A business recommendation method

Abstract

The application provides an enterprise recommendation method, comprising the following steps: obtaining information data of a target enterprise; performing clustering and grouping processing on sample data to obtain a plurality of sample groups; calculating deviation scores and weight scores of the sample data in each sample group; obtaining data scores of each sample data based on the deviation scores and the weight scores, summarizing all data scores to obtain total score data of the target enterprise; and when a user searches, recommending the corresponding target enterprise according to the ranking of the total score data. The sample data is processed by clustering and grouping through a clustering algorithm, and the index mean of the sample group is calculated. The deviation score is obtained based on the index mean and the index value deviation of each sample data. The weight score of each sample data is calculated based on the weighting method, and the weight score is adjusted through the deviation score to eliminate the influence of data differences, improve the calculation accuracy of the enterprise score ranking, and improve the user experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information data processing, in particular to an enterprise recommendation method. BACKGROUND

[0002] At present, when recommending enterprises to users, it is generally divided into hot enterprise recommendation or enterprise information recommendation based on user search records through collaborative filtering algorithm.

[0003] In the prior art, hot recommendation is mainly based on the case that the user has no search behavior before, and the enterprise is pushed according to the search times of the whole network, but the hot push does not mean that the enterprise is a high-quality enterprise; the traditional recommendation based on user search records cannot obtain ideal recommendation results for users because the influence of the enterprise itself is not considered, and the user experience is poor. SUMMARY

[0004] In view of the defects of the prior art, the purpose of the present application is to provide an enterprise recommendation method, which aims to solve the technical problem of poor user experience in the prior art.

[0005] In order to achieve the above purpose, the present application is realized by the following technical scheme: an enterprise recommendation method, comprising the following steps: obtaining information data of a target enterprise, wherein the information data comprises a plurality of sample data, and the sample data comprises an index value; clustering and grouping the plurality of sample data based on a clustering algorithm to obtain a plurality of sample groups; calculating the index mean value of the sample group, and obtaining the deviation score of the sample data based on the deviation amount of the index mean value and the index value; calculating the weight score of each sample data in the sample group based on a weighting method; obtaining the data score of each sample data based on the deviation score and the weight score, and obtaining the total score data of the target enterprise by summarizing all the data scores; when a user searches, the corresponding target enterprise is recommended according to the ranking of the total score data.

[0006] Compared with the prior art, the present application has the beneficial effects that: the sample data is clustered and grouped by the clustering algorithm, the different index mean values are calculated according to the specific conditions of each sample group, the deviation score is obtained based on the deviation amount of the index mean value and the index value of each sample data, the weight score of each sample data is calculated based on the weighting method, and the weight score is adjusted by the deviation score to eliminate the influence of data difference and improve the calculation accuracy of enterprise score ranking, so as to push high-quality enterprises to users and improve user experience.

[0007] According to one aspect of the above technical scheme, the step of recommending the corresponding target enterprise according to the ranking of the total score data specifically comprises:

[0008] construct an enterprise knowledge graph based on the information data;

[0009] recommend information of the knowledge graph of the target enterprise according to the ranking of the total score data.

[0010] According to an aspect of the above technical solution, when a user searches, the step of recommending the target enterprise according to the ranking of the total score data specifically includes:

[0011] When receiving an uncertain search signal, recommend the enterprise with the highest ranking of the total score data in each industry;

[0012] When receiving an accurate enterprise name, display the knowledge graph of the corresponding enterprise, and recommend a plurality of related enterprises according to the ranking of the total score data.

[0013] According to an aspect of the above technical solution, the step of constructing an enterprise knowledge graph based on the information data specifically includes:

[0014] Crawl annual report data of the target enterprise, and classify and identify structured data and unstructured data in the annual report data;

[0015] Perform knowledge extraction on the annual report data, and the knowledge extraction includes entity extraction, relationship extraction, and attribute extraction;

[0016] Store the structured data in a Neo4j graph database after being coded according to entities, relationships, and attributes;

[0017] Extract entities and relationships from the unstructured data using a natural language processing model, retain new relationships in a preset extracted entity set, and store the results in the Neo4j graph database to obtain the knowledge graph of the target enterprise.

[0018] According to an aspect of the above technical solution, after the step of crawling annual report data of the target enterprise and classifying and identifying structured data and unstructured data in the annual report data, the method further includes:

[0019] Perform structured processing on the unstructured data to obtain structured data with uniform formats.

