KR20190110084A - Esg based enterprise assessment device and operating method thereof - Google Patents

Abstract

Disclosed are a company evaluation performing apparatus based on ESG and an operation method thereof. According to one of various embodiments of the present invention, an ESG company evaluation apparatus may include: a news collecting unit collecting a plurality of news articles on the Internet, classifying the same by date or company, and analyzing similarity between the news articles to perform clustering for news articles having a similarity equal to or greater than a reference value; a news classifying unit classifying each of the news articles to determine which issue among environment, society, or hierarchy structure is related to each article; and an evaluation result deriving unit calculating ESG risk for a corresponding cluster in a cluster unit and calculating ESG company evaluation score based on the calculated value. The evaluation result deriving unit may include: an evidence level calculating unit which performs categorization for nouns extracted from the news articles and performs evidence level calculation in a cluster unit based on nouns included in each category; an ESG probability calculating unit which calculates probability which issue among environment, society, or hierarchy structure each cluster belongs to based on the category item extracted from the evidence level calculating unit; and a score calculating unit which calculates ESG related risks, company risks, and other risks for related issues in different ways and then calculates an ESG company evaluation score based on the calculated risks. According to one embodiment of the present invention, it is possible to increase calculation speed of deriving ESG company evaluation by automatically performing ESG company evaluation.

Description

ESG-Based Enterprise Assessment Device and How It Works {ESG BASED ENTERPRISE ASSESSMENT DEVICE AND OPERATING METHOD THEREOF}

Various embodiments of the present invention relate to an ESG-based enterprise evaluation performing apparatus and a method of operating the same, and collects and analyzes news articles on the Internet to analyze and score risks related to environmental, social, or governance issues of enterprises. To a device and a method of operating the same.

Recently, as companies pay more attention to risk management, they are evaluating their own companies and other companies in various ways and applying these evaluation results to investment, merger, and production line risk management.

In general, firms have generally assessed firms based on financial data that can be quantitatively acquired, but recently, a methodology of evaluating firms and analyzing risks based on non-financial data has emerged. Financial data disclosed by companies does not reflect adverse effects on the company, and the reliability of financial reports provided by the company is also questioned. The reason why corporate analysis is needed through non-financial data is that people's SNS activity through the internet is active, and the reputation of the company or product is spread, and the reputation of the company is shaken. The rumors may increase the risks of the company, but such events are difficult to analyze through financial data. Thus, by analyzing the company not only through financial data but also through non-financial data, more accurate evaluation of the company can be achieved.

In this trend, a methodology for analyzing non-financial data by dividing it into three themes of ESG (Environmnet, Social, Governance) is emerging.

There are companies that produce evaluation reports on companies based on non-financial data such as ESG. However, non-financial data is subjective and the report is relatively slow, making it difficult for companies to use it. In order to remedy these shortcomings, computer programs, etc., collect and automatically analyze news articles about ESG on the Internet to conduct corporate evaluations. There were many difficulties in the process.

Domestic Publication No. 10-2006-0096432

Various embodiments of the present invention are derived to solve the above problems, and aims to perform an ESG based evaluation of companies based on the collection of news articles on the Internet.

Another object of the present invention is to cluster news articles dealing with the same event through similarity analysis between collected news articles.

It is another object of the present invention to improve the performance in categorizing the news gathered by having the device performing the ESG enterprise assessment adopt a machine learning algorithm and perform the learning.

