KR20210127512A - System for small business credit loan based on artificial intelligence, and method thereof - Google Patents

Abstract

The present invention relates to a system for small business credit loans based on artificial intelligence and a method thereof. The system for small business credit loans based on artificial intelligence at least comprises: a credit evaluation unit collecting financial and non-financial data of small businesses including establishment sales, work history, and financial institution loan, building a credit evaluation model, and calculating the credit rating of the small businesses by using the built credit evaluation model; an alternative information generation unit evaluating an online site information, which advertises items of small businesses as well as social network service (SNS) activity information considerably affecting the sales of the small businesses, and reflecting the results in the credit rating in accordance with the credit evaluation model; a loan interest rate calculation unit calculating the loan interest rate for each credit rating, which is lower than that of the loan companies and which is outputted from the credit evaluation unit, by using the interest rate formed in the market; and a loan amount calculation unit calculating a loan amount based on the future cash flow information of the small businesses. The present invention aims to provide a system for small business credit loans based on artificial intelligence and a method thereof, which are capable of rapidly raising funds.

Description

AI-based small business credit loan system and method

The present invention relates to a credit loan, and more particularly, to an artificial intelligence-based small business credit loan system and method.

Domestic SMEs not only account for approximately 90% of total employment, but also account for approximately 60% of added value. As such, the country plays a pivotal role in the national economy, and as a major source of overall economic activity and employment, it has a significant impact on the national economy. However, in terms of business competitiveness and management financing compared to large corporations, the situation is poor, and in particular, the bank's avoidance of loans continues to be pointed out.

In particular, because of the large initial investment, small business owners have larger loans than borrowers who have only loans for individual businesses classified as SME loans. is at risk of becoming serious. Lack of credit, that is, lack of financial information, is pointed out as the main cause. In particular, in the case of small business owners, this phenomenon is more serious due to the lack of liquidity cash flow information along with reliable financial statements.

In order to solve this problem, it is required to develop a credit evaluation model for small business owners using big data and artificial intelligence. Various changes are being attempted in the field of corporate credit evaluation as the scope of big data utilization and the use of machine learning analysis methods are expanding around the world. Such attempts can be made more actively through Fintech companies. However, it is necessary to apply innovative technology to credit risk assessment through fintech, which is known to be more effective in the microfinance market, to provide reasonable financing to SMEs and small business owners with weak credit who are marginalized from existing financial services. .

Registered Patent Publication No. 10-1864006, 2018.05.28)

The problem to be solved by the present invention is to enable SMEs or small business owners with weak credit who are alienated from existing financial services to obtain reasonable interest rates and prompt financing with an appropriate loan amount rather than the existing high interest rates. For small business owners, financial information and various non-financial information are learned through a credit evaluation model built on the basis of AI technology, and small business credit evaluation is performed using the information of previously secured small business service members. It is to provide a credit loan system and method for small businesses based on artificial intelligence that enables rapid credit loans at reasonable interest rates.

The artificial intelligence-based small business credit loan system according to a temporary example of the present invention for achieving the above technical task builds a credit evaluation model by collecting the financial and non-financial data of at least small business owners, including business sales, business history, and borrowings from financial institutions. and a credit rating unit for calculating the credit rating of a small business by using the constructed credit rating model; An alternative information generating unit that evaluates SNS (Social Network Service) activity information that significantly affects the sales of small businesses and online site information that advertises items of small businesses and reflects them in the credit rating according to the credit evaluation model; a loan interest rate calculation unit that is lower than the loan business by using the interest rate formed in the market and calculates a loan rate for each credit rating output from the credit evaluation unit; and a loan amount calculation unit for calculating a loan amount based on the future cash flow information of the small business owner.

