Enterprise recommendation method and device based on big data, equipment and medium

Abstract

The application relates to an enterprise recommendation method, device, equipment and medium based on big data, which comprises the following steps: obtaining public Internet data of a target enterprise, obtaining various policy files and related public files of an industry to which the target enterprise belongs, analyzing each enterprise in the industry to form an enterprise list, obtaining data of declaration conditions met by each policy project declared by each enterprise in the enterprise list, performing data normalization processing on all the obtained original data, forming low-dimensional feature vector data, mapping the low-dimensional feature vector data to a high-dimensional feature space based on a support vector machine by using a kernel function, directly calculating the inner product difference between the centralized high-dimensional new data and the support vector of the high-dimensional feature space in the high-dimensional feature space, eliminating the enterprise corresponding to the high-dimensional new data with a negative inner product difference, and outputting similar enterprise recommendation data of the target enterprise after sorting, so that enterprise recommendation processing about the target enterprise is realized, and the enterprise recommendation accuracy is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of big data processing, and relates to an enterprise recommendation method and device based on big data, a computer device and a computer readable storage medium. BACKGROUND

[0002] With the development of Internet technology, information on the Internet is growing exponentially, which undoubtedly increases the difficulty of obtaining accurate and useful information. In the field of enterprise recommendation, in some scenarios, one or more enterprises similar to a target enterprise need to be obtained according to the name of the target enterprise, which requires corresponding recommendation according to the enterprise-related information on the Internet. In the existing related technology of enterprise recommendation, the feature information of an enterprise is obtained according to the public information (enterprise social security number, business registration, bidding information, etc.) on the Internet or the manually entered enterprise information (financial statements, operating information, etc.), and then the enterprise is recommended based on the recommendation algorithm according to the feature information.

[0003] However, due to the need for enterprise privacy protection, the cost of accurately obtaining and entering enterprise information manually is extremely high, and the feature information of an enterprise obtained from the public enterprise data on the Internet is relatively single and cannot fully reflect the characteristics of the enterprise, which also affects the accuracy of enterprise recommendation. That is, the traditional enterprise recommendation technology still has the technical problem of low recommendation accuracy. SUMMARY

[0004] To solve the above-mentioned problems in the traditional method, the present application provides an enterprise recommendation method and device based on big data, a computer device and a computer readable storage medium, which can greatly improve the accuracy of enterprise recommendation.

[0005] To achieve the above-mentioned purpose, the embodiments of the present application adopt the following technical solutions:

[0006] On the one hand, an enterprise recommendation method based on big data is provided, comprising the following steps:

[0007] Obtaining public Internet data corresponding to the name of a target enterprise and storing the data in a first-dimensional array;

[0008] Obtaining each policy document and each related public document published on each official website, parsing the policy document and the related public document, obtaining the public enterprise list and the declaration condition data of each policy item, and storing the data in a second-dimensional array;

[0009] Performing data normalization processing on the original data in the first-dimensional array and the second-dimensional array to obtain a normalized low-dimensional feature vector and mapping the low-dimensional feature vector to a high-dimensional feature space based on a support vector machine using a kernel function;

[0010] In the high-dimensional feature space, the high-dimensional new data corresponding to the low-dimensional feature vector is centralized, and the inner product difference of the centralized high-dimensional new data and the support vector of the high-dimensional feature space is calculated;

[0011] The enterprise corresponding to the high-dimensional new data with the negative inner product difference is excluded, and the enterprise corresponding to the high-dimensional new data with the positive inner product difference is determined as a similar enterprise;

[0012] The similar enterprises are sorted in descending order of the inner product difference, and the similar enterprise recommendation data of the target enterprise is output.

[0013] In one of the embodiments, the disclosed public Internet data includes the industry of the enterprise, the establishment time, the number of employees, the taxpayer qualification, the registered capital, the paid-in capital, the nature of the enterprise, the registered address, the number of software copyrights, the number of trademarks, the number of certificates, the number of work copyrights, the number of utility model patents, the number of invention patents, the number of design patents, the number of invention publications, the listing type dimensions, and the bidding data.

[0014] In one of the embodiments, the normalization method used in the data normalization process includes the minimum-maximum normalization, the Z-score standardization, the mean-variance normalization, the decimal scaling standardization, or the Log function conversion.

[0015] In one of the embodiments, the kernel function is the RBF kernel function.

[0016] In one of the embodiments, the method further includes the steps of:

[0017] According to the similar enterprise recommendation data, the policy project data matching the target enterprise is output by using the collaborative filtering algorithm.

[0018] In one of the embodiments, the method further includes the steps of:

[0019] According to the introduction script of the target enterprise, the introduction script matching each similar enterprise in the similar enterprise recommendation data is generated respectively.

[0020] On the other hand, an enterprise recommendation device based on big data is also provided, which includes:

[0021] The first acquisition module is configured to acquire public Internet data corresponding to the name of the target enterprise and store the data in a first dimension array;

[0022] The second acquisition module is configured to acquire each policy file and each related public file published on each official website, parse the each policy file and each related public file, obtain the public enterprise list and the declaration condition data of each policy project, and then store the data in a second dimension array;

[0023] The normalization mapping module is configured to perform data normalization on the original data in the first dimension array and the second dimension array, to obtain a low-dimensional feature vector after normalization and to map the low-dimensional feature vector into a high-dimensional feature space based on a support vector machine using a kernel function;

[0024] The score calculation module is configured to perform centering processing on the high-dimensional new data corresponding to the low-dimensional feature vector in the high-dimensional feature space, to calculate an inner product difference between the high-dimensional new data after centering and a support vector of the high-dimensional feature space;

[0025] The enterprise determination module is configured to exclude an enterprise corresponding to the high-dimensional new data with a negative inner product difference, and to determine an enterprise corresponding to the high-dimensional new data with a positive inner product difference as a similar enterprise.