[0020] According to an aspect of the above technical solution, after the step of obtaining structured data with uniform formats, the method further includes:

[0021] Perform data standardization processing on the sample data according to the following formula:

[0022] ;

[0023] In the formula, m represents the number of the sample data, and n represents the index value. represents a sample value of the i-th row and the j-th column, represents a sample value after normalization processing, represents taking a minimum value in the j-th column, represents taking a maximum value in the j-th column.

[0024] According to an aspect of the above technical solution, the step of performing clustering and grouping processing on the plurality of sample data based on a clustering algorithm to obtain a plurality of sample groups specifically includes:

[0025] The sample is clustered by row using a k-means clustering algorithm, and the original sample group S is divided into t sample groups after clustering, each sample group being marked as S1, S2, …, St. t wherein t is the number of clustering centers.

[0026] According to an aspect of the above technical solution, the step of calculating the mean value of the index of the sample group and obtaining the deviation score of the sample data based on the deviation amount of the mean value of the index and the index value specifically includes:

[0027] The standard deviation of the j-th index in the sample group S is calculated according to the following formula:

[0028]

[0029] wherein k = 1, …, t, is the number of samples in the k-th sample group, i = 1, …, , j = 1, …, n, n representing the index value, is the index value corresponding to the i-th row and the j-th column in the k-th sample group, is the mean value of the j-th column in the k-th sample group;

[0030] The deviation score of the sample data is calculated according to the following formula:

[0031] .

[0032] According to an aspect of the above technical solution, the step of calculating the weight score of each sample data in the sample group based on the weighting method specifically includes:

[0033] The correlation coefficient matrix of the sample group S is calculated according to the following formula:

[0034]

[0035] wherein and ​​​​​​respectively are index values corresponding to any two index pairs;

[0036] The information amount contained in each index in the kth sample group is calculated according to the following formula :

[0037] ;

[0038] In the formula, is a value corresponding to the ith row and jth column in the correlation coefficient matrix ;

[0039] The weight of each index in the kth sample group is calculated according to the following formula ;

[0040] ;

[0041] The CRITIC weight score of each sample data in the kth sample group is calculated according to the following formula:

[0042] .

[0043] According to an aspect of the above technical solution, the step of obtaining the data score of each sample data based on the deviation score and the weight score specifically includes:

[0044] The data score of the sample data is calculated according to the following formula:

[0045] ;

[0046] In the formula, is the data score of the ith sample in the kth sample group. BRIEF DESCRIPTION OF DRAWINGS

[0047] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings, in which:

[0048] Figure 1 is a flow chart of the enterprise recommendation method in the first embodiment of the present application;

[0049] Figure 2 is a flow chart of the enterprise recommendation method in the second embodiment of the present application;

[0050] The following detailed description will further describe the present application with reference to the above drawings. DETAILED DESCRIPTION

[0051] For the purpose of promoting the understanding of the present application, a more complete description of the application will be rendered by reference to specific embodiments thereof which are depicted in the accompanying drawings. These embodiments are non-limiting examples of the present application. Indeed, the present application can be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0052] It is to be understood that where the terms such as "on", "connected to", "coupled to" or "positioned on" are used herein, it is intended to mean that the element can be directly on another element or intervening elements can also be present. As used herein, the terms "vertical", "horizontal", "left", "right" and the like are merely used for the purpose of illustration and description.

[0053] 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. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0054] Referring to FIG. 1, a flowchart of a business recommendation method according to a first embodiment of the present application is shown, which includes the following steps: Figure 1

[0055] In step S100, information data of a target business is obtained, wherein the information data includes a plurality of sample data, and the sample data includes index values. Specifically, in this step, the information data includes annual report data and the like. The annual report is a collection of various operating information and accounting information, and contains data such as business basic information, accounting data, shareholder changes and shareholder information, which can comprehensively reflect the overall situation of the company. The annual report data is generally saved in a format such as pdf. The annual report data of a specified k (k≥1) years is crawled and saved in a specified directory. The sample data refers to all data used for calculation. One sample data usually refers to one piece of data, such as "Enterprise 1, reporting year, net profit, operating income". The index values correspond to the names in the data. In the example, the index value n=4, i.e. the index value of the sample data is 4.

[0056] In step S110, a clustering algorithm is used to cluster and group the plurality of sample data, and a plurality of sample groups are obtained. Specifically, in this step, the k-means clustering algorithm is used to cluster the sample data by rows, and the elbow method is used to determine the number of clustering centers.