The credit evaluation unit includes: a data collection unit for collecting financial and non-financial data of small business owners, including business place sales, business history, borrowings from financial institutions, monthly rent or rental deposit for business sites, and asset information outside the business place; Filtering out data that does not fit the credit rating model through data verification including missing value processing, outlier value processing, and data standardization processing for the collected small business financial and non-financial data a data verification unit that performs; and a model for calculating a credit rating model by selecting an independent variable candidate of the credit rating model from among the verified data and performing group classification, chi-square test, and t-test on the independent variable candidate to generate a statistically significant independent variable Includes output.

The model calculation unit generates a primary independent variable by performing group classification, chi-square verification, and t-test on the independent variable candidate, and determines the final independent variable by checking the stepwise selection method and multicollinearity for the primary independent variable. The stepwise selection method includes constructing a credit evaluation model through a mechanical variable addition operation based on the p-value while including the variables one by one in the model through a logistic regression model. The credit rating unit continues to collect financial and non-financial data of small business owners receiving loan services even after the credit rating model is built, and continues to upgrade the parameters of the credit rating model using the collected data whenever new data is collected. It is characterized by building an AI learning-based credit evaluation model that is learned by a model suitable for loan services.

The alternative information generation unit collects SNS (Social Network Service) activity information that significantly affects the sales of small business owners, including reputation and awareness (number of posts, number of followers, etc.) SNS information generation unit for scoring; and a subscription information generator for evaluating and scoring online site information (number of subscribers, number of visitors, number of posts, etc.) for introducing and advertising sales items and services of the small business owner.

The artificial intelligence-based small business credit loan system according to an exemplary embodiment of the present invention provides, for the small business subscribing to the loan service, the small business's financial, and a loan service DB for storing non-financial information, wherein the loan amount calculation unit generates net profit information and future cash flow information of a small business using the loan service DB, and referring to the future cash flow information, It is characterized in that the loan information is calculated by re-calculating the loan evaluation.

The artificial intelligence-based small business credit loan method according to a temporary example of the present invention for achieving the above technical task collects financial and non-financial data of small business owners, including at least the business sales, business history, and borrowings from financial institutions, of small business owners to group classification, chi-square constructing a credit rating model by generating statistically significant independent variables by performing verification and t verification; Social Network Service (SNS) activity information including social network service (SNS) reputation and awareness of sales items and services of small business owners and online site information that advertises small business items are evaluated and reflected in the credit rating according to the credit rating model. calculating a credit rating; calculating a loan interest rate for each credit rating at a lower interest rate than the loan business based on the interest rate formed in the market; and calculating the loan amount based on the future cash flow information of the small business establishment and the credit rating.

The loan amount calculation includes the steps of: collecting financial and non-financial information of small business owners, including business place sales information, purchase information, fixed assets, loan information, and sales and administrative expenses information after loan; generating future cash flow information using the collected data; and calculating the additional loan amount of the pre-loan small business using the future cash flow information.

According to the artificial intelligence-based small business credit loan system and method according to the present invention, financial information and various non-financial information are learned through a credit evaluation model built based on AI technology for small business owners lacking objective information, and secured small business By using the service member's information to evaluate the credit rating of small business owners, it is possible to quickly get a credit loan at a reasonable interest rate without requesting additional information as well as a high recovery rate through an appropriate loan amount and period.

And, according to the present invention, it is possible to adjust interest rates and provide additional loans by using cash flow information and net profit information or other financial institution usage information according to sales and purchases of continuous service members (customers).

In addition, according to the present invention, SMEs or small business owners with weak credit, which are excluded from existing financial services, can obtain not only a reasonable interest rate, but also a reasonable interest rate, with an appropriate loan amount, without submitting additional additional information.