Problems to be solved by the present invention are not limited to the above-mentioned problems, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

In order to achieve the above object, an embodiment of the present invention is an ESG company evaluation device for calculating the score by evaluating the enterprise in terms of ESG (Environmnet, Social, Governance), by collecting a plurality of news articles on the Internet A news collector configured to classify by date or company and perform clustering on news articles having a similarity or more than a reference value through similarity analysis between the news articles; A news classification unit for classifying each of the news articles as related to environmental, social or governance issues; And an evaluation result derivation unit configured to calculate an ESG enterprise evaluation score based on the calculated ESG risk for the cluster in cluster units, and the evaluation result derivation unit includes a category for nouns extracted from the news articles. An evidence level calculator configured to perform speech conversion and to calculate an evidence level in cluster units based on nouns included in each category item; An ESG probability calculator for calculating a probability of whether each cluster is included in an environment, a society, or a governance structure based on the category item extracted by the evidence level calculator; And a score calculation unit that calculates ESG enterprise evaluation scores based on the risks of ESG related risks, enterprise risks, and other related problems in different ways.

The evidence level calculation unit may perform ranking setting or evidence level calculation for each category item in a cluster unit based on the frequency of words included in each category item.

The ESG probability calculator may classify the category item extracted by the evidence level calculator into a set having a classification of environment, society, governance, corporate risk, and other related problems.

The ESG probability calculator may be based on a level of evidence calculated for the environment, society, or governance structure in calculating a probability of whether each cluster is included in an environment, a society, or a governance structure. have.

In order to achieve the above object, another embodiment of the present invention is a method for estimating the enterprise from the ESG (Environmnet, Social, Governance) perspective to calculate the score in the ESG company evaluation device, a plurality of news articles on the Internet Collecting and classifying by date or company and performing clustering on news articles having a similarity or more than a reference value through similarity analysis between the news articles; Classifying each of the news articles as relating to environmental, social or governance issues; Performing categorization on nouns extracted from the news articles, and performing evidence level calculation on a cluster basis based on nouns included in each category item; Calculating a probability of whether each cluster is included in an environment, a society, or a governance based on the extracted category item; And calculating an ESG enterprise valuation score based on the calculation of the risks for the ESG-related risks, the corporate risks, and other related problems in different ways, respectively.

According to an embodiment of the present invention, the ESG company evaluation is automatically performed, so that the speed at which the ESG company evaluation is derived can be improved.

According to another embodiment of the present invention, an enterprise evaluation apparatus may be provided in which the performance of classifying news articles by machine learning continues to be improved.

According to another embodiment of the present invention, an analysis of news articles written in various kinds of languages may be performed to evaluate a company based on the analysis.

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a conceptual diagram schematically illustrating a flow in which ESG company evaluation is performed according to an embodiment of the present invention.
Figure 2 is a block diagram schematically showing the configuration of the EGS company evaluation module according to an embodiment of the present invention.
3 is a diagram for explaining a method of categorizing nouns extracted from news articles by the evidence level calculator.
FIG. 4 is a diagram for describing a method of estimating, normalizing, and estimating an evidence level based on a numerical value of an item based on the number of words included in a category item.
5 is a diagram illustrating a result of the ESG probability calculation unit performing ESG probability calculation for each cluster according to an embodiment of the present invention.
6 is a block diagram schematically illustrating a process of deriving an ESG company evaluation score from a process of collecting news articles by an ESG company evaluation apparatus according to an embodiment of the present invention.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, "comprises" and / or "comprising" does not exclude the presence or addition of one or more other components in addition to the mentioned components. Like reference numerals refer to like elements throughout, and "and / or" includes each and all combinations of one or more of the mentioned components. Although "first", "second", etc. are used to describe various components, these components are of course not limited by these terms. These terms are only used to distinguish one component from another. Therefore, of course, the first component mentioned below may be a second component within the technical spirit of the present invention.

When any part of the specification is to "include" any component, this means that it may further include other components, except to exclude other components unless otherwise stated. In addition, the terms "... unit", "module", etc. described in the specification mean a unit for processing at least one function or operation, which may be implemented in hardware or software or a combination of hardware and software. .

1 is a conceptual diagram schematically illustrating a flow in which ESG company evaluation is performed according to an embodiment of the present invention.

ESG company evaluation disclosed in the present invention can be automated and performed by a program implemented in the form of computer software. That is, each of the methods illustrated in FIG. 1 may be performed by arithmetic processing by software installed in the ESG company evaluation apparatus 100 that performs ESG company evaluation.

