data, the present invention can quickly and conveniently obtain the industry trend parameter, market demand parameter and policy orientation index, and reflect the current industry development status through multiple dimensions, which is beneficial for the enterprise to expand timely or avoid risks and reduce investment, and can also assist the enterprise to segment the market to find business opportunities, and comprehensively utilize multi-dimensional information to assist the enterprise in innovative development. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings; Figure 1 is the system flow block diagram of the present invention; Figure 2 is the operation flow chart of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1: The present invention relates to an enterprise multi-dimensional information management system based on artificial intelligence. It includes a data collection unit for collecting internal enterprise data and public data. The internal enterprise data is input manually and stored in the data storage unit after input. The public data is obtained manually or by web crawler, and after being obtained, it is stored in the data storage unit after data cleaning. A data preprocessing unit for cleaning the data collected by the data collection unit and storing the cleaned data in the data storage unit. An information evaluation unit extracts the data from the data storage unit and calculates the industry trend parameter T, the market demand parameter K, and the policy orientation index P; then calculates the information evaluation parameter M and stores the information evaluation parameter M in the data storage unit to assist the enterprise in predicting and judging market development changes. An information warning unit obtains the information evaluation parameter M and evaluates whether the internal enterprise data meets the preset threshold. If it does not meet the threshold, a warning message is generated, stored in the data storage unit, and sent to the display unit. A display unit for displaying the data information, the information warning parameter M, and the warning message in the data storage unit.

[0019] In the present invention, the industry trend parameter T is obtained by using the TF-IDF model to obtain the public data in the data storage unit, extracting the keywords of the industry information data in the public data, defining the weights assigned to the keywords, calculating the number of documents n_d and the total keyword score t_d, then the industry trend index T = t_d / n_d, T = 0 to 1, and the total keyword score t_d is obtained by weighted calculation and summation of the TF-IDF values of the keywords in all documents.

[0020] In the TF-IDF model, the term frequency TF = n(w) / t(w), where n(w) represents the number of times the keyword w appears in the document, and t(w) represents the total number of words in the document; calculate IDF, IDF = log(n_d / w_d), n_d represents the number of documents, and w_d represents the number of documents containing the keyword w; calculate the TF-IDF value: TF-IDF = TF * IDF; calculate the total keyword score ; TF-IDF(w) represents the TF-IDF value of the keyword w, and q_w represents the weight of the keyword w.

[0021] For example, if 10 pieces of new energy-related article data are obtained, and then keywords are screened out, the keywords include new energy, demand, policy, etc. Taking new energy as an example, (1) calculate the term frequency TF = n(w) / t(w), where n(w) represents the number of times the keyword w appears in the document, and t(w) represents the total number of words in the document. The total number of words in the collected articles is 10,000, and "new energy" appears 200 times. Then the term frequency TF of new energy = 800 / 10,000 = 0.08; (2) Calculate IDF, IDF = log(n_d / w_d), where n_d represents the number of documents and w_d represents the number of documents containing the keyword w; Among 10 articles, new energy appears in 8 articles. Then the IDF of "new energy" = log(10 / 8) = 0.0969; (4) Calculate the TF-IDF value: TF-IDF = TF * IDF; then the TF-IDF of "new energy" = 0.08 * 0.0969 = 0.007752; Calculate the TF-IDF values of other keywords respectively; Calculate the total score of keywords ; TF-IDF(w) represents the TF-IDF value of keyword w, q_w represents the weight of keyword w, and the sum of keyword weights is 1; t_d is obtained by weighted calculation and summation of the TF-IDF values of keywords for all documents.

[0022] Calculate the market demand parameter K. Evaluate the market information data through the LSTM model, calculate the number of data entries i and the sentiment score ki_d, and then calculate the market demand parameter , where ki_d represents the sentiment score of the i-th market information data, represents the sum of the sentiment scores of i market information data.

[0023] The LSTM model can: Input: Text data such as user comments, social media posts, product feedback, etc.

[0024] Output: The sentiment classification of each piece of text (such as positive / negative sentiment), and / or the sentiment score (such as the sentiment intensity between 0 and 1). By classifying the sentiment of a large amount of text, we can summarize the overall market demand trend for a certain product or service and quantify it with the market demand parameter K; First, extract the sentiment information of users about the product or service from the market data.