1 is a block diagram showing the configuration of an artificial intelligence-based small business credit loan system according to an embodiment of the present invention.
2 is a block diagram showing an example of a detailed configuration of a credit evaluation unit that constructs a credit evaluation model that calculates and evaluates the credit of small business owners through the AI technique.
3 shows the data collection and processing flow of the credit evaluation unit.
4 shows variables defined for constructing a credit rating model.
FIG. 5 shows the properties of the variables defined in FIG. 4 .
6 is a block diagram showing a more detailed configuration of the alternative information generating unit.
7 shows the data processing flow of the alternative information generating unit.
8 shows an example of a score evaluation method according to alternative information.
9 shows an example of the interest rate setting process of the loan interest rate calculator.
10 shows an example of the loanable amount calculation process of the loan amount calculation unit.
11 shows a process for providing feedback after lending.
12 shows an example of a feedback detailed calculation process after loan.
13 shows the integrated data processing flow of each stage of the AI-based small business credit loan system according to the present invention.
14 is a flowchart illustrating an AI-based small business credit loan method according to the present invention.
15 is a flowchart illustrating in detail calculation of a loan amount after loan.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The configuration shown in the embodiments and drawings described in this specification is only a preferred embodiment of the present invention, and does not represent all the technical spirit of the present invention, so various equivalents that can be substituted for them at the time of the present application It should be understood that there may be variations and variations.

The artificial intelligence-based small business credit loan system according to the present invention evaluates credit when an existing small business customer or a new small business customer needs a loan. The final evaluation score is calculated by reflecting it in the evaluation. Then, it is a system that determines the final interest rate and loan period by reviewing the market loan interest rate information compared to the amount at the time of loan application and determining the appropriate loan interest rate according to the evaluation score.

If a small business customer who has applied for a loan agrees to the loan amount, loan interest rate, and loan period in this way, the loan is made promptly, and the interest rate is lowered by using the customer's sales and net profit information, cash flow information, or other financial institution usage information. They can make adjustments and provide additional loans.

An existing small business customer or a new small business customer subject to the credit loan service according to the present invention means a module having a function to settle sales for small business or a module for managing cash flow and profit and loss for small business A service having functions such as Refers to a small business customer registered on the server or a new small business customer. The present invention provides a credit loan service for existing small business customers or new small business customers. These customers already have basic information about small businesses stored, and information such as sales and cash flow are also calculated and stored. Therefore, credit evaluation can be performed through the credit evaluation module for small business owners without the need to request additional information from small business owners.

1 is a block diagram showing the configuration of an artificial intelligence-based small business credit loan system according to an embodiment of the present invention. 1, the artificial intelligence-based small business credit loan system according to an embodiment of the present invention includes a credit evaluation unit 110, an alternative information generation unit 120, a loan interest rate calculation unit 130 and a loan amount calculation unit ( 140), and may further include a loan service DB (140).

The credit evaluation unit 110 collects the financial and non-financial data of at least small business owners including business sales, business history, and borrowings from financial institutions to build a credit evaluation model, and the credit rating of the small business using the constructed credit evaluation model. to calculate The credit evaluation unit 110 continuously collects financial and non-financial data of small business owners receiving loan services even after the credit evaluation model is built, and uses the collected data whenever data is newly collected to determine the parameters of the credit evaluation model. It can be built as an AI learning-based credit evaluation model that is continuously upgraded and learned as a model suitable for loan services. That is, the credit evaluation unit 110 is a module that calculates credit rating through credit evaluation of small business owners through the AI technique, and the calculated credit evaluation model updates the parameters of the credit evaluation model when existing customer information is added to provide the latest credit evaluation model can be maintained.

The credit evaluation unit 110 may construct a credit evaluation model using machine learning. The machine learning refers to a series of algorithms or processing processes in which a computer is not programmed in advance and performs an appropriate operation on new data by learning a pattern from data. Machine learning techniques include artificial neural networks, decision tree learning methods, genetic algorithms, genetic programming, Gaussian process regression, linear fractional analysis). For the description of each model, reference may be made to general statistical books, so a detailed description thereof will be omitted. In addition, machine learning can be divided into supervised learning, unsupervised learning, and reinforcement learning according to the type of training data input to the learning system.