Referring to FIG. 1, the ESG company evaluation apparatus 100 may calculate a final company evaluation result through three steps. Referring to FIG. 1A, the ESG company evaluation apparatus 100 may first collect news articles, which are basic materials for performing corporate evaluation, on the Internet. In collecting news articles, the ESG company evaluation apparatus 100 determines whether a news is about a company or a topic through a morphological analysis and a similarity calculation between documents, and primarily classifies similar articles. Clustering can be performed.

Referring to FIG. 1B, the ESG company evaluation apparatus 100 may perform more precise news article classification based on the collected and primarily clustered news articles. In performing the news article classification operation, the ESG company evaluation apparatus 100 first determines whether the collected news is related to ESG, that is, at least one of environment, society, or governance structure, and then collects the news. You can determine and classify the news as it relates to the environment, society, or governance. Finally, the ESG company evaluation apparatus 100 may perform more detailed categorization for each news article classified into three subjects of environment, society, and governance.

Referring to FIG. 1C, the ESG company evaluation apparatus 100 may derive a final enterprise evaluation score based on the classified news articles. In this process, the ESG enterprise evaluation apparatus 100 may calculate the level of evidence for key words included in the news article, and calculate the probability of whether the news articles belong to the environment, society, or governance in cluster units. have.

For convenience of explanation, the ESG enterprise evaluation apparatus 100 of the present invention is shown and described as deriving the final ESG enterprise evaluation score through three steps, but such steps are divided or integrated so that fewer or more Of course, it can be implemented in steps.

2 is a block diagram schematically showing the configuration of the EGS company evaluation module 100 according to an embodiment of the present invention.

Referring to FIG. 2, the ESG company evaluation apparatus 100 includes a control unit 110, a news collection unit 120, a news classification unit 130, an evaluation result derivation unit 140, a communication unit 150, and a storage unit 160. It may be configured to include). In addition, the news collector 120 is a morphological analysis unit 121, the company and date classification unit 122 and the news clustering unit 123, the news classification unit 130 is a binary classification unit 131, ESG classification unit 132 and the category classifier 133, the evaluation result deriving unit 140 may include an evidence level calculator 141, an ESG probability calculator 142, and a score calculator 143, respectively.

For convenience of description, the subjects performing the respective roles in the ESG enterprise evaluation apparatus 100 are shown in the form of ~, but each of the parts may be sub-program modules operating in the ESG enterprise evaluation apparatus 100. . Such program modules include, but are not limited to, routines, subroutines, programs, objects, components, data structures, etc. that perform respective operations or execute particular abstract data types.

According to an embodiment, the controller 110 may include a news collector 120, a morpheme analyzer 121, a company and date classifier 122, a news clustering unit 123, a news classifier 133, and a binary classifier. 131, ESG classification unit 132, category classification unit 133, evaluation result derivation unit 140, evidence level calculation unit 141, ESG probability calculation unit 142, score calculation unit 143, communication unit The data flow between the 150 and the storage 160 may be controlled. That is, the controller 110 according to the present invention includes a news collector 120, a morphological analyzer 121, a company and date classifier 122, a news clustering unit 123, a news classifier 133, and a binary classification. The unit 131, the ESG classification unit 132, the category classification unit 133, the evaluation result derivation unit 140, the evidence level calculation unit 141, the ESG probability calculation unit 142, the score calculation unit 143, Each of the communicator 150 and the storage 160 may perform a unique function.

As described above, the news collector 120 may include a morpheme analyzer 121, a company and date classifier 122, and a news clustering unit 123. In order for the ESG enterprise evaluation apparatus 100 to perform an enterprise evaluation, it is necessary to collect only news articles related to the environment, society, or governance structure of a company from among numerous news articles published on the Internet, and collect news from a reliable source. Acquisition may not produce false assessments, such as by false news. The news collector 120 may collect news articles updated on the Internet at regular time intervals, and in collecting news articles, date information of each news article is published, media information, etc. that publish the news. Can be collected together. The news article collection of the news collector 120 may be performed through the communicator 150 in the ESG company evaluation apparatus 100.

