KR20190064749A - Method and device for intelligent decision support in stock investment - Google Patents

Abstract

The present invention relates to an intelligent stock investment decision support method including: a step (a) of collecting at least one among news, SNS, reports, public announcement data, financial statement data, and postings and comments in a finance-related community bulletin board in regard to an analysis target enterprise; a step (b) of screening a risky enterprise by using at least one among the collected public announcement data, financial statement data, and news; a step (c) of deriving an emotional value by word and document by using the news and SNS data, and deriving a credit risk level through variation trends of the emotional value by a word and a document; and a step (d) of deriving a signal based on the derived credit risk level and predicting the probability of the occurrence of a credit event by using the derived signal.

Description

TECHNICAL FIELD The present invention relates to an intelligent investment decision support method and apparatus,

The present invention relates to a method and apparatus for intelligent securities investment decision support.

Recently, as interest in artificial intelligence technology has been amplified, researches on machine learning, deep learning, and the like are being actively carried out. Particularly in the Fin-tech industry, innovations are being made that can make a big difference to existing financial services companies, focusing on payment systems, financial services, bank loan services, investment asset management, insurance and market infrastructure.

In addition, the pin-tec industry is becoming more and more active at the forefront of global ICT (information and communication technology) and start-up companies. Especially, we are analyzing customer 's tendency through Twitter' s analysis of one of the social network services, and providing customized information and services based on it.

However, this success is rare in Korea because the domestic mobile environment and technology are excellent, while the strict domestic financial service regulatory environment such as low autonomy and advance security deliberation hinders innovation activities in the domestic pin tech industry .

Machine learning, which is a field of artificial intelligence, can be predicted based on learned attributes through training data. There have been many attempts to predict stock price by applying machine learning to stock price data or financial statements, , Only at the predicted level and there are very few successful cases.

The present invention provides an intelligent securities investment decision support method and apparatus for supporting investment decision making through Opinion Mining and machine learning based on data collected from news and SNS.

In addition, the present invention predicts credit risk through opinion mining, emotional analysis and machine learning using data collected from news and SNS, defines a risk signal and predicts the probability of a credit event event in the end And to provide a method and apparatus for intelligent securities investment decision support.

According to an aspect of the present invention, there is provided an intelligent securities investment decision support method capable of supporting investment decision making through Opinion Mining and machine learning based on data collected from news and SNS.

According to an embodiment of the present invention, there is provided a method of supporting an intelligent securities investment decision, comprising the steps of: (a) selecting at least one of news, SNS, report, disclosure data, financial statement data, Collecting one; (b) screening a risk company using at least one of the collected disclosure data, financial statement data, and news; (c) deriving emotion values for each word and each document by using the news and SNS data, and deriving a credit risk through a variation of emotion values for each word and document; And (d) deriving a signal based on the derived credit risk, and estimating a credit event occurrence probability using the derived signal.

If the analysis target company is a listed company, deriving the disclosure index and the financial index using the disclosure data and the financial statement data; Applying the derived known index and financial index to the learned support vector machine (SVM) to classify the normal index and the financial index into a normal company and a risk company; Analyzing the occurrence frequency of the predetermined credit event keyword in the collected data to secondarily derive the risk company; And selecting a final risk candidate company by synthesizing the primary classification and the secondary derivation result.

The above credit event is a credit event, which includes capital increase, capital erosion, corporate bond, treasury stock disposition, liquidation merger, tangible accounting, free capitalization, private equity, major shareholder or major shareholder sale, debt guarantees for others, The occurrence of unpaid principal of the debentures, the disclosure of inquiries and the delayed reduction of the report.

In the step (b), if the analyzed entity is a non-listed company, the occurrence frequency of the credit event keyword may be derived from the collected news and disclosure data to screen the risky company.

 The step (c) may include the steps of: if the analysis target company is a listed company, assigning a sensitivity value using a stock price for each Opinion included in the collected data; Selecting key keywords based on positive and negative based on the emotion value; Selecting a meaningful word from the collected data to create a data set, deriving and embedding a vector value of each word included in the data set, Deriving a distance using the vector value between the core keyword and the word in the vector space and deriving an emotion value of each word based on the derived distance; Deriving emotion values for each document by totaling the emotion values of the words included in the document; And calculating the credit risk using the derived fluctuation of emotion value for each document.