Supervised learning means that an input variable and a target variable exist, while unsupervised learning is a method of learning data that does not have an answer to the input result when only the input variable exists and there is no target variable. And reinforcement learning is a learning technique that allows the agent defined in a specific environment to recognize the current state and select the action or action sequence that maximizes the reward among the selectable actions. The credit evaluation unit 110 may learn through supervised learning because the target variable is whether or not it corresponds to insolvency through various input variable data.

2 is a block diagram showing an example of the detailed configuration of the credit evaluation unit 110 to construct a credit evaluation model that calculates and evaluates the credit of small business owners through the AI technique. Referring to FIG. 2 , the credit evaluation unit 110 includes a data collection unit 210 , a data verification unit 220 , and a model calculation unit 230 . 3 shows the data collection and processing flow of the credit evaluation unit 110 .

An example of a credit evaluation model of an AI-based credit evaluation unit according to the present invention will be described with reference to FIGS. 2 and 3 . The data collection unit 210 collects the financial and non-financial data of the small business, including sales, business history, borrowings from financial institutions, monthly rent or rental deposit of the business place, and asset information outside the business place of the small business owner. When describing the data collection unit 210 in more detail, the data collection unit 210 first collects past data. (S310) Past data may include small business information, business place financial information, non-business site asset information, other loan information, non-financial information, and insolvency information. The past data may utilize information of a credit rating agency.

The distribution of accidents (e.g., bankruptcy, bankruptcy, etc.), which is a dependent variable, as a dependent variable through the past data of small business owners, is modeled through the data of the borrower who has never had an accident (no accident) and the borrower who has had at least one accident (the existence of an accident). can be calculated.

Looking at the process of significance verification and model calculation (S330) in detail in FIG. 3, the final variable selection task is performed using KS test, chi-square test, and t test for independent variables that can be obtained as target data for model construction. Independent variables are selected. In the process, the KS test, which tests normality, is first performed, and it can be judged as refined data necessary for analysis by checking whether the target independent variables have a normal distribution. For example, data outside the normal distribution can be judged as meaningless data. Then, to determine whether each independent variable is data that occurred by chance, the subject is extracted through a chi-square test. Finally, through the t-test, it is determined whether the data is composed of data that reflects the characteristics of the population.

Afterwards, in the data analysis, the final independent variables are selected by measuring the multicollinearity of whether there is a correlation between each independent variable. Multicollinearity refers to a phenomenon in which some predictors of the model used in regression analysis have a high degree of correlation with other predictors, which negatively affects data analysis. For example, for variables with multicollinearity, logistic regression analysis is performed to select these variables because the variables with larger regression coefficient estimates, Wald chi-square values, correct classification rates, and c statistic values have greater discriminating power. Do it.

In addition, the final independent variable is selected by performing a method of finding mechanically significant variables through a statistical technique called stepwise selection method. That is, through the logistic regression model, one variable is included in the model, and if the p-value is above a predetermined value, it is maintained as it is, and other explanatory variables are repeatedly added to complete the model.

The final created logistic regression model should be analyzed for performance evaluation. To this end, we evaluate the performance of the model by measuring the accuracy and precision of confirming the ratio of the error classification method and the classification model to the actual type.

4 shows variables defined for constructing a credit rating model. That is, variables for constructing a credit evaluation model can be defined as shown in FIG. 4 . For reference, in the case of commercial area information, which is important to small business owners in FIG. 4, there may be difficulties in defining and obtaining data, and there is a step in which the actual future cash flow is reflected when calculating the loanable amount on the AI credit loan system. Commercial area information was excluded from the independent variable.