[0026] The recommendation output module is configured to sort the similar enterprises in descending order of the inner product difference, and to output similar enterprise recommendation data of the target enterprise.

[0027] In one of the embodiments, the disclosed Internet data includes dimensions of an industry to which an enterprise belongs, a time length of establishment, a number of employees, a taxpayer qualification, registered capital, actually paid-in capital, enterprise nature, a registration address, a number of software copyrights, a number of trademarks, a number of certificates, a number of work copyrights, a number of utility model patents, a number of invention patents, a number of design patents, a number of invention publications, and a listing type and bidding data.

[0028] In another aspect, a computer device is provided, which includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the above-described enterprise recommendation method based on big data are implemented.

[0029] In yet another aspect, a computer readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-described enterprise recommendation method based on big data are implemented.

[0030] One of the above technical solutions has the following advantages and beneficial effects:

[0031] The enterprise recommendation method, device, equipment and medium based on big data, by obtaining the public internet data of the target enterprise first, then obtaining various policy files and related public files of the industry to which the target enterprise belongs, and then analyzing each enterprise in the industry and forming an enterprise list, and obtaining the declaration condition data of each enterprise in the enterprise list, then performing data normalization processing on all the original data obtained, forming low-dimensional feature vector data and mapping to a high-dimensional feature space based on a support vector machine using a kernel function, directly calculating the inner product difference of the centralized high-dimensional new data and the support vector of the high-dimensional feature space in the high-dimensional feature space, eliminating the enterprises corresponding to the high-dimensional new data with negative inner product difference, outputting the similar enterprise recommendation data of the target enterprise after sorting, and realizing the enterprise recommendation processing of the target enterprise.

[0032] Compared with the traditional method, the technical solution adds the declaration conditions possessed by the enterprises obtained through reverse deduction on the network to the feature information of the enterprises, so as to effectively enrich the feature information of the enterprises, and avoids directly calculating the distance from the high-dimensional new data to the optimal hyperplane in the high-dimensional feature space in the data classification and prediction process, but directly uses the support vector to simplify the classification calculation, so as to finally realize fast and accurate enterprise recommendation output, and achieve the purpose of greatly improving the accuracy of enterprise recommendation. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0034] Figure 1 A flowchart of an embodiment of the enterprise recommendation method based on big data;

[0035] Figure 2 A flowchart of an embodiment of the enterprise recommendation method based on big data;

[0036] Figure 3 A flowchart of an embodiment of the enterprise recommendation method based on big data;

[0037] Figure 4 A flowchart of an embodiment of the enterprise recommendation method based on big data;

[0038] Figure 5 A module structure diagram of an embodiment of the enterprise recommendation device based on big data. DETAILED DESCRIPTION

[0039] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0040] Unless otherwise defined, 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 application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0041] It should be noted that a reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another.

[0042] Those skilled in the art can understand that the embodiments described herein can be combined with other embodiments. The term "and / or" used in the specification and claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0043] In the research design of the present application, the inventors found that various organization websites publish policy documents and relevant public documents of the policy supported by the enterprise reporting policy. By analyzing the policy documents and the relevant public documents of the policy, the enterprise obtained policy projects can be obtained (the principle is that the policy documents can analyze all relevant data of the reporting conditions; the relevant public documents corresponding to the policy documents can analyze the list of enterprises supported by the policy projects). By judging the requirements of the policy projects, the conditions possessed by the enterprises obtaining the policy projects can be deduced in reverse. Adding the conditions possessed by the enterprises to the characteristic information of the enterprises can effectively enrich the characteristic information of the enterprises and be used to effectively improve the accuracy of enterprise recommendation. For example:

[0044] 1. A enterprise obtains the "national enterprise technology center" recognition in the relevant public documents on the website.

[0045] 2. Analyzing the policy reporting conditions of "national enterprise technology center" shows that the R&D expenses of the previous year are greater than or equal to 15 million, the number of employees is greater than or equal to 300, etc.

[0046] 3. Therefore, it is deduced that the R&D expenses of A enterprise in the previous year are greater than or equal to 15 million, the number of employees is greater than or equal to 300, etc.

[0047] 4. More dimensional characteristic information is obtained by the third step, so that the enterprise portrait is enriched, and the accuracy of enterprise recommendation is improved.

[0048] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0049] Please refer to Figure 1 In one embodiment, a big data-based enterprise recommendation method is provided, including the following processing steps S11-S16:

[0050] S11, obtaining public internet data corresponding to the name of the target enterprise and storing it in a first dimension array.

[0051] It can be understood that the target enterprise can be, but is not limited to, an enterprise that the user currently focuses on, wants to find or search for, or a to-be-analyzed enterprise that wants to know which policy projects can be declared. The public internet data refers to the enterprise-related public information of the target enterprise publicly disclosed on the internet, such as but not limited to enterprise business registration information, industry information, region information, enterprise insurance number information, and establishment time, which can be used to generate the enterprise portrait of the target enterprise.