[0057] ​Step S120, calculating the index mean value of the sample group, and obtaining the deviation score of the sample data based on the deviation amount of the index mean value and the index value. In this step, first, the data is clustered and grouped, and the index mean value and the deviation amount of the index value in each sample group are calculated, and the deviation score of each sample data is obtained based on the index mean value and the deviation amount.

[0058] Step S130, calculating the weight score of each sample data in the sample group based on the weighting method. Specifically, in this step, the weighting method adopts objective weighting method, which can adopt entropy weight method, CRITIC weight method, information weight method or independence weight method, etc., and the weight score of the sample data is calculated based on the weight value and the index value of the sample data.

[0059] Step S140, obtaining the data score of each sample data based on the deviation score and the weight score, and summarizing all the data scores to obtain the total score data of the target enterprise.

[0060] Step S150, when the user searches, the corresponding target enterprise is recommended according to the ranking of the total score data.

[0061] In summary, the enterprise recommendation method in the above embodiment of the present application, the sample data is clustered and grouped by the clustering algorithm, and different index mean values are calculated according to the specific conditions of each sample group, the deviation score of each sample data is obtained based on the deviation amount of the index mean value and the index value, the weight score of each sample data is calculated based on the weighting method, and the deviation score is adjusted to the weight score, the influence of data difference is eliminated, the calculation accuracy of enterprise score ranking is improved, and high-quality enterprises are pushed to the user to improve the user experience.

[0062] Please refer to Figure 2 , which is the flow chart of the enterprise recommendation method in the second embodiment of the present application, including the following steps:

[0063] Step S200, the annual report data of the target enterprise is crawled, and the structured data and unstructured data in the annual report data are classified and identified. Understandably, in this step, before crawling the annual report data of the target enterprise, the list of target enterprises that need to be stored in the REDIS database is first needed, then the multi-threaded crawler technology is used to obtain the enterprise name in REDIS that needs to be crawled, the related data of the target enterprise in the specified website is crawled, the corresponding enterprise name in the REDIS database is deleted after each batch of enterprise data is crawled, and the crawling is stopped until the REDIS database is empty, and the enterprise data after the crawling is stored in the specified enterprise database. In addition, since the content and format of the annual report data of the listed enterprise are unified, for the above-mentioned pdf format annual report data of the enterprise, a parsing program is written by using python. For example, financial data and enterprise employee data are stored in tables, and financial, employee and other labels can be preset, and the corresponding data is parsed according to the labels. The parsed data is divided into structured data and unstructured data according to the type. The structured data such as enterprise basic information and financial information is stored in a relational database table, and the semi-structured or unstructured data such as enterprise news public opinion and commodity evaluation information is stored in a non-relational database such as MongoDB, for subsequent calculation and analysis. The above-mentioned single structured data and unstructured data are sample data, and each sample data includes an index value corresponding thereto.

[0064] Step S210, the unstructured data is structured to obtain structured data with uniform format. The format of the original data can be many, such as table data and text data, and we need to integrate different types of data into a uniform format, like a two-dimensional table, which is data structuring. Specifically, when the data itself is structured type data, only simple preprocessing is needed, including filling of missing data, correction of error data, etc.; when the data is unstructured type data, such as news public opinion text data, the data is classified (a small part of sample data can be labeled first, and then the model is trained) by using a classification algorithm for prediction, and finally the number of each type of sample after the statistics is used as a feature of the enterprise.

[0065] Preferably, in this step, in order to facilitate subsequent clustering processing, the sample data needs to be standardized (data normalization processing). The above step S210 further includes the following steps:

[0066] Step S211, the sample data is standardized according to the following formula:

[0067] ;

[0068] In the formula, m represents the number of sample data, n represents the index value, indicates the sample value of the i-th row and the j-th column. denotes the standardized sample value, denotes taking the minimum value in the jth column, denotes taking the maximum value in the jth column. The purpose of data normalization processing is to eliminate the influence of dimension. Some index values are particularly large (such as 100000), and some index values are particularly small (such as 10), which differ by 10000 times. If the values are not normalized, the index with a small value is easily excluded in model calculation.

[0069] Step S220, clustering and grouping processing is performed on a plurality of sample data based on a clustering algorithm, and a plurality of sample groups are obtained. In the embodiment, the above step S220 specifically includes the following steps:

[0070] Step S221, using a k-means clustering algorithm, the sample is clustered by row, and the original sample group S is divided into t sample groups after clustering, and each sample group is marked as S1, S2, …, St. t , wherein t is the number of clustering centers.