[0025] Then, use the LSTM model to classify the user sentiment and obtain the sentiment score (positive score) of each comment.

[0026] Then calculate the market demand parameter: K approaches 1: The market feedback is positive and the demand is strong; K approaches 0: The market feedback is negative and the demand is insufficient; The specific steps are as follows: S151 Obtain market information data: The market information data includes user reviews, social media data, and online questionnaire feedback; perform sentiment annotation on the obtained market information data to obtain a sequence of numbers. For example, obtain the following data: "This product is very easy to use and has very powerful functions!", "The price is a bit expensive, but the overall quality is good.", "It's not worth the price at all, and the experience is very bad.", "The service attitude is very good, and the logistics is also very fast.", "The product quality is too poor to be used at all!", Perform manual annotation on the above data, where 1 represents positive and 0 represents negative. labels = [1, 1, 0, 1, 0].

[0027] S152 Text preprocessing: Convert the text data into numerical word vectors. First, clean the text, including removing punctuation marks, converting to lowercase, and removing stop words (such as "de", "shi", "zhe"); then perform word vector conversion.

[0028] The specific steps of word vector conversion include: (1) Word segmentation: Use a word segmentation function to split the text into a sequence of words; for example, given a sentence "This mobile phone has a very high cost performance"; split it into the sequence {"This", "mobile phone", "cost performance", "very high"}; (2) Word normalization: Replace words with standard words with the same meaning, such as replacing "wonderful" with "good" and "purchase" with "buy"; (3) Perform word indexing: Use a dictionary mapping to convert each word into a unique index to obtain a sequence of numbers. For example, {"mobile phone", "cost performance", "very high"} is mapped through the dictionary to obtain the sequence of numbers {12, 45, 78}; (4) Convert to word vectors: Use pre-trained Chinese word vectors (such as Word2Vec, GloVe, FastText, BERT) to convert words into vectors of a fixed dimension; (5) Unify the data length: Fill or truncate the obtained word vectors so that the word vector data has a unified length; in this embodiment, the set length is 10. If the length of the word vector data is less than 10, fill it with 0 to 10. For example, if the word vector is {12, 45, 67, 89, 33, 22}, the length is less than 10 and needs to be filled to 10 digits to get {12, 45, 67, 89, 33, 22, 0, 0, 0, 0}; if the word vector data exceeds 10, keep the first 10 digits of the data.

[0029] S153 Build an LSTM model: It includes a word embedding layer for converting text data into word vectors; a long short-term memory layer for extracting the time series features of the text and learning the emotional expression pattern of the text; and a fully connected layer with a sigmoid activation function for finally outputting an emotional score between 0 and 1. The word embedding layer sets the vocabulary size to 5000 (at most 5000 words), the word vector dimension to 64 (each word is mapped to a 64-dimensional vector), and the sentence length to 10 (each text contains at most 10 words). The long short-term memory layer is set to 128 dimensions.

[0030] S154 Train the LSTM model: Train it with the prepared market information data and emotional labels, with 2 data items per batch and 5 training rounds. S155 Use the trained LSTM model for emotion prediction: Input the market information data into the trained LSTM model to obtain the emotional score ki_d, where the emotional score ki_d is greater than 0 and less than 1. For example, for the data "This product is really good, highly recommended!", the emotional score output by the LSTM model is 0.9213; for the data "The usage experience is very poor, not recommended for purchase", the emotional score output by the LSTM model is 0.1532.

[0031] S156 Calculate the market demand parameter K: Sum up the emotional scores ki_d obtained in S155, and then calculate the average value of the emotional scores, which is the market demand parameter K. The formula is ; When the calculated market demand parameter K is less than 0.5, it indicates that the market sentiment is negative and the demand may be low. If K is greater than 0.5, it indicates that the market sentiment is positive and the demand is good. The closer K is to 1, the more positive the market feedback and the relatively stronger the market demand.

[0032] In this embodiment, the policy orientation index P evaluates the policy information data through the TF-IDF model, calculates the number of data items i and the policy orientation index value Pi, and then calculates the policy orientation index , where Pi represents the policy orientation index value of the i-th policy information data item, represents the sum of the policy orientation index values of i policy information data items.