The step (c) may include the steps of: defining an emotional dictionary when the analyzed entity is a non-listed company; Selecting a meaningful word from the collected data to create a data set, deriving and embedding a vector value of each word included in the data set, Deriving an emotion value for each word in the vector space using the emotion dictionary; Deriving emotion values for each document by summing emotion values of the words included in the document; And calculating the credit risk using the derived fluctuation of emotion value for each document.

The step of defining the emotional dictionary includes the steps of: collecting data expressing emotion for each stock opinion-related opinion; And deriving an emotion value of each word using the na ㅿ and bayes classifier for the collected data.

In the step (d), the derivation of the signal may be derived using a daily variation of the emotion value for each document.

The derived signal includes a signal for maintaining positive or negative for a predetermined period of time (for example, three days), a signal for increasing or decreasing the daily average sensibility value, a signal for crossing the daily average sensibility value, A ratio of values, and a weekly emotion value average increase rate signal per document.

(D) collecting data of companies in the same industry and performing a logistic regression analysis to check whether a predetermined credit event is generated within a predetermined period to derive effective signals for each industry; Constructing a regression equation using the industry specific effective signal; And estimating a credit event occurrence probability by reflecting the derived signal in a regression equation of the same kind of industry.

According to another aspect of the present invention, there is provided an apparatus for executing an intelligent securities investment decision support method capable of supporting projection decision making through Opinion Mining and machine learning based on data collected from news and SNS.

According to an embodiment of the present invention, there is provided a computer apparatus in which an intelligent securities investment decision support method is implemented, the computer apparatus comprising: a communication unit; A memory in which at least one instruction is stored; And a processor for executing the instructions stored in the memory, wherein the instructions executed by the processor include: (a) news related to the company to be analyzed, SNS, report, disclosure data, financial statement data, And at least one of a comment; (b) screening a risk company using at least one of the collected disclosure data, financial statement data, and news; (c) deriving emotion values for each word and each document by using the news and SNS data, and deriving a credit risk through a variation of emotion values for each word and document; And (d) deriving a signal based on the derived credit risk, and predicting a credit event occurrence probability using the derived signal.

By providing the intelligent securities investment decision support method and apparatus according to an embodiment of the present invention, projection decision making can be supported through Opinion Mining and machine learning based on data collected from news and SNS.

In addition, the present invention predicts credit risk through opinion mining, emotional analysis, and machine learning using data collected from news and SNS, and defines a risk signal to predict the probability of corporate credit event eventually .

1 is a block diagram schematically illustrating an internal configuration of an intelligent securities investment decision support apparatus according to an embodiment of the present invention;
2 illustrates an internal configuration of a data acquisition module according to an embodiment of the present invention;
3 is a diagram illustrating an internal configuration of a risk company screening module according to an embodiment of the present invention;
4 is a flowchart showing a method for screening a risk company by evaluating the financial soundness of a listed company according to an embodiment of the present invention.
5 is a diagram illustrating a financial indicator according to an embodiment of the present invention.
Figure 6 is a diagram illustrating disclosure indices in accordance with an embodiment of the present invention.
FIG. 7 is a view showing a result of comparing the result of prediction of a risk company with a result of actual corporate credit by evaluating the financial soundness through SVM based on disclosure indicators and financial indicators of a test company according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating an example of the frequency of occurrence of credit events in each collected data for each company according to an embodiment of the present invention; FIG.
Figure 9 is a method for evaluating the financial health of an unlisted company in accordance with an embodiment of the present invention.
FIG. 10 illustrates an internal structure of a credit risk calculation module according to an embodiment of the present invention; FIG.
11 is a diagram illustrating a method for evaluating a credit risk of a listed company according to an embodiment of the present invention.
12 illustrates a method for evaluating a credit risk of an unlisted company in accordance with an embodiment of the present invention.
FIG. 13 is a view for explaining emotion value assignment by opinion according to an embodiment of the present invention; FIG.
FIG. 14 is a diagram for explaining emotion dictionary according to an embodiment of the present invention; FIG.
15 illustrates a detailed configuration of a credit event occurrence probability prediction module according to an embodiment of the present invention.
16 is a flowchart illustrating a method of predicting a credit event occurrence probability according to an embodiment of the present invention.
FIG. 17 is a diagram illustrating signals according to an embodiment of the present invention; FIG.

As used herein, the singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise. In this specification, the terms "comprising ", or" comprising "and the like should not be construed as necessarily including the various elements or steps described in the specification, Or may be further comprised of additional components or steps. Also, the terms "part," " module, "and the like described in the specification mean units for processing at least one function or operation, which may be implemented in hardware or software or a combination of hardware and software . Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention is intended to primarily screen risky business groups by using disclosure data, financial indicators, and text-based methods, and to detect risk signals through opinion mining for selected companies and finally to predict credit events.