[0071] It can be understood that the prior art technology based on knowledge graph calculation of importance (score) is to uniformly calculate the subjective and objective weight of all sample indexes, and then recommend to the user according to the importance score. This calculation method does not consider the difference between different sample groups, but in fact, the essence of recommendation is to classify the population and then recommend to different groups. The embodiment classifies the sample by row using a clustering algorithm, and divides the sample into different groups; then, the improved score calculation formula is used for calculation in different groups, which can reduce the difference between different sample groups.

[0072] Step S230, calculating the mean value of the index of the sample group, and obtaining the deviation score of the sample data based on the deviation amount of the index mean value and the index value. It can be understood that the above index mean value is the sum of the index values in the sample group / sample number. In the embodiment, the above step S230 specifically includes the following steps:

[0073] Step S231, the standard deviation of the jth index in the kth sample group is calculated according to the following formula:

[0074]

[0075] In the formula, k = 1, …, t, is the number of samples in the kth sample group, i = 1, …, , j = 1, …, n, n represents the index value, is the index value corresponding to the ith row and jth column in the kth sample group, ​​​the mean value of the index of the jth column in the kth sample group;

[0076] Step S232, calculating the deviation score of the sample data according to the following formula:

[0077] .

[0078] Step S240, calculating the weight score of each sample data in the sample group based on the weighting method. In the embodiment, the above step S240 specifically includes the following steps:

[0079] Step S241, calculating the correlation coefficient matrix of the sample group according to the following formula: :

[0080] ;

[0081] wherein, xij and xji are the index values corresponding to any two columns of indexes;

[0082] Step S242, calculating the information amount contained in each index in the kth sample group according to the following formula: :

[0083] ;

[0084] wherein, xij and xji are the index values corresponding to any two columns of indexes;

[0085] Step S243, calculating the weight of each index in the kth sample group according to the following formula: ;

[0086] ;

[0087] Step S244, calculating the CRITIC weight score of each sample data in the kth sample group according to the following formula:

[0088] .

[0089] Step S250, obtaining the data score of each sample data based on the deviation score and the weight score. Specifically, in the embodiment, the above step S250 specifically includes the following steps:

[0090] Step S251, calculating the data score of the sample data according to the following formula:

[0091] ;

[0092] wherein,​​​​ Score the data of the i-th sample in the k-th class sample group.

[0093] In addition, in some application scenarios of the embodiment, two data sets S1 and S2 are formed by respectively using the score results calculated based on the CRITIC weight and the score results calculated based on the scheme, the scores in S1 and S2 are used as the prediction values of the model, S1 and S2 are respectively divided into two parts according to a ratio of 7:3 to be used as training and prediction, a regression algorithm is used to calculate the data, and the mean square error of the model is used as the judgment standard. The result shows that the mean square error of the model using the scheme is lower.

[0094] In step S260, knowledge extraction is performed on the annual report data, and the knowledge extraction includes entity extraction, relation extraction and attribute extraction.

[0095] In step S270, the structured data is stored in a Neo4j graph database after being encoded according to entities, relations and attributes; the unstructured data is used to extract entities and relations by using a natural language processing model, new relations in a preset entity set are retained, and the result is stored in the Neo4j graph database to obtain a knowledge graph of the target enterprise. Specifically, in this step, in the graph database, entities are stored as points of the graph database, relations between entities are stored as edges, and attributes are stored as attributes of the points.

[0096] In step S280, when an uncertain search signal is received, the enterprise with the highest total score data ranking in each industry is recommended. Specifically, in this step, the recommended information of the enterprise with the highest score includes basic information stored in the enterprise knowledge graph and processed text statistical information.

[0097] In step S290, when an accurate enterprise name is received, the knowledge graph of the corresponding enterprise is displayed, and a plurality of related enterprises are recommended according to the ranking of the total score data.

[0098] In summary, the enterprise recommendation method in the above embodiment can cluster and group the sample data by using a clustering algorithm, calculate different index means according to the specific conditions of each sample group, obtain a deviation score based on the deviation of the index mean and the index value of each sample data, adjust the weight score through the deviation score, eliminate the influence of data difference, improve the calculation accuracy of enterprise score ranking, push high-quality enterprises to users, recommend enterprises to users from the perspective of enterprise comprehensive index, enrich the amount of user-related enterprise data, and also provide high-quality enterprises in the industry to users without certainty search, reduce search time and improve search efficiency.