[0033] Specifically, it is manifested as: Define supportive keywords and restrictive keywords, and assign weights to the supportive keywords and restrictive keywords. Calculate the supportive score S_positive of the supportive keywords and the negative score S_negative of the restrictive keywords in the policy information data of the public data through the TF-IDF model to obtain the support strength parameter where \(S_i\) represents the numerical value of the support strength parameter of the \(i\)-th policy information data. Among them, the positive support score (\(S_{positive}\)) = \(\sum\) (weight of positive support keywords × TF-IDF (positive support keywords)), and the negative score (\(S_{negative}\)) = \(\sum\) (weight of restrictive keywords × TF-IDF (restrictive keywords)); \(S_i\) is normalized to ensure that \(S_i\in[0,1]\).

[0034] Define the policy coverage parameter \(C_i\) to represent the coverage of the \(i\)-th policy information data, where \(C_i\in[0,1]\); \(C_i=\frac{actual\ coverage\ range}{theoretical\ maximum\ coverage\ range}\); \(C_i\) is an estimated value representing a trend. For example, if the subsidy amount or coverage reaches the upper limit of the industry expectation, then \(C = 1\); if the policy only covers a very small part of the market or has no clear scope, then \(C\approx0\).

[0035] Define the implementation intensity parameter \(E_i\) to represent the implementation intensity of the \(i\)-th policy information data, where \(E_i\in[0,1]\); Define that the implementation intensity parameter \(E_i\) reflects the actual binding force and timeliness of the policy. The value of \(E_i\) is adjusted according to the specific policy. For example: For long-term policies over 3 years, the score is relatively high, so \(E = 1\); For short-term policies less than 1 year, the score is relatively low, so \(E = 0.3\); For mandatory policies (such as "enforced execution"), the score is relatively high, and for encouraging policies, the score is relatively low; Calculate the policy orientation index where \(i\) represents the total number of policy information data obtained, and \(P_i\) represents the numerical value of the policy orientation index of the \(i\)-th policy information data, where \(P_i = p_1\times S_i + p_2\times C_i + p_3\times E_i\); among them, \(p_1\), \(p_2\), and \(p_3\) are the weight parameters of the policy orientation index, \(p_1\), \(p_2\), and \(p_3\) are all greater than 0, and \(p_1 + p_2 + p_3 = 1\); in this embodiment, \(p_1 = 0.7\), \(p_2 = 0.1\), and \(p_3 = 0.2\); The calculation methods of the positive support score \(S_{positive}\) and the negative score \(S_{negative}\) are the same as those in step S14 and will not be elaborated here; Among them, the calculation steps of the numerical value of the policy orientation index \(P_i\) include: S161 Obtain policy information data: The policy information data includes public policy documents, tax documents, laws and regulations documents, and departmental documents; S162 Define keywords: Define positive support keywords and restrictive keywords, and assign weights to positive support keywords and restrictive keywords; S163 Calculate the supportive score S_positive: Calculate the supportive score S_positive of the supportive keywords in the policy information data in the public data through the TF-IDF model. Supportive score (S_positive) = ∑(supportive keyword weight × TF-IDF(supportive keyword)); S164 Calculate the negative score S_negative: Calculate the negative score S_negative of the restrictive keywords in the policy information data in the public data through the TF-IDF model. Negative score (S_negative) = ∑(restrictive keyword weight × TF-IDF(restrictive keyword)); S165 Calculate the support strength parameter Si: Support strength parameter where Si represents the value of the support strength parameter of the i-th policy information data; S166 Define the policy coverage parameter Ci, which represents the coverage of the i-th policy information data, and Ci = 0~1; S167 Define the execution intensity parameter Ei, which represents the execution intensity of the i-th policy information data, and Ei = 0~1; S168 Calculate the policy orientation index value Pi: Calculate the policy orientation index value Pi through the formula Pi = p1 * Si + p2 * Ci + p3 * Ei, where p1, p2, and p3 are the weight parameters of the policy orientation index, and p1 + p2 + p3 = 1. The operation steps of the information early warning unit include: S181 Obtain enterprise internal data: Enterprise internal data includes R & D data, enterprise cost data, enterprise human resource data, and enterprise sales data; S182 Set internal thresholds: Set the upper and lower thresholds for R & D data, set the upper and lower thresholds for enterprise cost data, set the upper and lower thresholds for enterprise human resource data, and set the upper and lower thresholds for enterprise sales data; S183 Set the threshold of the information evaluation parameter M: Set the threshold of the information evaluation parameter M according to the enterprise development plan, denoted as ZX, which represents the enterprise's expectation for market development. Generally, the average value of the first three information evaluation parameters M is used as ZX. For example, the sum of the information evaluation parameters M in the first three quarters is divided by 3 to obtain the information evaluation parameter M threshold ZX in the fourth quarter; S184 Evaluate enterprise internal data: Obtain information evaluation parameter M and compare it with ZX. If M is less than ZX, it indicates that the overall market situation is less than expected. Then, respectively compare whether the enterprise's innovation and R & D data meet the lower threshold or whether the enterprise's sales data meet the lower threshold. If one of the judgment conditions is determined to be otherwise, an early warning message is generated and sent to the display unit; otherwise, no early warning message is generated. When both are not otherwise, no early warning message is generated.