Referring to FIG. 2 , after the data collection unit 210 collects past data and defines independent variables, the data verification unit 220 performs data verification and purification (S320). The data verification unit 220 collects the data. We filter out data that do not fit into the credit rating model through data verification including missing value processing, outlier value processing, and data standardization processing for financial and non-financial data of a small business owner. . That is, the data verification unit 220 may perform missing value processing, outlier value processing, and data standardization processing for data verification and purification. (S320) Here, missing value processing and outlier value processing The data to be processed means, for example, that the actual input data must be an integer, but it refers to data that contains a negative value, a character string, or a non-existent value. Since such data can affect the calculation of the credit rating model, it is necessary to have a treatment plan, and this operation is called data purification operation.

Variable data of actual small business owners may have significant bad data in nature. These data will continue to exist in the future, but since there are no missing values in the dependent variable (insolvency) data of small business owners who have these variables, they can all be used as utility data. However, for example, when there is no value on a specific independent variable, there are various methods such as reflecting the average of the variables in statistical techniques for data standardization for these data, but in the credit rating model, the dependent variable ( Non-significant variables for default) are excluded and a model can be created. Therefore, in this case, it is treated as 0.

FIG. 5 shows the properties of the variables defined in FIG. 4 , and data processing can be performed by defining the properties of the variables as shown in FIG. 5 . An embodiment of the present invention is to establish a credit evaluation model for the restaurant business by limiting the industry to the restaurant business. When the industry is limited to the catering industry, characteristics of the catering industry may be reflected.

The characteristics of the catering industry are many individual entrepreneurs, and the employees are mainly relatives or part-timers. The catering business requires a lot of money to equip the initial interior and kitchen equipment, and it is mainly financed by loans from financial institutions through personal collateral loans. Until recently, local commercial districts were important to generate sales, but geographic location is moving away from an important factor in sales due to excessive rent and activation of food delivery. In particular, food taste and service are important, and it is a characteristic of the catering industry that, depending on how active social network service (SNS) with these contents is in addition to improved financial information, it can significantly contribute to sales.

In the present invention, the credit evaluation model in the credit evaluation unit 110 based on AI can be developed by setting the financial part as a main variable item, and the embodiment according to the present invention adds alternative information to small business owners to credit evaluation. It is possible to build a credit rating system in the form of

Meanwhile, referring to FIG. 2 , past data is collected through the data collection unit 210 , and after data verification and purification are performed through the data verification unit 220 , the model calculation unit 230 verifies the significance of the data and Calculate the credit rating model. In the significance verification and model calculation step (S330), final variable selection, AI model calculation, and credit score grading are performed. (S330)

The model calculation unit 230 selects an independent variable candidate of the credit evaluation model from among the verified data, performs group classification, chi-square verification, and t verification on the independent variable candidate to generate a statistically significant independent variable, Calculate the evaluation model. The model calculation unit 230 generates a primary independent variable by performing group classification, chi-square verification, and t verification on the independent variable candidate, and confirms the stepwise selection method and multicollinearity for the primary independent variable, and finally Independent variables are selected, and the stepwise selection method builds a credit evaluation model by adding variables based on p-value while including variables one by one through a logistic regression model.

The model calculating unit 230 will be described in more detail. The final variable selection of the model in the model calculation unit 230 is based on the statistically selected results when selecting the independent variables to construct the credit evaluation model, considering the suitability of the business point of view when lending, and combining the optimal variables. In this case, there are various methods for selecting statistically significant variables. In an embodiment of the present invention, the final independent variable can be selected in the following way.

The independent variable can be selected using group classification, chi-square test, and t-test for the independent variable (refer to FIG. 4) for model construction. Through this, a logistic regression model is calculated through a statistical program (S330). When the model is calculated, scores can be graded through past sample data, and 6 groups are classified based on this score. However, it should be checked whether a phenomenon occurs that results worse than the low-scoring group, even though the score is high, as the score result is lowered.