The morpheme analysis unit 121 may analyze the entire text of the collected news articles in a morpheme unit that is a minimum unit having a meaning. When analyzing the news article written in Korean, the morpheme analysis unit 121 may select one of various Korean morpheme analyzers and analyze the same. According to an embodiment, there are various types of Korean morphological analyzers such as kiwi, HAM, HLX, Mecab, etc. The morphological analyzer 121 may perform morphological analysis using one of them and use the result. According to another embodiment, the morpheme analysis unit 121 may analyze the news written in a language other than Korean by using a morpheme analyzer corresponding to the corresponding language.

The company and date classification unit 122 may classify the collected news articles by company and date. The company and date classification unit 122 may determine which company the article is about by using a company dictionary in which company names are constructed in the form of a database. The company and date classification unit 122 may determine which company the article is about, and then check the date the article is published, and organize the news articles by period. According to an embodiment of the present disclosure, the company and date classification unit 122 may classify news articles by companies, and then classify news related to the same company by a week period or a month period. Such period-based classification may be used to calculate the latest news articles in a weighted manner in the ESG company evaluation apparatus 100 to perform the company evaluation later.

The news clustering unit 123 may configure related news, which covers the same issue in the collected news articles, into a cluster. The news clustering unit 123 may perform a similarity calculation between news articles to confirm whether a plurality of news articles are related news dealing with the same issue.

According to an embodiment of the present disclosure, the news clustering unit 123 may perform similarity calculation between news articles by performing cosine similarity calculation between each vector after vectorizing each document.

The news clustering unit 123 may vectorize each news article using TF-IDF (Term Frequency-Inverse Document Frequency). TF (Term Frequency) is a value representing the frequency of occurrence of a specific word in a document. The higher this value, the more important the word in the document. The frequency at which a particular word appears in a plurality of collected news articles may be represented as a document frequency (DF), which is an inverse document frequency (IDF). If a word appears frequently among multiple news articles, it is an IDF rather than a DF because the word becomes a conventional word and cannot be a key word in the news article. The TF-IDF is defined as a value obtained by multiplying the TF and the IDF, and the manner in which the news clustering unit 123 calculates the TF-IDF for the news article according to an embodiment of the present invention may use Equation 1. .

In Equation 1, tf _ij denotes the frequency of occurrence of the word i in the news article j, and df _i denotes the number of news articles including the word i in the group consisting of news articles.

The news clustering unit 123 may vectorize each news article based on Equation 1, and calculate a similarity between news articles based on the vector value of each news article.

According to an embodiment, the news clustering unit 123 may calculate the similarity between news articles through Equation 2 above. In Equation 2, A and B are the respective vectors, and A _i and B _i are the i th component in the A and B vectors, respectively. The cosine similarity between news articles is calculated as a number between 0 and 1, where 0 means that the news articles are independent of each other, and 1 means that the news articles are completely identical to each other.

After calculating the similarity between news articles, the news clustering unit 123 may determine whether the news articles are similar or not based on a reference value set through an experiment in advance. Can tie Hereinafter, the term 'cluster' in the specification means a set of news articles determined by the news clustering unit 123 as similar news.

As described above, the news classifier 130 may include a binary classifier 131, an ESG classifier 132, and a category classifier 133. The news classifier 130 may classify the news articles collected by the news collector 120 according to three criteria. The news classifier 130 may first classify the news articles through the binary classifier 131, then classify the news articles through the ESG classifier 132, and finally classify the news articles through the category classifier 133. . In this process, news articles can be classified on a more detailed basis.