[0052] Specifically, the enterprise recommendation device can obtain the public internet data corresponding to the name of the target enterprise through enterprise name search collection, crawler crawling, user input or upload, etc., and then store the obtained public internet data in the first dimension array configured in advance, so that the processing algorithm can be quickly called in the subsequent data calculation, classification processing, etc.

[0053] S12, obtaining each policy file and each related public file published on each official website, parsing each policy file and each related public file, obtaining the public enterprise list and the declaration condition data of each policy project, and storing them in a second dimension array.

[0054] It can be understood that the each policy file and each related public file can be the policy file and the related public file of the policy supported by the enterprise for declaration published on the official website of each department of each region. The enterprise recommendation device can parse the policy file and the related public file of the policy to obtain the policy project obtained by the enterprise in the policy formula file. Moreover, the enterprise policy device judges the declaration condition of the policy project, and can reversely deduce the actual condition possessed by the enterprise obtaining the policy project. The actual condition possessed by the enterprise is added to the characteristic information of the enterprise, which can more effectively enrich the characteristic information of the enterprise, so as to effectively guarantee the accuracy of the enterprise recommendation.

[0055] Specifically, the enterprise recommendation device can, but is not limited to, acquire each policy file and each related public file published on each website through website search collection, crawler crawling, user input or upload and the like, then parse the acquired files, and can store the obtained enterprise list and the corresponding declaration condition data of each policy item into a pre-configured second dimension array, so that the processing algorithm can be quickly called in subsequent data calculation, classification processing and the like.

[0056] S13, performing data normalization processing on the original data in the first dimension array and the second dimension array to obtain a normalized low-dimensional feature vector and map to a high-dimensional feature space based on a support vector machine using a kernel function.

[0057] It can be understood that data normalization is a process of scaling data to a specific range, and the purpose is to eliminate the dimensional differences between different features, so that the data can better adapt to the processing of various machine learning algorithms. In the embodiment, data normalization processing is adopted to form each low-dimensional feature vector of each enterprise, each low-dimensional feature vector contains multiple dimensions of features, and each dimension represents a sample feature of data. These low-dimensional feature vectors can be used as the portrait expression of each enterprise. In the field of big data, different data normalization methods are suitable for different data situations. Therefore, according to the type and format of the original data, the data use needs of the machine learning algorithm and the like, a suitable existing data normalization method can be selected to perform the above data normalization processing, which can also improve the accuracy and stability of the machine learning algorithm.

[0058] For nonlinear classification or regression problems in machine learning algorithms such as support vector machines, kernel functions are commonly used function tools for mapping low-dimensional data to high-dimensional feature spaces, so that originally linearly inseparable problems are converted into linearly separable problems. In the embodiment, a kernel function is also used to realize the mapping processing of the low-dimensional feature vector to the high-dimensional feature space.

[0059] S14, in the high-dimensional feature space, performing centering processing on the high-dimensional new data corresponding to the low-dimensional feature vector, and calculating the inner product difference of the centered high-dimensional new data and the support vector of the high-dimensional feature space.

[0060] It can be understood that after the original data is mapped to a new high-dimensional feature space by the kernel function, a linear classifier (such as a support vector machine) can be used to classify the data in the new high-dimensional feature space. In this way, even if the original data is nonlinearly separable in the low-dimensional space, a linear decision boundary can be found in the high-dimensional feature space, that is, an optimal classification hyperplane is constructed to maximize the distance of data points of different categories in the high-dimensional feature space to the hyperplane.

[0061] Specifically, the general process of solving the optimal classification hyperplane in the high-dimensional feature space can be: calculating the mean value of all samples in the feature space; subtracting the mean value from each sample respectively to obtain a centralized sample matrix; solving the covariance matrix of the sample matrix; solving the eigenvectors and eigenvalues of the covariance matrix; selecting the eigenvector with the largest eigenvalue as the hyperplane normal vector; solving the bias parameter according to the distance formula of the sample point to the hyperplane; obtaining the expression of the final optimal classification hyperplane; and classifying the input data samples using the optimal classification hyperplane.

[0062] Then, the inventors found in practical research that in the distance formula of the sample point to the optimal classification hyperplane in the high-dimensional feature space, the norm of the hyperplane normal vector is difficult to calculate; and the support vectors are the data points closest to the optimal classification hyperplane, and their distance to the optimal classification hyperplane is 1. For any input sample point, the distance of the sample point to the optimal classification hyperplane can be simplified as the inner product difference between the sample point and the support vector. Therefore, the support vector can be directly used to simplify the classification calculation process, and direct calculation of the distance of the high-dimensional new data to the optimal classification hyperplane is avoided. In this way, the processing steps of using the support vector machine to classify new data are improved as follows: centralizing the high-dimensional new data, calculating the inner product of the high-dimensional new data and the support vector, calculating the inner product difference and judging the positive and negative of the inner product difference, and classifying the high-dimensional new data according to the positive and negative.

[0063] S15, removing the enterprise corresponding to the high-dimensional new data with the negative inner product difference, and determining the enterprise corresponding to the high-dimensional new data with the positive inner product difference as the similar enterprise.

[0064] S16, sorting the similar enterprises in descending order of the inner product difference, and outputting the similar enterprise recommendation data of the target enterprise.

[0065] Specifically, the corresponding enterprise with a negative score (i.e., inner product difference) is removed from the queue of similar enterprises, and the positive sample (i.e., the corresponding enterprise with a positive inner product difference) is sorted in descending order of the score value to obtain the similar enterprises of the target enterprise and the similarity of each similar enterprise, which can be used as enterprise recommendation data output. The output data format can be, but is not limited to, a data table, a similar enterprise map or other forms of documents, as long as it is convenient for users to use.