In this case, credit events that can be confirmed by the disclosure data include capital increase, capital infiltration, corporate bond, treasury stock disposal, liquidation merger, segregation accounting, free capital reduction, private equity fund, sale of largest shareholder or liquor shareholder, Regeneration procedure, occurrence of delinquency of loan principal, occurrence of non-payment of interest on principal of bond, delay of inquiry, and delayed reduction of report.

A brief description of each credit event follows. Each credit event type will briefly describe each credit event type briefly in order to facilitate understanding and explanation in economic terms.

A rights offering implies the sale of additional stock through the stock market upon receipt of funds. In the case of rights offering, the share of existing shareholders tends to be diluted because the number of shares to be traded increases, that is, the increase in the number of stocks implies that the share of the company should be shared with more people, which affects the stock price decline .

Capital incapacity is a state where capital deficit has occurred and the amount of capital that the original company has decreased due to the deficit or other reasons, and the accumulated capital loss is less than the capital due to accumulated deficits. Capital erosion is generally a condition for the cancellation of the listing. If the capital erosion exceeds 50%, the reason for the incorporation of the management stock is considered. If the capital is eroded more than 50% . The easiest way to get rid of this capital erosion is the 'free potato'.

Corporate bonds are classified into BW (bond with warrants) and CB (convertible bonds) depending on the type. BW is divided into BWs and BWs. If the BWs are issued for the capital increase, the right to receive new stocks is preferentially granted. BWs are corporate bonds to which the BWs are entitled. It is a corporate bond that includes an opportunity to buy additional listed shares.

BW has increased its number of shares by issuing new shares, but has not changed its earnings. As the number of shares will increase without any improvement in earnings, it will lead to a decline in share prices as the share of profits per share will decrease.

If the issuance volume of the warrants is equal to the daily trading volume, it is not a risk factor. However, if the issuance volume is more than 10% of the stake, the stock price will be burdened.

Issuance of convertible bonds is not good news for existing shareholders. Exclude abuse of convertible bonds In normal cases, issuance of a telephone call means that the company needs funds. This may be a positive investment in the development of an OR company if cash is needed as the financial condition worsens.

Treasury stock means stocks issued by the Company that are acquired and held by the Company under certain conditions or reasons. The disposition of treasury stock means the operation financing of the enterprise. The disposal of treasury stocks has a negative impact on share prices as stocks increase on the market. Companies that secure funds through disposition of treasury stock can raise the value of the company through new investments, which may be short-term negative, but may be favorable in the long run.

The liquidation merger means increasing the face value by combining stocks with small face value, which is a way to reduce the number of shares and increase the stock price. Parallel merging is a commonly used method of raising share prices in corporations by increasing stock prices while reducing the number of shares. Or because the price of a low stock price can be cheaper, it has an effect of preventing it, and it also has an advantage that the turnover rate is lowered but the stock price is relatively easy to manage.

It is difficult to grasp on the financial statements such as exaggerated accounting and reverse accounting, but it can have a direct influence on the corporate default and the cancellation of the listing on the occasion of one occurrence.

Free bonus is generally obtained by taking stock of the potato ratio and incinerating it without any compensation to the shareholders. In the worst case for shareholders, capital reduction is made by filling the loss with potato profit. However, there are cases where the company carries out additional capital increase to raise stock price. As most companies do not pay compensation to shareholders, the total amount of assets is unchanged, but the share price may fall when free of charge.

A private equity fund is a fund that collects funds from a small number of investors and manages them in stocks and bonds. Private equity refers to a strategy to attract investment funds from a small number of long-term investors, to acquire and restructure corporate and financial institutions, and then to sell or re-sell the investment funds. Private equity funds are exploited as a means of supporting family members or transferring internal funds because of the nature of being able to buy shares in a particular company without restrictions.

Largest shareholder or lump-sum shareholder selling means that the shareholders of the company sell shares to the company. If the largest shareholder changes frequently, it is a sign that the business enterprise is unstable. There is also the possibility of insider trading or unfair trading behavior, which can be a negative factor.

The guarantee of the debt to another person is to guarantee the loan if the third party (the largest shareholder, the management, the other legal person, etc.) it means. This is bad news because unexpected debt is suddenly generated and ancillary expenses can be incurred.