[0036] The lower threshold of the enterprise's innovation and R & D data represents the minimum R & D data index set by the enterprise in a sluggish market environment. If the enterprise's innovation and R & D data are less than the lower threshold of the enterprise's innovation and R & D data, it indicates that the enterprise's R & D does not meet the market development situation and the product competitiveness decreases. At this time, if it is determined to be otherwise, an early warning message is generated and sent to the display unit; Compare whether the enterprise's sales data meet the lower threshold. The lower threshold of the enterprise's sales data represents the minimum sales business target set by the enterprise in a sluggish market environment. If the enterprise's sales data are less than the lower threshold of the enterprise's sales data, it indicates that the enterprise's sales situation cannot meet the basic development requirements of the enterprise. At this time, if it is determined to be otherwise, an early warning message is generated and sent to the display unit; The above judgment is made from the perspective of enterprise development. In the case of a market environment less than expected, it is necessary to meet the minimum R & D and sales performance to maintain the development of the enterprise. If it cannot be met, early warning and adjustment of the enterprise's strategic deployment are required in a timely manner.

[0037] If M is less than ZX, respectively compare whether the enterprise's cost data exceed the lower threshold and whether the enterprise's human resources data exceed the lower threshold. If one of the judgment conditions is determined to be yes, an early warning message is generated and sent to the display unit; otherwise, no early warning message is generated. Among them, the lower threshold of the enterprise's human resources data is the maximum human resources (such as the number of personnel) in a sluggish market environment. If the actual human resources data exceed the lower threshold of the enterprise's human resources data, it means that the enterprise's human resources management exceeds the limit and increases the enterprise's operation burden. At this time, if it is determined to be yes, an early warning message is generated and sent to the display unit; The lower threshold of the enterprise's cost data is the maximum enterprise cost (such as procurement cost) in a sluggish market environment. If the actual enterprise cost data exceed the lower threshold of the enterprise's cost data, it means that the enterprise's cost management exceeds the limit and increases the enterprise's operation burden. At this time, if it is determined to be yes, an early warning message is generated and sent to the display unit; The above judgment is made from the perspective of enterprise operation. In the case of a market environment less than expected, it is necessary to meet cost reduction and avoid expansion to maintain the development of the enterprise. If it exceeds the limit, early warning and adjustment of the enterprise's strategic deployment are required in a timely manner.

[0038] If M is greater than ZX, indicating that the market environment is better than expected, then compare whether the enterprise's innovation and R & D data meets the upper limit threshold or whether the enterprise's sales data meets the upper limit threshold respectively. If the determination is negative, generate a warning message and send it to the display unit; otherwise, no warning message is generated. The upper limit threshold of the enterprise's innovation and R & D data represents the minimum R & D data index set by the enterprise when the market environment is good. If the enterprise's innovation and R & D data is less than the upper limit threshold of the enterprise's innovation and R & D data, it indicates that the enterprise's R & D efforts are insufficient and it is difficult to meet the enterprise's development needs. At this time, if the determination is negative, generate a warning message and send it to the display unit. Compare whether the enterprise's sales data meets the upper limit threshold. The upper limit threshold of the enterprise's sales data represents the minimum sales business target set by the enterprise when the market environment is good. If the enterprise's sales data is less than the upper limit threshold of the enterprise's sales data, it indicates that the enterprise's sales situation cannot meet the requirements of the enterprise's rapid development. At this time, if the determination is negative, generate a warning message and send it to the display unit. The above judgments are made from the perspective of the enterprise's development. In the case where the market environment is greater than expected, it is necessary to meet the requirements of improving R & D capabilities and sales performance to promote the rapid development of the enterprise. If it cannot be met, it is necessary to give a warning in a timely manner and adjust the enterprise's strategic deployment.