[0066] The above enterprise recommendation method based on big data first acquires public Internet data of a target enterprise, then acquires various policy files and related public files of an industry to which the target enterprise belongs, further analyzes each enterprise of the industry and forms an enterprise list, simultaneously acquires data of declaration conditions met by each enterprise in the enterprise list, then performs data normalization processing on all original data acquired in the foregoing, forms low-dimensional feature vector data and maps to a high-dimensional feature space based on a support vector machine by using a kernel function, directly calculates an inner product difference of the high-dimensional new data after centering and the support vector of the high-dimensional feature space in the high-dimensional feature space, eliminates an enterprise corresponding to the high-dimensional new data with a negative inner product difference, outputs similar enterprise recommendation data of the target enterprise after sorting, and realizes enterprise recommendation processing about the target enterprise.

[0067] Compared with a traditional method, the above technical solution adds declaration conditions possessed by an enterprise obtained through reverse deduction on the Internet to feature information of the enterprise, so as to effectively enrich the feature information of the enterprise, and avoids directly calculating a distance from high-dimensional new data to an optimal hyperplane in a high-dimensional feature space in a data classification and prediction process, but directly uses a support vector to simplify classification calculation, so as to finally realize fast and accurate enterprise recommendation output, and achieve the purpose of greatly improving enterprise recommendation accuracy.

[0068] In an embodiment, the public Internet data includes an industry to which an enterprise belongs, a time length of establishment, a number of employees, a taxpayer qualification, registered capital, actually paid-in capital, enterprise nature, a registration address, a number of software copyrights, a number of trademarks, a number of certificates, a number of work copyrights, a number of utility model patents, a number of invention authorization, a number of appearance design, a number of invention publication, a listing type dimension, and bidding data. Therefore, the low-dimensional feature vector can also correspond to data of each year of the enterprise, for example, can include the number of employees, the registered capital, the actually paid-in capital, the registration address, the number of software copyrights, the number of trademarks, the number of certificates, the number of work copyrights, the number of utility model patents, the number of invention authorization, the number of appearance design, and the number of invention publication, and the like. By using these basic Internet data and adding declaration condition data possessed by the enterprise, the enterprise portrait can be effectively enriched, and the accuracy of enterprise recommendation can be improved.

[0069] In an embodiment, a normalization method used in the data normalization processing process can include minimum-maximum normalization, Z-score standardization, mean-variance normalization, decimal scaling standardization, or Log function conversion.

[0070] Specifically, different normalization methods are suitable for different data situations, and selecting the appropriate normalization method can improve the accuracy and stability of machine learning algorithms. Among them, Min-Max Scaling: also known as deviation standardization, is used to linearly map the original data to the interval [0, 1]. Z-score standardization: used to standardize data according to its mean and standard deviation, so that the data conforms to the standard normal distribution. Decimal scaling standardization: used to scale data by moving the decimal point position, so that the absolute value of the data is less than 1. Log function conversion: for features with large data values and large variation ranges, log transformation can be performed to make the data more conform to the normal distribution. Mean-variance normalization (Standard Scaling): subtract the mean from the data and then divide by the standard deviation, so that the data conforms to the standard normal distribution.

[0071] Taking Min-Max Scaling as an example: take the maximum value max and the minimum value min of the feature data of each day in the above two arrays; for each feature data x, use the following formula for normalization processing:

[0072]

[0073] After all features are normalized, a new normalized array is obtained for subsequent processing.

[0074] For example, mean-variance normalization, the normalization calculation formula is as follows:

[0075]

[0076] Where x represents the original data, μ represents the mean of the original data, σ represents the standard deviation of the original data, and z represents the converted data. The specific standardization process can be divided into the following steps: calculate the mean and standard deviation of the original data; standardize each original data, including subtracting the mean from the original value to get the difference; divide the difference by the standard deviation to get the standardized value; after obtaining the standardized values of all data, they can be used for subsequent classification, clustering, and regression analysis processing. Take the number of employees and registered capital as an example, take the data of six enterprises over the years, and convert them into the following enterprise feature matrix [[100 5080 200 90 150],[150 80 120 400 70 200]], and then perform normalization processing through the above algorithm.

[0077] The specific application of other normalization methods can also be understood by referring to their own processing flow. By using the above data normalization processing method, the normalization processing of the original data can be efficiently and accurately completed.

[0078] In one embodiment, the kernel function is a RBF kernel function. It can be appreciated that the RBF (Radial Basis Function) kernel function is a commonly used kernel function for non-linear classification or regression problems in machine learning algorithms such as SVM (Support Vector Machine). The RBF kernel function is defined as follows: for two samples x i and x j in the input space, the output value of the RBF kernel function is:

[0079]

[0080] where σ is the bandwidth of the RBF kernel function, x i -x j represents the Euclidean distance between vectors x i and x j , μ represents the mean of the original data, σ represents the standard deviation of the original data, and z represents the transformed data. The RBF kernel function has the characteristic of mapping the original input space to a high-dimensional feature space, thereby making the originally linearly inseparable problem linearly separable. At the same time, due to the self-similarity of the RBF kernel function (i.e. each sample point has the greatest similarity with itself), it has good effect on problems with self-similarity such as spatio-temporal data.