A bankruptcy means that a note or check returned on maturity can not be paid normally until the end of the business day. A bankrupt company is a company that is not capable of repaying its debts. The bankruptcy has a direct impact on the stock market and investment, and the bankruptcy corporation will suffer disadvantages such as suspension of checking transactions and appointment of management items.

The corporate rehabilitation process is a system that provides an opportunity for the debt to recover from the excesses of the enterprise. The corporate rehabilitation process is worthwhile when a company is worth continuing its business, but it will not be able to cover its debt with operating profit due to over-investment and financial accidents. If the regeneration procedure is not possible, the listing may be abolished.

Disclosure of disclosures refers to disclosures in which the Exchange requests for an answer and responds within one-half of a day when there is a significant change in the price or volume of transactions, such as rumors, reports, and sovereign rights regarding major management issues. The inquiry disclosure may have a small impact on the stock market, but it is necessary for the exchange to be interested because it is an announcement requiring a direct reply from a specific event.

Audit report delays occur when there is significant uncertainty within an entity, when the entity fails to present self-reliance, or when the company refuses to correct an important distortion indication. The rejection of the audit report can also be viewed as a related credit event, and a 'fair' company can also grasp the financial imperfections by looking at the operating loss.

FIG. 1 is a block diagram schematically illustrating an internal configuration of an intelligent securities investment decision support apparatus according to an embodiment of the present invention. FIG. 2 illustrates an internal configuration of a data collection module according to an embodiment of the present invention FIG. 3 is a diagram illustrating an internal structure of a risk company screening module according to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a risk company screening module according to an embodiment of the present invention. FIG. 5 is a diagram illustrating a financial indicator according to an embodiment of the present invention. FIG. 6 is a diagram illustrating disclosure indicators according to an embodiment of the present invention. FIG. The result of comparing the predicted value of the risk company with the actual corporate credit result by evaluating the financial soundness through the SVM based on the disclosure index and the financial index of the test company according to the embodiment of the invention FIG. 8 is a view for explaining an example of the frequency of occurrence of credit events in collected data for each company according to each keyword according to an embodiment of the present invention, and FIG. 9 FIG. 10 is a diagram illustrating an internal configuration of a credit risk calculation module according to an embodiment of the present invention, FIG. 11 is a diagram illustrating an internal configuration of a credit risk calculation module according to an embodiment of the present invention FIG. 12 is a diagram illustrating a method of evaluating a credit risk of a listed company according to an embodiment of the present invention. FIG. 12 is a diagram illustrating a method of evaluating a credit risk of an unlisted company according to an embodiment of the present invention. FIG. 14 is a diagram for explaining emotion dictionary according to an embodiment of the present invention, and FIG. 15 is a diagram for explaining emotion value assignment according to an embodiment of the present invention. FIG. 16 is a flowchart illustrating a method of predicting a credit event occurrence probability according to an embodiment of the present invention, and FIG. 17 is a flowchart illustrating a method of predicting a credit event occurrence probability according to an embodiment of the present invention. 1 is a diagram illustrating signals according to an embodiment of the invention.

1, an intelligent securities investment decision support apparatus 100 according to an embodiment of the present invention includes a data collection module 110, a risk company screening module 115, a credit risk calculation module 120, a credit event And an occurrence probability prediction module 125. [

The data collection module 110 is a means for collecting data on the on-line and SNS data of the analysis target company and storing the data on the database.

The data collection module 110 can collect various information related to the analysis target company on-line and on the SNS. However, since the purpose of the present invention is to support decision making of securities investment, the data collection module 110 can collect data related thereto. The data collection module 110 can collect and store data related to the analysis target company on-line and on the SNS through a web crawling technique. Since the web crawling technique itself is obvious to those skilled in the art, a detailed description thereof will be omitted.

For example, the data collection module 110 may collect news, reports, SNS, postings and comments in a community-related bulletin board related to the analyzed company, disclosure data, financial statement data, and the like, and store the collected data in a database.

FIG. 2 schematically shows an internal configuration of the data acquisition module 110. As shown in FIG. Referring to FIG. 2, the data collection module 110 according to an exemplary embodiment of the present invention collects news related to a company to be analyzed and stores the collected news in a news database, collects SNS data and stores the SNS data in SNS data, Can be collected and stored in the financial statement database, and the disclosure data can be collected and stored separately in the disclosure data base

The risky enterprise screening module 115 is a means for analyzing enterprise collected data (for example, disclosed data and financial statement data) to evaluate the financial health and screening the risky enterprise.

It can be classified into two types: the case where disclosure data and financial statement data are available and the case where it is impossible to use. In other words, listed companies can use disclosure data and financial statement data, but unlisted companies can not use disclosure data and financial statement data.