[0099] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0100] The above-described embodiments only express several implementation manners of the present application, which are described in a more specific and detailed manner, but cannot be understood as a limitation on the patent scope of the present application. It should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the concept of the present application, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for recommending businesses, characterized in that, Includes the following steps: Obtain information data of the target company, wherein the information data includes several sample data, and the sample data includes indicator values; The sample data are clustered and grouped based on a clustering algorithm to obtain several sample groups; Calculate the mean of the indicators for the sample group, and obtain the deviation score of the sample data based on the deviation between the mean of the indicators and the indicator values. The weighted score of each sample data in the sample group is calculated based on the weighting method; Based on the deviation score and the weight score, a data score is obtained for each sample data. All data scores are then aggregated to obtain the total score data for the target company. When a user searches, the system recommends the corresponding target company based on the ranking of the total score data. The step of recommending corresponding target companies based on the ranking of the total score data specifically includes: Construct an enterprise knowledge graph based on the aforementioned information data; Based on the ranking of the total score data, the corresponding knowledge graph information of the target enterprise is recommended; When a user performs a search, the steps of recommending corresponding target companies based on the ranking of the total score data specifically include: When no definite search signal is received, the company with the highest total score in each industry is recommended. When an accurate company name is received, the corresponding company's knowledge graph is displayed, and several related companies are recommended according to the ranking of the total score data. The specific steps of clustering several sample data based on a clustering algorithm to obtain several sample groups include: The k-means clustering algorithm is used to cluster the samples row by row, dividing the original sample group S into t sample groups, each labeled as S1, S2, ..., Sn. t , where t is the number of cluster centers; The steps of calculating the mean of the indicators for the sample group and obtaining the deviation score of the sample data based on the deviation between the mean of the indicators and the indicator values ​​specifically include: The sample ensemble is calculated using the following formula. The standard deviation of the j-th indicator : ； In the formula, k = 1, ..., t, Let i be the number of samples in the k-th class, i = 1, ..., j = 1, ..., n, where n represents the number of index values. Let j be the index value of the i-th sample in the k-th sample group. Let be the mean value of the j-th indicator in the k-th sample group; The deviation score of the sample data is calculated using the following formula: ； The steps for obtaining a data score for each sample data based on the deviation score and the weight score specifically include: The data score for the sample data is calculated using the following formula: ； In the formula, Score the data of the i-th sample in the k-th class sample group. The sample data is weighted and scored.

2. The enterprise recommendation method according to claim 1, characterized in that, The steps for constructing an enterprise knowledge graph based on the information data specifically include: The system crawls the annual report data of the target company and classifies and identifies the structured and unstructured data in the annual report data. Knowledge extraction is performed on the annual report data, including entity extraction, relation extraction, and attribute extraction. The structured data is encoded by entity, relation, and attribute and then stored in the Neo4j graph database; The unstructured data is used to extract entities and relationships using a natural language processing model. Newly added relationships are retained in a preset set of extracted entities, and the results are stored in the Neo4j graph database to obtain the knowledge graph of the target enterprise.

3. The enterprise recommendation method according to claim 2, characterized in that, After crawling the annual report data of the target company and classifying and identifying the structured and unstructured data in the annual report data, the method further includes: The unstructured data is processed to obtain structured data with a uniform format.

4. The enterprise recommendation method according to claim 3, characterized in that, After the step of obtaining structured data with a uniform format, the method further includes: The sample data is standardized according to the following formula: ； In the formula, m represents the number of sample data, and n represents the index value. This represents the sample value in the i-th row and j-th column. This represents the standardized sample values. This means taking the minimum value in the j-th column. This indicates taking the maximum value in the j-th column.

5. The enterprise recommendation method according to claim 1, characterized in that, The steps for calculating the weighted scores of each sample data in the sample group based on the weighting method specifically include: The sample ensemble is calculated using the following formula. Correlation coefficient matrix : ； In the formula, and These are the indicator values ​​corresponding to any two columns of indicators; The information content of each indicator in the k-th sample group is calculated using the following formula. : ； In the formula, The correlation coefficient matrix The value corresponding to the i-th row and j-th column; The weights of each indicator in the k-th class of samples are calculated using the following formula. ; ； Calculate the CRITIC weight score for each sample in the k-class sample group using the following formula: 。

Images