The binary classification unit 131 may determine whether each of the collected news articles is related to the ESG and classify the data in a true or false manner. That is, the binary classification unit 131 determines whether the collected news is data that can be used to evaluate the ESG company. If it is determined to be related to at least one of the environment, society, and governance, the binary classification unit 131 classifies as TRUE, otherwise, FALSE. Can be classified. Therefore, the binary classification unit 131 classifies the news articles that are not necessary for the ESG company evaluation, such as sports, entertainment, politics, art, etc. as FALSE, so that the ESG company evaluation apparatus 100 processes the amount of news articles to be processed. Can be significantly reduced.

According to an embodiment, the ESG classification unit 132 may classify news articles classified as TRUE in the binary classification unit 131 into three issues of environment, society, and governance. Thereafter, the classified news articles may be classified into more subdivided categories by the category classification unit 133. That is, the issues of environment, society, and governance are divided into more detailed subcategories, and one news article belongs to any one of subcategories by the category classification unit 130.

According to an embodiment, the binary classifier 131, the ESG classifier 132, and the category classifier 133 may classify news articles using a machine learning algorithm.

Looking at how the binary classifier 131, the ESG classifier 132, and the category classifier 133 classify using a machine learning algorithm, it is possible to first extract only nouns from news articles through natural language analysis. . According to an embodiment, the binary classifier 131, the ESG classifier 132, and the category classifier 133 may use KoNLP, which is a type of natural language analysis package, to extract only nouns from a news article composed of Korean. . According to another exemplary embodiment, the binary classifier 131, the ESG classifier 132, and the category classifier 133 may use an analysis tool suitable for a corresponding language to extract a noun from a news article composed of another language.

The binary classifier 131, the ESG classifier 132, and the category classifier 133 may extract a noun only from a news article and then generate a matrix using the above-described TF-IDF through Equation 1. The matrix generated using the TF-IDF may be generated for each cluster classified by the news clustering unit 123. According to an embodiment, in the generated matrix, a row may include a word (noun) included in each news article, and a column may list a news article including the word for each word (noun). Can be.

The binary classifier 131, the ESG classifier 132, and the category classifier 133 may then perform a classification suitable for the role of each classifier through a machine learning algorithm based on a matrix generated using the TF-IDF. Can be. According to an embodiment, the binary classification unit 131, the ESG classification unit 132, and the category classification unit 133 employ a specific machine learning algorithm and then receive the training data pre-classified by the user to perform the learning. Therefore, the sorting ability can be improved. Subsequently, test data may be input to the binary classification unit 131, the ESG classification unit 132, and the category classification unit 133, which have been trained to some degree through the training data, and each classification unit that has passed the test corresponds to the corresponding classification unit. Machine learning algorithms can be used to classify news articles.

According to an embodiment, the binary classification unit 131, the ESG classification unit 132, and the category classification unit 133 may be machines such as Multinomial Bayes, Bernoulli Bayes, Stochastic Gradient Descend (SGD), Linear SVC, Perceptron, Random Forest, and the like. One of the running algorithms can be selected and employed.

Evaluation result deriving unit 140 according to an embodiment may perform the ESG company evaluation based on the news articles collected by the news collector 120 and classified by category by the news classification unit 130. That is, the evaluation result deriving unit 140 may perform the scoring for each company based on the ESG-related article and perform the company evaluation through the score. As described above, the evaluation result deriving unit 140 may include an evidence level calculating unit 141, an ESG probability calculating unit 142, and a score calculating unit 143.

The evidence level calculator 141 may classify nouns extracted from news articles into a plurality of word categories. According to an embodiment, the evidence level calculator 141 may classify nouns into environmental damage (D_Env), company stakeholder (S_Company), general damage (General Dagmae, D_gen), and the like.

3 is a diagram for explaining a method of categorizing nouns extracted from news articles by the evidence level calculator 141.