Therefore, these will be separately described separately.

FIG. 3 shows the internal structure of the risk enterprise screening module 115. As shown in FIG. 3, the risky company screening module 115 can screen the risky enterprise by changing the financial soundness evaluation method according to whether the enterprise type is a listed company or a non-listed company.

First, a method of evaluating the financial soundness of a listed company and screening a risky company will be described in detail with reference to FIG.

In step 410, the hazardous enterprise screening module 115 structures the data collected by the enterprise.

In the data structuring stage, financial indicators are derived by calculating detailed items affecting credit events in financial statement data based on data collected by each company.

In addition, the financial statements can be used to derive business performance, profitability, and stability indicators based on the SPL rating model, and then reflect each of these indicators at a specified rate to ultimately derive financial indicators. That is, the risky enterprise screening module 115 can screen the risky enterprise by evaluating the financial soundness of the analyzed enterprise based on the derived financial indicators.

The derivation method for each sub-index for deriving the financial indicators and the reflection ratio of each sub-index are as shown in FIG.

Referring to FIG. 5, a method of deriving detailed indicators for deriving respective financial indicators will be briefly described.

As shown in FIG. 5, in order to derive the financial indicators, it is necessary to derive the business performance index, the profitability index and the stability index respectively, and the business performance index, the profitability index and the stability index are calculated by 30%, 35% Can be used to derive financial indicators.

A method of deriving each index will be described.

The business performance index can be derived by calculating 45% and 55% of the asset size index and asset quality index, respectively. At this time, the asset size indicator and the asset quality indicator can be derived by any one of the methods defined in FIG.

In addition, the profitability index can be derived by calculating 60% and 40% of the profitability index after deriving the quality index of business performance, indicators and performance. The formula for deriving the quality index of business performance, indicators and performance is shown in FIG.

The stability index can be derived by deriving the financial burden index and the reimbursement capacity index by the formulas defined in FIG. 5, respectively, and then calculating the stability index by reflecting 55% and 45%, respectively.

In addition, the risk company screening module 115 derives public indicators. For example, the risk company screening module 115 may analyze the disclosure data of a past period (for example, 2 years or 4 years) including the year of the credit event to use as an indicator. The risk company screening module 115 can derive the disclosure index by deriving the number of occurrences of the indicators shown in FIG. 6 in deriving the disclosure index.

In an embodiment of the present invention, an example of summary of disclosure data affecting the occurrence of a credit event is shown in FIG.

If the financial indicators and each sub-item are derived through the data structure, and the disclosure index (number of disclosures) is derived, the risk company screening module 115 applies the structured data to the support vector machine (SVM) And classified into a company and a risk company. Since the SVM algorithm itself is a well-known technique, a detailed description thereof will be omitted.

The SVM assumes that the collected past corporate data is learned by distinguishing the credit event occurrence group and the normal repayment group. In this way, we use financial indicators (detail items to evaluate business performance, profitability, and stability) and non-financial indicators (disclosure indicators (cases) that are considered to affect creditworthiness) as independent variables to the learned SVM, It is possible to screen the normal enterprise and the risk enterprise primarily by setting whether or not the event occurs.

FIG. 7 illustrates the result of comparing the result of predicting the risk company with the actual corporate credit result by evaluating the financial soundness through the SVM based on the disclosure index and the financial index of the test company.

As shown in FIG. 7, it is possible to classify the normal company and the risk company by applying the company-specific disclosure index and the financial index to the SVM. However, as shown in FIG. 7, the risk company screening module 115 evaluates the financial soundness of the SVM based on the disclosure indicators and the financial indicators of each company. When the risk firm is predicted, the financial soundness is evaluated as good , And there may be cases where it is actually a risk company.

As shown in FIG. 7, it can be seen that, in the case of the right-wing enterprise predicted as a defaulting company through the SVM analysis, it is in agreement with the actual de-listing result. However, the results of the SVM analysis show that the company is estimated to be a normal company, but the actual listing is canceled or the company is out of work.

Accordingly, in step 420, the risk company screening module 115 can identify the risk companies by determining whether a predetermined credit event keyword is generated from the collected data.

For example, the risky enterprise screening module 115 may count the number of occurrences of credit event keywords in the collected data, respectively, and analyze whether a credit event occurs to screen risky companies.

FIG. 8 shows an example of sorting the risk companies after sorting the frequency of occurrence of credit events in the collected data for each company by respective keywords.