[0039] If M is greater than ZX, compare whether the enterprise's cost data exceeds the upper limit threshold and whether the enterprise's human resources data exceeds the upper limit threshold respectively. If the determination is positive, generate a warning message and send it to the display unit; otherwise, no warning message is generated.

[0040] Among them, the upper limit threshold of the enterprise's human resources data is the maximum human resources (such as the number of personnel) when the market environment is good. If the actual human resources data exceeds the upper limit threshold of the enterprise's human resources data, it means that the enterprise's human resources management has exceeded the limit and there is a tendency of blind expansion. At this time, if the determination is positive, generate a warning message and send it to the display unit. The upper limit threshold of the enterprise's cost data is the maximum enterprise cost (such as procurement cost) when the market environment is good. If the actual enterprise cost data exceeds the upper limit threshold of the enterprise's cost data, it means that the enterprise's cost management has exceeded the limit and increased the enterprise's operation burden. At this time, if the determination is positive, generate a warning message and send it to the display unit. The above judgments are made from the perspective of the enterprise's operation. In the case where the market environment is less than expected, it is necessary to meet the requirements of reducing costs and avoiding expansion to maintain the enterprise's development. If it exceeds the limit, it is necessary to give a warning in a timely manner and adjust the enterprise's strategic deployment.

[0041] The above thresholds are determined based on the enterprise's internal data and industry data, and need to be appropriately adjusted according to the actual situation to avoid the situation where the thresholds are too large or too small.

[0042] Calculate the information evaluation parameter M based on the obtained industry trend parameter T, market demand parameter K, and policy orientation index P. The calculation formula is: , where , are the exponential factors of the industry trend parameter T and the market demand parameter K respectively, and both are greater than 0; is the influence factor of the policy orientation index P.

[0043] reflects the influence of the industry trend. If T is larger, the trend is stronger, and M grows faster; reflects the influence of the market demand trend. If K is larger, the market demand is relatively strong, and M grows faster; represents the policy orientation situation. P is between [0, 1], controls its influence degree on M, and uses a power function to control the influence. suppresses M when P is low and amplifies M when P is high.

[0044] When T and K are larger, the exponential function and make M grow rapidly, indicating that the market opportunity is larger; at this time, when P≈0, indicating that the policy is extremely unfavorable, makes M close to 0, indicating that the market risk is higher; when P≈1, indicating that the policy is extremely favorable, has a smaller influence on M and keeps M at a high level, indicating that the market is supported by the policy.

[0045] Further explanation: Set to 2, to 1.5, to 0.8. When the industry trend parameter T = 0.8, it indicates that the industry trend is strong. The market demand parameter K indicates that the market demand is average. The policy orientation index P = 0.5 indicates that the policy support is average; then calculate , and calculate that M is approximately 7.01; Due to the relatively strong industry trend T = 0.8 and the medium market demand K = 0.6, the exponential amplification effect makes M larger. However, the average policy support P = 0.5 causes M to be somewhat suppressed, but still remains at a relatively high value of 7.01, indicating that there are strong opportunities in the market but policy risks need to be concerned about.

[0046] In order to further represent the influence of the information evaluation parameter M on market changes, M can be normalized. Then M1 = M / M(max), where M(max) is the maximum value in historical data. Normalizing M1 to [0, 1] can more intuitively reflect the influence of the industry trend, market demand, and policy orientation.