Referring to FIG. 3, the evidence level calculator 141 may finally perform categorization by determining which items of the extracted nouns correspond to which items in the ESG and which elements are also related to the items. The categorization of such nouns may be different from the categorization performed by the category classification unit 133 described above. That is, the categorization performed by the category classifier 133 categorizes each news article into subcategories finer than the issues of environment, society, and governance, and the categorization performed by the evidence level calculator 141 is news. It may be a categorization of nouns extracted from articles. That is, the category classifier 133 performs categorization on each of the news articles, and the evidence level calculator 1410 performs categorization on the nouns. The entities belonging to the categories differ from the news articles and the nouns, respectively. can do.

According to an embodiment, the evidence level calculator 141 performs the categorization as described above, and then a few words are added to each category item for each cluster, which is a collection of news articles classified by the news cluttering unit 123. Can be calculated. Such words may be nouns extracted by the evidence level calculator 141 and categorized as shown in FIG. 3.

The evidence level calculator 141 may set the rank of each category item based on the number of words included in the category item, and may calculate the evidence level of each category item based on the normalized number by normalizing the number of words. .

FIG. 4 is a diagram for describing a method of the evidence level calculating unit 141 calculating the level of evidence based on the number of words included in the category item, normalizing the item, and calculating the evidence level based on the normalized value.

Referring to FIG. 4, how many words are included in each category and which words are included in one cluster is displayed, and the ranking for each category is set based on the number of words included. In addition, the number of words included in each category is shown as a normalized value, and the evidence level is displayed based on the normalized value. The evidence level calculator 141 may calculate the evidence level by classifying the normalized numerical value by the section according to a predetermined reference value.

The ESG probability calculator 142 may calculate a probability of whether each cluster, which is a set of similar news, belongs to an environment, a society, or a governance structure, based on the category of the category used by the evidence level calculator 141. have.

The ESG probability calculator 142 may define the category items used by the evidence level calculator 141 in a smaller number of sets before calculating the probability of which issue each cluster belongs to. According to one embodiment, such sets may be classified as environmental, social, governance, corporate risk and other related issues.

The ESG probability calculator 142 may calculate the level of evidence on environmental, social or governance issues for a particular cluster using Equation 3 above. As an example of how the ESG probability calculator 142 calculates the evidence level for the environment, in Equation 3, C _i means the i-th category among the categories for the environment, and n means the number of categories for the environment. Where E (C _i− ) is calculated as described above in the evidence level calculator 141 as the evidence level of the i th category for the environment. That is, the evidence level calculated by the evidence level calculator 141 is an evidence level for categories used to classify the extracted nouns, and the evidence level calculated by the ESG probability calculator 142 is based on environmental, social, and governance structures. Level of evidence for three types of issues.

The ESG probability calculator 142 may calculate a probability of which of the three types of clusters are related to the environment, society, and governance.

The ESG probability calculator 142 may calculate the ESG probability of each cluster by using Equation 4 above.

5 is a diagram illustrating a result of ESG probability calculation for each cluster performed by the ESG probability calculator 142 according to an embodiment of the present invention.

Referring to FIG. 5, for a plurality of clusters, evidence levels and three ESG probabilities are shown for three types of issues: environment, society, and governance. If you add up the probabilities of whether a cluster is related to environmental, social, or governance issues, you can see that it becomes 1.

According to an embodiment, the score calculator 143 may calculate the ESG company valuation score using the auxiliary indicators calculated after calculating various auxiliary indicators before calculating the ESG company valuation score, which is the final stage of ESG company evaluation. have. Such supplementary indicators and ESG company valuation score can be calculated through the level of evidence and ESG probability calculated as described above.

 According to one embodiment, the various supplementary indicators may be composed of an ESG risk score, a corporate risk score and a relevance score. Such auxiliary indicators are not limited to the above three types but may be defined in various numbers and calculation methods.