Up to now, referring to FIGS. 4 to 8, a method of screening a risk company by evaluating the financial soundness of a listed company has been described.

Referring to FIG. 9, a method for evaluating the financial soundness of a non-listed company will be described.

In step 910, the risk company screening module 115 defines credit events by company / article based on the collected news and disclosure reports.

Unlisted companies are not allowed to use financial statement data, so unlisted companies should use only news or public disclosure reports to assess their financial soundness. Therefore, we define certain credit events for each unlisted company by deriving certain keywords related to credit events from news and disclosure reports.

In step 915, the risk company screening module 115 grasps the occurrence of the credit event-related keywords on an article-by-article basis.

In step 920, the risky enterprise screening module 115 firstly screens the risky enterprise among unlisted companies by calculating the occurrence frequency (frequency of keyword occurrence) and the growth rate of the credit event by time.

As shown in FIG. 9, since the risk company screening module 115 can not use the financial statement data in the case of the unlisted company, the risk company can be selected based on the occurrence of the credit event based on the news and the disclosure data.

Referring again to FIG. 1, the credit risk calculation module 120 is a means for deriving a credit risk for companies selected as a risk company.

The credit risk calculation module 120 is a means for deriving the current credit risk for the companies selected as the risk enterprise. Here, the credit risk refers to the index using emotional value at the level of text and document, change of emotion value, etc. using news and SNS data.

FIG. 10 shows the detailed structure of the credit risk calculation module 120. As shown in FIG. As shown in FIG. 10, the credit risk calculation module 120 may have a different configuration depending on a listed company and a non-listed company.

First, in order to facilitate understanding and explanation, a method of deriving a credit risk of a listed company will be described with reference to FIG.

In step 1110, the credit risk calculation module 120 preprocesses the collected data for the analysis target company. It eliminates unnecessary information and advertisements from the collected data for the analyzed companies in the news or SNS, and can be parsed for emotional analysis and word embedding analysis.

In step 1115, the credit risk calculation module 120 allocates the emotion value by opinion using the stock price.

The credit risk calculation module 120 can assign affirmative or negative emotional values to each opinion data from the preprocessed data using the stock price.

The credit risk calculation module 120 according to an exemplary embodiment of the present invention calculates the credit risk of the current day based on the stock price on the day of the first day, The emotional value can be assigned to the opinion data on the 1st, 2nd and 3rd days, respectively. Here, the emotion value may be positive or negative.

After the emotion value is assigned to the opinion data, the credit risk calculation module 120 allocates emotion value of each word using the Navie Bayes classifier to the emotion-assigned opinion data . Here, words appearing more frequently in affirmative data through Naïve Bayes classification can be derived as negative numbers than those appearing more frequently in positive and negative data.

Since the Naïve Bayes classification itself is a known technique, a detailed description of deriving the polarity value of each word will be omitted based on this.

In step 1120, the credit risk calculation module 120 selects the largest n (natural number) positive and negative words having the greatest absolute value of the emotion value among the words to which the emotion value is assigned by the Naïve Bayes classification, as the core keyword .

In step 1125, the credit risk calculation module 120 extracts a meaningful word from the collected data, and derives and embeds a vector value of the extracted words.

For example, the credit risk calculation module 120 removes noise such as symbols, numbers, and the like, which are unnecessary information for the collected data. The credit risk calculation module 120 may then extract each word from the collected data and tag the part of speech for each word. At this time, it is possible to create a data set of words by removing unnecessary data except nouns and verbs.

The credit risk calculation module 120 may generate a vector value for each word included in the word data set and embed it. The dimension of the vector may be n-dimensional.

The credit risk calculation module 120 may generate a vector value based on the frequency of simultaneous occurrence of words for a word data set.

For example, the credit risk calculation module 120 may embed each word using word2vec. Here, since word2vec is obvious to those skilled in the art, a detailed description thereof will be omitted.

In step 1130, the credit risk calculation module 120 derives emotion values for each word and each document using the distance between each word embedded in the word and the core keyword.

For example, since each word and core keyword are respectively transformed into a vector value and embedded, a distance can be derived based on the difference between the core keyword and the vector value of each word in the vector space. Based on this, Can be derived.

For example, the emotion value of each document can be calculated by multiplying the occurrence frequency of the word in the document by the emotion value of each word by using the emotion value per word, and summing them.

In step 1135, the credit risk calculation module 120 calculates a credit risk based on a change in emotion value of each document, and uses the calculated credit risk to detect a danger signal before a credit event occurs. At this time, the emotion value for each day can be derived by using the average value of emotion value for each document calculated above.