[0047] Implementation Two: The operating steps of the enterprise multi-dimensional information management system based on artificial intelligence include: S11 Collect enterprise internal data: Enterprise internal data includes enterprise basic information, enterprise innovation and R & D data, enterprise financial data, enterprise human resources data, and enterprise internal business data, and further includes R & D data, enterprise cost data, enterprise human resources data, and enterprise sales data; Enterprise basic information includes information such as registered capital, industry, establishment time, registered address, and enterprise scale; Enterprise internal data is entered and maintained manually. S12 Collect public data: Public data includes industry information, market information, and policy information. Information is collected manually or legal data is obtained using a crawler, and then the data is screened manually; Industry information includes industry reports, information disclosed by industry associations, information on national official websites, information on industry data websites, etc.; Market information includes user reviews, social media data, online questionnaire feedback, etc.; Policy information includes publicly available policy documents, tax documents, laws and regulations documents, departmental documents, etc. S13 Data cleaning: De-duplicate, handle missing values, and handle outliers for the collected enterprise internal data and public data, and format and store the cleaned data in the data storage unit. S14 Evaluate the industry trend parameter T: Obtain the keywords of the industry information data in the public data through the TF-IDF model, define the weights assigned to the keywords, calculate the number of documents n_d and the total keyword score t_d, then the industry trend index T = t_d / n_d, T = 0~1, and the total keyword score t_d is obtained by weighted calculation and summation of the TF-IDF values of the keywords in all documents. TF-IDF is a method for measuring the importance of keywords in text; it determines the uniqueness of a certain word for the current document by calculating the frequency of a word in a certain document (TF) and the universality of the word in all documents (IDF). S15 Evaluate the market demand parameter K: Use the market information data in the public data collected in S12 to construct an LSTM model, score the new market information data after training the LSTM model, calculate the number of data items i and the sentiment score ki_d, and then calculate the market demand parameter , where ki_d represents the sentiment score of the i-th market information data item, represents the sum of the sentiment scores of i market information data items; The LSTM model can: Input: Text data such as user reviews, social media posts, product feedback, etc.

[0048] Output: The sentiment classification of each piece of text, such as positive / negative sentiment, and / or sentiment scores, such as the sentiment intensity between 0 and 1. By classifying the sentiment of a large amount of text, we can summarize the overall market demand trend for a certain product or service and quantify it with the market demand parameter K; K approaches 1: The market feedback is positive and the demand is strong; K approaches 0: The market feedback is negative and the demand is insufficient; S16 Evaluate the policy orientation index P: Define supportive keywords and restrictive keywords, and assign weights to the supportive keywords and restrictive keywords. Calculate the supportive score S_positive of the supportive keywords and the negative score S_negative of the restrictive keywords in the policy information data in the public data through the TF-IDF model to obtain the support strength parameter , where Si represents the numerical value of the support strength parameter of the i-th policy information data; among them, the supportive score (S_positive) = ∑(weight of supportive keyword × TF-IDF (supportive keyword)), and the negative score (S_negative) = ∑(weight of restrictive keyword × TF-IDF (restrictive keyword)); Si is normalized to ensure Si = 0 to 1; Define the policy coverage parameter Ci, which represents the coverage of the i-th policy information data, and Ci = 0 to 1; Define the implementation intensity parameter Ei, which represents the implementation intensity of the i-th policy information data, and Ei = 0 to 1; Calculate the policy orientation index , where i represents the total number of policy information data obtained, and Pi represents the numerical value of the policy orientation index of the i-th policy information data, where Pi = p1 * Si + p2 * Ci + p3 * Ei; where p1, p2, and p3 are the weight parameters of the policy orientation index, p1, p2, and p3 are all greater than 0, and p1 + p2 + p3 = 1; S17 Obtain the information warning parameter M: Dynamically store the evaluated industry trend parameter T, market demand parameter K, and policy orientation index P, and then calculate the information warning parameter , where , are the exponential factors of the industry trend parameter T and the market demand parameter K respectively and are both greater than 0; is the influencing factor of the policy orientation index P; S18 Determine whether to generate a warning message: Set the enterprise internal data threshold, and then set the threshold of the information evaluation parameter M according to the enterprise development plan. Combine the enterprise internal data, the enterprise internal data threshold, the information evaluation parameter M, and the threshold of the information evaluation parameter M to determine whether to generate a warning message through evaluation; S19 Information Display: The display unit obtains the data information in the data storage unit, the information warning parameter M and the warning information, and performs data visualization display, which is used to reflect the internal data situation of the enterprise, facilitate the monitoring of the enterprise operation situation, and at the same time reflect the changes in industry trends, market demands and policy orientations. The M value is displayed in real time through the dashboard, and the multi-dimensional information of the enterprise is displayed through the display unit.