According to an embodiment, the score calculator 143 may finally calculate the ESG company evaluation score in the same manner as in Equation 5 above. In Equation 5, ESGrisk, CompanyRisk, and Relevance represent various auxiliary indicators. In Equation 5, ESGrisk, CompanyRisk, and Relevance represent various auxiliary indicators. ESGrisk can be calculated by summing and calculating how many risks related to the environment, society, and corporate structure of a set of news articles classified into a specific cluster, and CompanyRisk is a set defined by ESG probability calculation unit 142. Among the sets defined by the ESG probability calculation unit 142 and other categories and word categories classified into the set, Relevance may be a value related to a set related to corporate risk and word categories classified into the set. It may be a related value.

The communication unit 150 according to an embodiment enables communication between the ESG company evaluation apparatus 100 and an external device. In detail, the ESG company evaluation apparatus 100 may communicate with the user terminal of the corresponding apparatus, and the news collector 120 may collect the published news through the Internet connection.

The storage unit 160 according to an embodiment may store data necessary for the operation of the ESG company evaluation apparatus 100. The storage unit 160 may store the collected news articles, classification information on the news articles, scoring information, learning history of machine learning algorithms, and the like in the form of data.

In the manner as described above, the ESG enterprise evaluation apparatus 100 may finally derive the ESG enterprise evaluation score for each cluster, which is a set of similar articles.

6 is a block diagram schematically illustrating a process of deriving an ESG company evaluation score from a process of collecting news articles by the ESG company evaluation apparatus 100 according to an embodiment of the present invention.

Referring to FIG. 6, the ESG company evaluation apparatus 100 may collect news articles, which are basic data in performing enterprise evaluation, at regular or irregular time intervals on the Internet (S601).

Thereafter, the ESG company evaluation apparatus 100 may classify the collected news articles by company and date (S603), and analyze the corresponding articles using an appropriate stemmer according to the language in which the collected news articles are written. (S605).

The ESG company evaluation apparatus 100 vectorizes each news article based on the news articles for which morphological analysis is completed, calculates similarity among news articles through similarity analysis between vectors, and collects related news and performs clustering. It may be (S607). In this process, the ESG company evaluation apparatus 100 may vectorize each news article by using the TF-IDF, and calculate the similarity between each news article through cosine similarity calculation.

The ESG company evaluation apparatus 100 may classify whether each of the news articles collected belongs to an environment, a society, and a governance issue, and a category belonging to each of the environmental, social, and governance issues. (S609). In this process, the ESG enterprise evaluation apparatus 100 may preemptively classify whether each news article is related to performing ESG enterprise evaluation in the form of TRUE or FALSE, and select an appropriate machine learning algorithm in each classification step. By adopting and generating the training data, the ESG company evaluation apparatus 100 may perform the learning through the corresponding algorithm. The user of the EGG enterprise evaluation apparatus 100 may verify the ESG enterprise evaluation apparatus 100 that has learned through the machine learning algorithm and the training data through the test data, and may allow each classification of the step S609 to be performed. .

The ESG company evaluation apparatus 100 may calculate an evidence level value for each ESG issue of the corresponding cluster in each cluster unit classified in step S607, and calculate an ESG probability value for the corresponding cluster (S611). In this process, the ESG enterprise evaluation apparatus 100 may classify nouns extracted from news articles into a plurality of word categories and classify them into sets such as environment, society, governance, corporate risk, and other related issues. Subsequently, the ESG company evaluation apparatus 100 may calculate an evidence level value for the environment, society, and governance structure among the sets, and also calculate an ESG probability value.

Finally, the ESG company evaluation apparatus 100 may calculate the final ESG company evaluation score based on the evidence level value and the ESG probability value calculated in step S611 (S613). In this process, the ESG company evaluation apparatus 100 may utilize the ESG probability value calculated in step S611 and the evidence level value of the set classified as the enterprise risk.