For example, the credit risk calculation module 120 can calculate the credit risk through the trend of the change in emotion value per document, that is, whether it is maintained for three days or more.

The credit risk according to an exemplary embodiment of the present invention represents an integrated monitoring indicator utilizing the sensitivity score (emotion value) and the trend of the emotional score at the level of text and document using collected news and social data. Therefore, credit risk indicators can be used to define signals that could pose risks to future investments.

 Up to this point, a method for evaluating the credit risk of a listed company using FIG. 11 has been described.

Hereinafter, a method for evaluating a credit risk for a private company will be described in detail with reference to FIG.

In step 1210, the credit risk calculation module 120 defines an emotion dictionary.

Since unlisted companies without market price information are difficult to allocate emotion value from the stock price perspective, they will define emotional dictionaries for terms that are frequently used in financial markets and use them separately.

For example, based on all company-related opinions registered in portal finance such as Naver, the emotional value of each opinion is assigned first, and the emotion value of each word is assigned through the Naïve Bayes classification. If the document is negative, the negative emotion value is assigned to the words in the document, and the process is repeated for the entire document, whereby the emotion score for each word can be derived.

An example of assignment of emotion value by opinion is shown in Fig. A part of the predefined emotion dictionary based thereon is shown in Fig.

As shown in FIG. 14, each word in the emotion dictionary is divided into positive and negative words, and each word has emotion value.

In step 1215, the credit risk calculation module 120 extracts a meaningful word from the collected data, and then derives and embeds a vector value of the extracted words.

In step 1220, the credit risk calculation module 120 calculates the emotion value for each word / document by using the emotion dictionary.

For example, the credit risk calculation module 120 may propagate emotion values by word using a predefined emotion dictionary. Then, the emotion value for each document can be derived by using the emotion value for each word.

In step 1225, the credit risk calculation module 120 calculates a credit risk based on a change in emotion value of each document, and uses the calculated credit risk to detect a danger signal before a credit event occurs. At this time, the emotion value for each day can be derived by using the average value of emotion value for each document calculated above.

For example, the credit risk calculation module 120 can calculate the credit risk through the trend of the change in emotion value per document, that is, whether it is maintained for three days or more.

Referring again to FIG. 1, the credit event occurrence probability prediction module 125 predicts the probability of occurrence of an individual credit event through logistic regression analysis for companies with high credit risk.

FIG. 15 shows a detailed configuration of the credit event occurrence probability prediction module 125, and a method of predicting a credit event occurrence probability based on the credit event occurrence probability prediction module 125 is shown in FIG.

This will be described in more detail with reference to FIG.

In step 1610, the credit event occurrence probability prediction module 125 derives a signal from the credit risk viewpoint by using a trend (variation) of the emotion value per line. 17, signals according to an embodiment of the present invention are defined.

A signal for maintaining positive or negative for a predetermined period of time (for example, three days), a signal for increasing / decreasing the daily average sensation value, a signal for crossing the daily average sensation value, a ratio signal of the sensibility value per document, Weekly emotion value average increase rate signal per document can be respectively defined.

In step 1615, the credit event occurrence probability prediction module 125 defines an effective signal at the industrial level using the derived signal.

Collect data from companies in the same industry and perform logistic regression analysis to construct a regression equation using only the influential signals (effective signals). For example, through the logistic regression analysis, the validity signals that affect credit events are defined differently for each credit event and can be used to construct forecast models for each industry and credit event. .

In step 1620, the credit event occurrence probability prediction module 125 predicts a credit event occurrence probability (probability) using a signal derived for the analysis object company.

For example, the credit event occurrence probability prediction module 125 predicts the probability of a credit event occurrence by inputting a signal corresponding to an analysis target company in a credit event prediction model (i.e., a regression formula) corresponding to the analysis target company have.

As described above, the intelligent securities investment decision support apparatus 100 according to an embodiment of the present invention collects online and SNS data for an analysis target company, and then, using Opinion Mining and machine learning, By detecting risk signals, it is possible to make forecasts in the highly volatile stock market.

In addition, the intelligent securities investment decision support apparatus 100 according to FIG. 1 may be a computer. If the intelligent securities investment decision support apparatus 100 is a computer, it may include a memory and a processor, and each of the configurations described in FIG. 1 may be a logical configuration performed in the processor.