[0081] When using the RBF kernel function, the size of the bandwidth parameter σ needs to be adjusted, and cross-validation and other methods are usually used to select it. If σ is large, the output value of the kernel function changes relatively smoothly with distance, and has weak localization characteristics; if σ is small, the output value of the kernel function changes relatively sensitively with distance, and has strong localization characteristics. Therefore, it can be flexibly selected according to actual use needs.

[0082] Specifically, an example of spatial mapping using the RBF kernel function is as follows: in this example, there is a two-dimensional sample data set X and its corresponding classification label Y. For the data set X in this example, the kernel matrix between each pair of samples is calculated. First, a suitable σ value needs to be selected, and different σ values can be tried, for example σ = 1. Then, the RBF kernel matrix is calculated, and the obtained RBF kernel matrix is as follows: [[1.0.60653066 0.60653066 0.00033546 0.00012341] [0.60653066 1.0.13533528 0.00183156 0.00067067] [0.60653066 0.13533528 1.0.00091188 0.00033546] [0.00033546 0.00183156 0.00091188 1.0.60653066] [0.00012341 0.00067067 0.00033546 0.60653066 1.]]

[0088] In the same way, until the completion of the mapping of all data processing, that is, the original data through the RBF kernel function is quickly mapped to a new high-dimensional feature space.

[0089] Further, after the completion of the low-dimensional feature vector mapping to the high-dimensional feature space based on the support vector machine, the optimal classification hyperplane can be solved, for example:

[0090] Calculate the mean value of all samples in the high-dimensional feature space, denoted as μ; the above RBF kernel matrix is taken as an example (the same as in this embodiment), then μ = [0.30326 0.30326 0.30326 0.30326 0.30326].

[0091] Subtract μ from each sample to obtain the centralized sample matrix X;

[0092] X = [[0.69697399 0.30182766 0.30182766 -0.29990674 -0.28989119] [0.30182766 0.69697399 0.05683738 0.116866 0.42786001]

[0094] [0.30182766 0.05683738 0.69697399 -0.28801512 -0.28989119]

[0095] [-0.29990674 0.116866 -0.28801512 0.69697399 0.30182766]

[0096] [-0.28989119 0.42786001 -0.28989119 0.30182766 0.69697399]]

[0097] Calculate the covariance matrix XX' of the sample matrix X;

[0098] XX' = [[0.48358984 0.17124736 0.17124736 0.0. [0.17124736 0.48358984 0.02399147 0.04655467 0.18360223] [0.17124736 0.02399147 0.48358984 0.0. [0.0.04655467 0.0.48358984 0.17124736] [0.0.18360223 0.0.17124736 0.48358984]]

[0103] Find the eigenvectors and eigenvalues of the covariance matrix XX'. Note the sign of the eigenvectors.

[0104] Eigenvectors:

[0105] [-0.40824829 -0.40824829 0.81649658 0.0.

[0106] [-0.40824829 0.81649658 0.40824829 0.0.

[0107] [-0.40824829 0.40824829 -0.81649658 0.0. [0.0.0.0.1. [0.0.0.0.1.

[0110] Eigenvalues: [2.41597182 1.41597182 1.41597182 0.48358984 0.48358984]

[0112] Choose the eigenvector with the largest eigenvalue as the hyperplane normal vector w.

[0113] w = [-0.40824829 -0.40824829 0.81649658]

[0114] According to the formula for the distance from a sample point x to the hyperplane Let d = 1, find the bias parameter b. For example, b = -1.2267573. Here, w T represents the L2 norm of the vector w.

[0115] Finally, the optimal classification hyperplane expression is:

[0116] -0.40824829x1 -0.40824829x2 + 0.81649658x3 -1.2267573 = 0

[0117] where x1, x2 and x3 represent three sample points on the optimal classification hyperplane respectively.

[0118] After obtaining the optimal classification hyperplane, it can be used to classify new data, for example:

[0119] 1. Center the new data sample x, subtract the mean μ, to obtain the centered sample x'.

[0120] 2. Calculate the inner product of x' and the hyperplane normal vector w, plus the bias parameter b, to obtain the formula for the distance of x' to the optimal classification hyperplane:

[0121] 3. If d > 0, the new data sample is located in the positive direction of the hyperplane, and is classified as a positive sample; if d < 0, the new data sample is located in the negative direction of the hyperplane, and is classified as a negative sample.

[0122] 4. Since in high-dimensional feature space, the norm ||w|| of the hyperplane normal vector w in the hyperplane distance formula is difficult to calculate, support vectors can be used to simplify the calculation. Support vectors are data points closest to the hyperplane, with a distance of 1 to the hyperplane. Then for any new data x', its distance d to the hyperplane can be expressed as:

[0123] d = (w T x') - (w T x0)

[0124] where x0 is the support vector.

[0125] 5. Therefore, directly calculate the difference between (w T x') and (w T x0), if the difference > 0, the new data is a positive sample, otherwise it is a negative sample.

[0126] 6. Finally, since w is the hyperplane normal vector, then (w T x0) is the distance of the support vector to the hyperplane, which can be calculated according to the hyperplane equation.

[0127] Finally, the steps of using support vector machine to classify new data can be changed to: 1) center the new data; 2) calculate the inner product of the new data and the support vector; 3) calculate the difference of the inner product and judge the positive and negative of the difference; 4) classify the new data according to the positive and negative. This avoids directly calculating the distance of the new data to the hyperplane, simplifying the classification calculation process. As Figure 2 shown, is an implementation process diagram of applying the above enterprise recommendation method, wherein the similarity threshold can be selected as 0.