[0049] By collecting the internal data and public data of the enterprise, multi-dimensional collection, analysis and display of the information required for the enterprise development are carried out, which is beneficial to the management of the multi-dimensional information of the enterprise; collecting and managing the multi-dimensional information, mining the internal connection of the information data, evaluating and analyzing to obtain the information evaluation parameter, which can predict the impact of industry development, market demand and policy orientation on the future development of the enterprise, and assist the enterprise in strategic layout and resource allocation; according to the collected data, the industry trend parameter, market demand parameter and policy orientation index can be obtained quickly and conveniently, and the current industry development situation can be reflected through multiple dimensions, which is beneficial to the enterprise to expand in a timely manner or avoid risks and reduce investment, and can also assist the enterprise to segment the market to find business opportunities, and comprehensively utilize the multi-dimensional information to assist the enterprise in innovative development.

[0050] In the present invention, the setting of the threshold value is for the convenience of comparison. Regarding the size of the threshold value, it depends on the amount of sample data and the number of base values set by those skilled in the art for each group of sample data; as long as the proportional relationship between the parameter and the quantified value is not affected.

[0051] The above formulas are all obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value. The coefficients in the formulas are set by those skilled in the art according to the actual situation. As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An enterprise multi-dimensional information management system based on artificial intelligence, characterized in that: It includes a data collection unit for collecting internal enterprise data and public data; A data preprocessing unit, used for cleaning the data collected by the data collection unit and storing the cleaned data in the data storage unit; The information evaluation unit extracts the data in the data storage unit and calculates the industry trend parameter T, the market demand parameter K and the policy orientation index P; then calculates the information evaluation parameter M to assist enterprises in predicting and judging market development changes; The information warning unit obtains the information evaluation parameter M and evaluates whether the internal data of the enterprise meets the preset threshold. If not, a warning message is generated and sent to the display unit; The display unit is used to display the data information, information warning parameter M and warning information in the data storage unit.

2. According to claim 1, the enterprise multi-dimensional information management system based on artificial intelligence is characterized in that: The calculation of the industry trend parameter T obtains the public data in the data storage unit through the TF-IDF model, extracts the keywords of the industry information data in the public data, defines the keyword allocation weight, calculates the number of documents n_d and the total keyword score t_d, then the industry trend index T=t_d / n_d, T=0~1, and the total keyword score t_d is obtained by weighted calculation and summation of the keyword TF-IDF values ​​of all documents.

3. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 2 is characterized in that: In the TF-IDF model, the word frequency TF=n(w) / t(w) is calculated, where n(w) represents the number of times the keyword w appears in the document, and t(w) represents the total number of words in the document; IDF is calculated, IDF=log(n_d / w_d), n_d represents the number of documents, and w_d represents the number of documents containing the keyword w; TF-IDF value is calculated: TF-IDF=TF*IDF; the total score of the keyword is calculated ; TF-IDF(w) represents the TF-IDF value of keyword w, and q_w represents the weight of keyword w.

4. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 1 is characterized in that: The market demand parameter K is evaluated by the LSTM model on the market information data, the number of data items i and the sentiment score ki_d are calculated, and then the market demand parameter is calculated. , where ki_d represents the sentiment score of the ith market information data, Represents the sum of sentiment scores of i pieces of market information data.

5. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 4 is characterized in that: The steps of building the LSTM model include: S151 obtains market information data: the market information data includes user comments, social media data, and online questionnaire feedback; sentiment annotation is performed on the obtained market information data to obtain a set of number series; S152 Text preprocessing: Convert text data into numerical word vectors. First, clean the text, including removing punctuation, converting to lowercase, and removing stop words; then perform word vector conversion; S153 builds an LSTM model: including a word embedding layer, which is used to convert text data into word vectors; a long short-term memory layer, which is used to extract the time series features of the text and learn the emotional expression pattern of the text; a fully connected layer, which uses a sigmoid activation function to finally output a sentiment score between 0 and 1; S154 Training LSTM model: training with prepared market information data and sentiment labels; S155 uses the trained LSTM model to perform sentiment prediction: input market information data into the trained LSTM model to obtain the sentiment score ki_d, which is greater than 0 and less than 1; S156 calculates the market demand parameter K: sums the sentiment scores ki_d obtained in S155, and then calculates the average of the sentiment scores, which is the market demand parameter K. The formula is: .

6. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 1 is characterized in that: The policy orientation index P evaluates the policy information data through the TF-IDF model, calculates the number of data items i and the policy orientation index value Pi, and then calculates the policy orientation index , where Pi represents the policy orientation index value of the i-th policy information data, Represents the sum of the policy orientation index values ​​of i policy information data;.

7. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 6 is characterized in that: The calculation steps of the policy orientation index value Pi include: S161 Obtaining policy information data: Policy information data includes public policy documents, tax documents, legal and regulatory documents, and departmental documents; S162 Define keywords: define supporting keywords and restrictive keywords, and assign weights to supporting keywords and restrictive keywords; S163 Calculate the support score S_positive: Calculate the support score S_positive of the support keywords of the policy information data in the public data through the TF-IDF model, support score (S_positive) = ∑(support keyword weight × TF-IDF (support keyword)); S164 calculates the negativity score S_negative: calculates the negativity score S_negative of the restrictive keywords in the policy information data in the public data through the TF-IDF model, and the negativity score (S_negative) = ∑(restrictive keyword weight×TF-IDF (restrictive keyword)); S165 calculation supports force parametersSi: supports force parameters , Si represents the support strength parameter value of the i-th policy information data; S166 defines the policy coverage parameter Ci, which represents the coverage of the i-th policy information data, Ci=0~1; S167 defines the execution intensity parameter Ei, which represents the execution intensity of the i-th policy information data, Ei=0~1; S168 calculates the policy orientation index value Pi: the policy orientation index value Pi is calculated by the formula Pi=p1*Si+p2*Ci+p3*Ei, where p1, p2, and p3 are the weight parameters of the policy orientation index, and p1+p2+p3=1.

8. The enterprise multi-dimensional information management system based on artificial intelligence according to claim 1 is characterized in that: The operation steps of the information early warning unit include: S181 obtains internal enterprise data: internal enterprise data includes R&D data, enterprise cost data, enterprise human resources data, and enterprise sales data; S182 sets internal thresholds: sets the upper and lower thresholds of R&D data, sets the upper and lower thresholds of enterprise cost data, sets the upper and lower thresholds of enterprise human resources data, and sets the upper and lower thresholds of enterprise sales data; S183 Setting the threshold of the information evaluation parameter M: setting the threshold of the information evaluation parameter M according to the enterprise development plan; S184 evaluates the internal data of the enterprise: obtains the information evaluation parameter M and compares it with the set information evaluation parameter M threshold. If M is less than the set information evaluation parameter M threshold, compares whether the enterprise's innovation and R&D data meets the lower limit threshold and whether the enterprise's sales data meets the lower limit threshold. If it is determined that it is not, a warning message is generated and sent to the display unit. Otherwise, no warning message is generated. If M is less than the set information evaluation parameter M threshold, compare whether the enterprise cost data exceeds the lower limit threshold and whether the enterprise human resources data exceeds the lower limit threshold. If it is determined to be yes, a warning message is generated and sent to the display unit, otherwise no warning message is generated; If M is greater than the set information evaluation parameter M threshold, compare whether the enterprise's innovation and R&D data meets the upper threshold, and whether the enterprise's sales data meets the upper threshold. If it is determined to be negative, a warning message is generated and sent to the display unit, otherwise no warning message is generated; If M is greater than the set information evaluation parameter M threshold, compare whether the enterprise cost data exceeds the upper threshold and whether the enterprise human resources data exceeds the upper threshold. If it is determined to be yes, a warning message is generated and sent to the display unit, otherwise no warning message is generated.

9. An enterprise multi-dimensional information management system based on artificial intelligence according to claims 1-8, characterized in that: The calculation formula of the information evaluation parameter M is: ,in , are the industry trend parameter T and market demand parameter K index factors, both of which are greater than 0; It is the influencing factor of the policy orientation index P.

Images