The apparatus and method according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions to be recorded on a computer-readable medium may be those specially designed and constructed for the present invention or may be known and available to those of ordinary skill in the computer software arts. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Includes hardware devices specifically configured to store and execute program instructions such as magneto-optical media and ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like.

The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

The embodiments of the present invention have been described above. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

Claims (12)

In the intelligent securities investment decision support method,
(a) collecting at least one of a news, an SNS, a report, a disclosure data, financial statement data, a posting in a financial community board, and a comment related to an analysis target company;
(b) screening a risk company using at least one of the collected disclosure data, financial statement data, and news;
(c) deriving emotion values for each word and each document by using the news and SNS data, and deriving a credit risk through a variation of emotion values for each word and document; And
(d) deriving a signal based on the derived credit risk, and predicting a credit event occurrence probability using the derived signal.
The method according to claim 1,
The step (b)
If the analysis target company is a listed company, deriving the disclosure index and the financial index using the disclosure data and the financial statement data;
Applying the derived known index and financial index to the learned support vector machine (SVM) to classify the normal index and the financial index into a normal company and a risk company;
Analyzing the occurrence frequency of the predetermined credit event keyword in the collected data to secondarily derive the risk company; And
And selecting a final analysis candidate company based on the first classification and the second derivation result.
3. The method of claim 2,
The above credit event is a credit event, which includes capital increase, capital erosion, corporate bond, disposal of treasury stock, liquidation merger, segregation accounting, free capital reduction, private equity fund, sale of largest shareholder or lump sum shareholder, debt guarantee for others, The occurrence of non-payment of interest on the bond, the disclosure of the inquiry, and the delayed report of the reduction of taxes.
The method according to claim 1,
The step (b)
Wherein if the analysis target company is a non-listed company, the frequency of occurrence of a credit event keyword is extracted from the collected news and disclosure data, and the risk company is screened.
The method of claim 1, wherein the step (c)
If the analysis target company is a listed company, assigning a sensitivity value using a stock price for each opinion included in the collected data;
Selecting key keywords based on positive and negative based on the sensibility polarity value;
Selecting a meaningful word from the collected data to create a data set, deriving and embedding a vector value of each word included in the data set,
Deriving a distance using the vector value between the core keyword and the word in the vector space and deriving an emotion value of each word based on the derived distance;
Deriving emotion values for each document by totaling the emotion values of the words included in the document; And
And calculating a credit risk using the derived fluctuation of emotion value for each document.
The method of claim 1, wherein the step (c)
Defining an emotional dictionary when the analyzed entity is a non-listed company;
Selecting a meaningful word from the collected data to create a data set, deriving and embedding a vector value of each word included in the data set,
Deriving an emotion value for each word in the vector space using the emotion dictionary;
Deriving emotion values for each document by summing emotion values of the words included in the document; And
And calculating a credit risk using the derived fluctuation of emotion value for each document.
The method according to claim 6,
The step of defining the emotional dictionary comprises:
Collecting data expressing a clear emotion for each opinion; And
And deriving an emotion value of each word using the naive bayes classifier for the collected data.
The method according to claim 1,
In the step (d) above,
Wherein the intelligent securities investment decision support method is derived using the daily fluctuation trend of emotion value of each document.
9. The method of claim 8,
The derived signal includes a signal for maintaining positive or negative for a predetermined period of time (for example, three days), a signal for increasing or decreasing the daily average sensibility value, a signal for crossing the daily average sensibility value, Value ratio signal and a weekly emotion value average increase rate signal for each document.
The method according to claim 1,
The step (d)
Collecting data of companies in the same industry and performing logistic regression analysis to check whether a predetermined credit event occurs within a predetermined period to derive effective signals for each industry;
Constructing a regression equation using the industry specific effective signal; And
And estimating a credit event occurrence probability by reflecting the derived signal in a regression equation of the same kind of industry.
Readable recording medium having recorded thereon a program code for performing the method according to any one of claims 1 to 10.
1. A computer device in which an intelligent securities investment decision support method is implemented,
A communication unit;
A memory in which at least one instruction is stored; And
A processor for executing instructions stored in the memory,
The instructions executed by the processor,
(a) collecting at least one of a news, an SNS, a report, a disclosure data, financial statement data, a posting in a financial community board, and a comment related to an analysis target company;
(b) screening a risk company using at least one of the collected disclosure data, financial statement data, and news;
(c) deriving emotion values for each word and each document by using the news and SNS data, and deriving a credit risk through a variation of emotion values for each word and document; And
(d) deriving a signal based on the derived credit risk, and predicting a credit event occurrence probability using the derived signal.

Images