[0128] In another embodiment, the kernel function can also be selected as a linear kernel function, a polynomial kernel function, a sigmoid kernel function, or a Laplace kernel function. It can be understood that in addition to the RBF kernel function, there are other kernel functions that can be used for nonlinear classification or regression problems in machine learning algorithms such as support vector machines, some of which are more suitable for specific data sets or tasks than the RBF kernel function, such as the following commonly used kernel functions:

[0129] Linear Kernel: The inner product of two vectors in the input space is calculated, which is suitable for linearly separable or approximately linearly separable data.

[0130] Polynomial Kernel: The inner product of two vectors in the input space is calculated after polynomial expansion, which is suitable for data with certain nonlinear characteristics.

[0131] Sigmoid Kernel: The inner product of two vectors in the input space is calculated after Sigmoid function transformation, which is suitable for data sets with symmetrically distributed features.

[0132] Laplace Kernel: The inner product of two vectors in the input space is calculated after the Euclidean distance between them is substituted into the Laplace function, which is suitable for data sets with noise or outliers.

[0133] In addition to the above several kernel functions, some other kernel functions can also be used, such as ANOVA kernel function and Bessel kernel function, etc. In practical applications, the most suitable kernel function can be selected according to the characteristics of the data set and the requirements of the task, so as to achieve better processing effect.

[0134] In one embodiment, as shown in Figure 3 the above method can further include the following processing step S17:

[0135] S17, using a collaborative filtering algorithm to output policy project data of the matching target enterprise according to the similar enterprise recommendation data.

[0136] It can be understood that after obtaining the similar enterprise recommendation data of the target enterprise, the enterprise recommendation device can also be instructed to use a collaborative filtering algorithm to predict the policy items that the target enterprise may be eligible for according to the similar enterprise recommendation data, filter out the items that the target enterprise has declared, obtain the policy items that the target enterprise may be eligible for and has not declared, and match the policy items to the target enterprise, output the policy item data of the target enterprise, that is, directly recommend the policy items that the target enterprise is also suitable for declaring to the user, so as to further improve the application range of enterprise recommendation.

[0137] Specifically, according to the similar enterprise recommendation data (which can be in the form of a similar enterprise list), the policy items obtained by all similar enterprises are obtained, for example, a matrix can be used to represent, where the row represents the similar enterprise, the column represents the policy item, and each element of the matrix represents a binary variable (0 represents not obtained, 1 represents obtained) whether to obtain the item.

[0138] Item prediction: According to the statistical data of the policy items obtained by similar enterprises, the policy items that the target enterprise may be eligible for are predicted. The specific method is to find the top K (which can be flexibly valued according to actual needs) similar enterprises with the highest similarity to the target enterprise, then calculate the weighted average similarity of them to the target enterprise, and weight the statistical data of the policy items corresponding to the similar enterprises according to the size of the weighted average similarity. Finally, filter out the items that the target enterprise has declared, and output the predicted eligible undeclared items. Here, existing collaborative filtering algorithms can be used to achieve, for example:

[0139] Suppose there are three similar enterprises B, C and D, and their obtained policy item data is as follows:

[0140] Item B C D A 1 0 0 B 1 1 0 C 0 1 1 D 0 1 0

[0141] The similarity of the target enterprise A to the similar enterprises is 0.9, 0.6 and 0.8 respectively, and the predicted data of the target enterprise is obtained by weighted average according to the similarity:

[0142] Item Prediction result A 0.84 B 0.6 C 0.72 D 0.18

[0143] Since the target enterprise A has obtained items B and C, these items are filtered out, and the final output of the predicted eligible undeclared items is A and D.

[0144] In an embodiment, as shown in Figure 4 the above method can further include the following processing step S18:

[0145] S18, generating a promotion script matched to each similar enterprise in the similar enterprise recommendation data according to the promotion script of the target enterprise.

[0146] It is understandable that after obtaining similar enterprise recommendation data for the target enterprise, the enterprise recommendation device can be instructed to generate recommendation texts corresponding to each similar enterprise based on the pre-made recommendation texts for the target enterprise, such as targeted marketing promotion texts. This results in recommendation texts that can also be directly applied to these similar enterprises, thereby further expanding the applicability of enterprise recommendations and improving the execution efficiency and accuracy of automated marketing tasks.

[0147] Furthermore, based on the already obtained list of similar companies recommended to the target companies, intelligent marketing tasks can be carried out. For example, personalized marketing plans can be generated based on the list of projects that similar companies have not yet applied for, to encourage similar companies to apply for these projects and increase their success rate. For instance, marketing plans can be sent to similar companies via email or SMS, providing market analysis intelligence data and planning suggestions related to the projects, introducing successful application case data and experience information, or the organizer can offer similar companies some preferential measures to attract them to apply for these projects. Alternatively, existing intelligent marketing platforms can be used to conduct in-depth analysis of the aforementioned relevant data of similar companies to understand their customer needs and consumption habits, and based on the analysis results, suggestions for adjusting the marketing strategies of similar companies can be output.