According to an embodiment of the present invention, ESG company evaluation apparatus 100 provides a function for providing a basis for the score is obtained when an individual or a company using the ESG company evaluation results request the basis for the evaluation score for a particular company. It may include. That is, when the evaluation scores for articles of a specific company are calculated, the evaluation scores can provide information that the influence of the specific scores on the articles is affected by the frequency of a specific word, and search for articles containing the words. It can also be used.

As such, when the ESG enterprise evaluation apparatus 100 disclosed by the embodiments of the present invention performs an automated enterprise evaluation, when similar news articles about a specific company are published on the Internet, the news articles are displayed in the environment, society, It is possible to determine which issues of governance are relevant and how critical the risks are to those issues.

Meanwhile, the ESG enterprise evaluation apparatus 100 according to an embodiment of the present invention may also be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored.

For example, the computer-readable recording medium may be a ROM, a RAM, a CD-ROM, a magnetic tape, a hard disk, a floppy disk, a removable storage device, a nonvolatile memory (Flash memory). Optical data storage.

The computer readable recording medium can also be distributed over computer systems connected over a computer network so that the computer readable code is stored and executed in a distributed fashion.

In the above, embodiments of the present invention have been described with reference to the accompanying drawings, but those skilled in the art to which the present invention pertains may implement the present invention in other specific forms without changing the technical spirit or essential features. I can understand that. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive.

100: ESG company evaluation device 110: control unit
120: news collection 121: morphological analysis
122: corporate and date classification unit 123: news clustering unit
130: news classification unit 131: binary classification unit
132: ESG classification unit 133: category classification unit
140: evaluation result deriving unit 141: evidence level calculating unit
142: ESG probability calculator 143: Score calculator

Claims (5)

In the ESG company evaluation device for calculating the score by evaluating the company in terms of ESG (Environmnet, Social, Governance),
A news collector configured to collect a plurality of news articles on the Internet, classify them by date or by company, and perform clustering on news articles having a similarity or more through a similarity analysis between the news articles;
A news classification unit for classifying each of the news articles as related to environmental, social or governance issues; And
It includes an evaluation result derivation unit that calculates ESG enterprise valuation scores based on the calculated and calculated ESG risk for the cluster on a cluster basis.
The evaluation result deriving unit may include: an evidence level calculator configured to categorize nouns extracted from the news articles and to calculate an evidence level in cluster units based on nouns included in each category item;
An ESG probability calculator for calculating a probability of whether each cluster is included in an environment, a society, or a governance structure based on the category item extracted by the evidence level calculator; And
ESG enterprise evaluation apparatus comprising a score calculation unit for calculating the risk for ESG-related risks, corporate risks and other related problems in a different manner, and then calculates ESG enterprise valuation scores based thereon.
The method of claim 1,
The evidence level calculation unit, ESG enterprise evaluation apparatus for performing the ranking setting or evidence level calculation for the category item in the cluster unit based on the frequency of the words included in each category item.
The method of claim 1,
And the ESG probability calculating unit classifies the category item extracted by the evidence level calculating unit into a set having a classification of environment, society, governance, corporate risk, and other related problems.
The method of claim 1,
The ESG probability calculation unit is based on the level of evidence calculated for the environment, society or governance in calculating the probability of each cluster is included in the issue of environmental, social or governance, ESG enterprise evaluation device.
In the method by which the ESG company evaluation device evaluates the company in terms of ESG (Environmnet, Social, Governance) to calculate the score,
Collecting a plurality of news articles on the Internet, classifying them by date or company, and performing clustering on news articles having a similarity or more than a reference value through similarity analysis between the news articles;
Classifying each of the news articles as relating to environmental, social or governance issues;
Performing categorization on nouns extracted from the news articles, and performing evidence level calculation on a cluster basis based on nouns included in each category item;
Calculating a probability of whether each cluster is included in an environment, a society, or a governance based on the extracted category item; And
A method for calculating the score of an ESG enterprise evaluation device, comprising calculating the ESG enterprise valuation score based on the calculation of the risk for the ESG related risk, the enterprise risk and other related problems in different ways.

Images