[0148] It should be understood that, although Figures 1 to 4 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order requirement for the execution of these steps; they can be executed in other orders. Figures 1 to 4 At least some of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0149] Please see Figure 5In one embodiment, a big data-based enterprise recommendation device 100 is provided, comprising a first acquisition module 11, a second acquisition module 12, a normalization mapping module 13, a score calculation module 14, an enterprise determination module 15, and a recommendation output module 16. The first acquisition module 11 is configured to acquire public Internet data corresponding to the name of a target enterprise and store the data in a first dimension array. The second acquisition module 12 is configured to acquire each policy document and each related public document published on each official website, parse the policy document and the related public document, obtain a public enterprise list and each policy item declaration condition data, and store the data in a second dimension array. The normalization mapping module 13 is configured to perform data normalization processing on the original data in the first dimension array and the second dimension array, obtain a normalized low-dimensional feature vector, and map the low-dimensional feature vector to a high-dimensional feature space based on a support vector machine using a kernel function. The score calculation module 14 is configured to perform centering processing on the high-dimensional new data corresponding to the low-dimensional feature vector in the high-dimensional feature space, and calculate the inner product difference between the centered high-dimensional new data and the support vector of the high-dimensional feature space. The enterprise determination module 15 is configured to exclude the enterprise corresponding to the high-dimensional new data with a negative inner product difference, and determine the enterprise corresponding to the high-dimensional new data with a positive inner product difference as a similar enterprise. The recommendation output module 16 is configured to sort the similar enterprises in descending order of the inner product difference, and output similar enterprise recommendation data of the target enterprise.

[0150] The above-described big data-based enterprise recommendation device 100 acquires public Internet data of a target enterprise, then acquires various policy documents and related public documents of an industry to which the target enterprise belongs, parses each enterprise in the industry to form an enterprise list, acquires declaration condition data of each enterprise in the enterprise list, performs data normalization processing on all the original data, forms low-dimensional feature vector data, maps the low-dimensional feature vector data to a high-dimensional feature space based on a support vector machine using a kernel function, directly calculates the inner product difference between the centered high-dimensional new data and the support vector of the high-dimensional feature space in the high-dimensional feature space, excludes the enterprise corresponding to the high-dimensional new data with a negative inner product difference, and outputs similar enterprise recommendation data of the target enterprise in a sorted order, thereby achieving enterprise recommendation processing on the target enterprise.

[0151] Compared with the traditional method, the above technical solution adds the declaration conditions possessed by the enterprise obtained through reverse deduction on the Internet to the feature information of the enterprise, thereby effectively enriching the feature information of the enterprise. In addition, in the data classification and prediction process, the distance from the high-dimensional new data to the optimal hyperplane in the high-dimensional feature space is directly calculated, and the support vector is directly used to simplify the classification calculation, thereby finally achieving fast and accurate enterprise recommendation output, and achieving the purpose of greatly improving the accuracy of enterprise recommendation.

[0152] In one embodiment, the disclosed Internet data includes the industry to which the enterprise belongs, the establishment time, the number of employees, the taxpayer qualification, the registered capital, the actually paid-in capital, the nature of the enterprise, the registered address, the number of software copyrights, the number of trademarks, the number of certificates, the number of works copyrights, the number of utility model patents, the number of invention patents, the number of design patents, the number of invention publications, the listing type dimensions, and the bidding data.

[0153] In one embodiment, the normalization method used in the data normalization process includes minimum-maximum normalization, Z-score standardization, mean-variance normalization, fractional scaling standardization, or Log function conversion.

[0154] In one embodiment, the kernel function is an RBF kernel function.

[0155] In one embodiment, the above-mentioned big data-based enterprise recommendation device 100 can further include a project matching module for outputting the policy project data of the target enterprise according to the most similar enterprise in the similar enterprise recommendation data.

[0156] In one embodiment, the above-mentioned big data-based enterprise recommendation device 100 further includes a recommendation generation module for generating a recommendation script matched to each similar enterprise in the similar enterprise recommendation data according to the recommendation script of the target enterprise.

[0157] For specific limitations of the big data-based enterprise recommendation device 100, please refer to the corresponding limitations of the big data-based enterprise recommendation method in the above, which will not be repeated here. Each module in the above-mentioned big data-based enterprise recommendation device 100 can be realized by software, hardware, and their combinations in whole or in part. The above-mentioned modules can be embedded in or independent of a device with data processing function in hardware form, or stored in the memory of the aforementioned device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module. The aforementioned device can be, but is not limited to, various types of data computing and processing devices in the art.

[0158] In one embodiment, a computer device is also provided, comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following processing steps when executing the computer program: obtaining public Internet data corresponding to the name of a target enterprise and storing in a first dimension array; obtaining each policy document and each related public document published on each official website, parsing each policy document and each related public document, obtaining the public enterprise list and the declaration condition data of each policy item, and storing in a second dimension array; performing data normalization processing on the original data in the first dimension array and the second dimension array, obtaining a normalized low-dimensional feature vector, and mapping to a high-dimensional feature space based on a support vector machine using a kernel function; in the high-dimensional feature space, performing centering processing on the high-dimensional new data corresponding to the low-dimensional feature vector, calculating the inner product difference of the centered high-dimensional new data and the support vector of the high-dimensional feature space; removing the enterprises corresponding to the high-dimensional new data with negative inner product difference, and determining the enterprises corresponding to the high-dimensional new data with positive inner product difference as similar enterprises; sorting the similar enterprises in descending order of the inner product difference, and outputting the similar enterprise recommendation data of the target enterprise.

[0159] It can be understood that, in addition to the memory and the processor mentioned above, the computer device described above also includes other software and hardware components not listed in the specification, which can be determined according to the specific model of the computer device in different application scenarios, and the specification will not be listed in detail.

[0160] In one embodiment, the processor executing the computer program can also implement the steps or sub-steps added in each embodiment of the enterprise recommendation method based on big data described above.

[0161] In one embodiment, a computer readable storage medium is also provided, which stores a computer program, and the computer program is executed by a processor to implement the following processing steps: obtaining public Internet data corresponding to the name of a target enterprise and storing in a first dimension array; obtaining each policy document and each related public document published on each official website, parsing each policy document and each related public document, obtaining the public enterprise list and the declaration condition data of each policy item, and storing in a second dimension array; performing data normalization processing on the original data in the first dimension array and the second dimension array, obtaining a normalized low-dimensional feature vector, and mapping to a high-dimensional feature space based on a support vector machine using a kernel function; in the high-dimensional feature space, performing centering processing on the high-dimensional new data corresponding to the low-dimensional feature vector, calculating the inner product difference of the centered high-dimensional new data and the support vector of the high-dimensional feature space; removing the enterprises corresponding to the high-dimensional new data with negative inner product difference, and determining the enterprises corresponding to the high-dimensional new data with positive inner product difference as similar enterprises; sorting the similar enterprises in descending order of the inner product difference, and outputting the similar enterprise recommendation data of the target enterprise.

[0162] In one embodiment, the computer program, when executed by the processor, further implements the steps or sub-steps added in each of the above-mentioned big data-based enterprise recommendation method embodiments.

[0163] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the processes of the above-mentioned embodiments can be included. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus dynamic random access memory (Rambus DRAM, RDRAM) and interface dynamic random access memory (DRDRAM).

[0164] Each technical feature of the above embodiments can be combined arbitrarily. To make the description concise, each technical feature of the above embodiments is not described in all possible combinations, but as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0165] The above embodiments only express several implementation manners of the present application, and the description is specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present application, some modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A big data-based enterprise recommendation method, characterized in that, Including the following steps: Obtain publicly available internet data corresponding to the name of the target company and store it in the first dimension array; the publicly available internet data includes the company's industry, years of establishment, number of employees, taxpayer qualifications, registered capital, paid-in capital, company nature, registered address, number of software copyrights, number of trademarks, number of certificates, number of works copyrights, number of utility model patents, number of invention patents, number of design patents, number of invention publications, listing type dimension, and bidding data; Obtain the policy documents and related public notices published on the official websites, parse the policy documents and related public notices, obtain the list of publicized enterprises and the application conditions data for each policy project, and then store them in the second dimension array; The original data in the first dimension array and the second dimension array are normalized to obtain normalized low-dimensional feature vectors, which are then mapped to a high-dimensional feature space based on support vector machine using a kernel function. In the high-dimensional feature space, the high-dimensional new data corresponding to the low-dimensional feature vector is centered, and the inner product difference between the centered high-dimensional new data and the support vectors of the high-dimensional feature space is calculated. The companies corresponding to the high-dimensional new data with negative inner product differences are removed, and the companies corresponding to the high-dimensional new data with positive inner product differences are identified as similar companies. The similar companies are sorted in descending order of their inner product differences, and the similar company recommendation data of the target company is output.

2. The enterprise recommendation method based on big data according to claim 1, characterized in that, The normalization methods used in data normalization processing include min-max normalization, Z-score normalization, mean-variance normalization, decimal scaling normalization, or Log function transformation.

3. The enterprise recommendation method based on big data according to claim 2, characterized in that, The kernel function is the RBF kernel function.

4. The enterprise recommendation method based on big data according to claim 2, characterized in that, It also includes the following steps: Based on the similar enterprise recommendation data, a collaborative filtering algorithm is used to output policy project data that matches the target enterprise.

5. The enterprise recommendation method based on big data according to claim 2, characterized in that, It also includes the following steps: Based on the promotional copy of the target company, promotional copy matching each similar company in the similar company recommendation data is generated.

6. A big data-based enterprise recommendation device, characterized in that, include: The first acquisition module is used to acquire publicly available internet data corresponding to the name of the target company and store it in the first dimension array; The publicly available internet data includes the company's industry, years of establishment, number of employees, taxpayer qualifications, registered capital, paid-in capital, company type, registered address, number of software copyrights, number of trademarks, number of certificates, number of copyrights for works, number of utility model patents, number of invention patents granted, number of design patents, number of invention publications, listing type, and bidding data. The second acquisition module is used to acquire the policy documents and related public notices published on the official websites, parse the policy documents and related public notices, obtain the list of publicized enterprises and the application conditions data of each policy project, and then store them in the second dimension array. The normalization mapping module is used to normalize the original data in the first dimension array and the second dimension array to obtain normalized low-dimensional feature vectors and map them to a high-dimensional feature space based on support vector machine using a kernel function. The scoring calculation module is used to center the high-dimensional new data corresponding to the low-dimensional feature vector in the high-dimensional feature space, and calculate the difference of the inner product between the centered high-dimensional new data and the support vector of the high-dimensional feature space. The enterprise identification module is used to remove enterprises corresponding to the high-dimensional new data with negative inner product differences, and to identify enterprises corresponding to the high-dimensional new data with positive inner product differences as similar enterprises. The recommendation output module is used to sort the similar enterprises in descending order of their inner product differences and output the similar enterprise recommendation data of the target enterprise.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the enterprise recommendation method based on big data as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the enterprise recommendation method based on big data as described in any one of claims 1 to 5.

Images