As a result, when data (independent variables) of new and existing customers are input according to the AI credit rating model, a model that can calculate ratings is completed. This model is a system in which the parameters of the model are upgraded when continuous customer data is obtained. If the existing credit rating agency develops a specific model, it is continuously updated while studying the periodicity of the model. However, the credit rating model according to the present invention is The parameters of the model are updated every time it occurs, so that the model is always up-to-date. In this way, rather than creating and calculating a model limited to data from other credit rating agencies in the past, customer information about the loan service according to the present invention is continuously updated and developed into a model suitable for the service, becoming the most recent model suitable for the service. will be.

Meanwhile, referring to FIG. 1 , in addition to calculating a credit rating through the financial and non-financial data of the small business in the credit rating unit 110 , active activities such as SNS may be used for credit rating. This is to reflect the fact that the direction of future sales changes very dramatically depending on the active business form, even if the credit rating of small business owners is good. In addition, financial and non-financial data for credit evaluation are judged based on data generated in the past, which is insufficient to determine future business directions. For this, subscription information and SNS usage information of small business owners can be reflected as input variables. When subscription information and SNS usage information are input, the final score is reflected through a point system, and this score is reflected as a weight in the credit rating score.

The alternative information generating unit 120 evaluates SNS (Social Network Service) activity information that significantly affects the sales of small businesses and online site information that advertises items of small business owners, and reflects them in the credit rating according to the credit rating model.

6 is a block diagram showing a more detailed configuration of the alternative information generating unit 120 , and FIG. 7 is a data processing flow of the alternative information generating unit 120 . Referring to FIG. 6 , the alternative information generation unit 120 includes an SNS information generation unit 610 and a subscription information generation unit 620 .

SNS information generation unit 610 is a social network service (SNS) activity that significantly affects the sales of small business owners, including reputation and awareness (number of posts, number of followers, etc.) Information can be collected and scored. The subscription information generator 620 may score by evaluating online site information (number of subscribers, number of visitors, number of posts, etc.) that introduces and advertises sales items and services of the small business owner.

Subscription information defines sites that can introduce the business, and may include membership registration period information and information on other subscription companies. have. (S710) In order to reflect the alternative information including subscription information and SNS usage information in the credit evaluation model, the correlation with the insolvent information is calculated, the alternative information is scored by AI, and the alternative information is scored by calculating the weight through the AI score. You can. (S720)

When the score according to the alternative information is calculated in this way, it can be reflected as a weight in the credit evaluation model for each score range. 16 shows an example of a score evaluation method according to alternative information. The alternative information is calculated by scoring subscription information (member registration information, other subscription company information) and SNS use information (number of posts, number of followers) through this. It can be reflected as a weight in the credit rating score. Therefore, it is desirable that alternative information is reflected as a weight in the credit evaluation calculation score, rather than changing the credit evaluation model. Using a predetermined weighted rate, for example, if the credit rating score is 800 points, the rate is 0.5, and the alternative information score calculated according to the grade is 90 points, the final score will be determined as 845 (=800+(0.5 *90)). can

Referring to FIG. 2 , the data collection unit 210 for the small business subscribing to the loan service, the small business's financial and non-financial information including sales information, purchase information, fixed assets, loan information, and sales and management cost information of the small business establishment. When information is collected, the collected data is stored and managed in the loan service DB 150 . At this time, the data collection unit 210 may collect data through the credit rating agency 212 and the financial institution 214 through the network by obtaining the consent of the small business owner subscribed to the loan service, the POS system 218 and the delivery agency Data may also be collected from service providers 216 .

Next, it is the stage of setting the loan interest rate range according to the information on the market interest rate, which is a very important stage that can affect the profit and loss of the lending company. If the loan interest rate is high, the loan interest rate can be high, so you can make a profit, but the loan generation can be reduced by that much. In addition, the service for small business owners has an image of a social enterprise service, so the service reputation may change depending on the loan interest rate. However, if the loan interest rate is low, the loan profit will be lowered, and if the risk of insolvency of small business owners with poor credit quality increases, there may be a situation in which the loan loss provisions need to be accumulated by that much. In order to determine such an appropriate loan interest rate, it is necessary to determine an appropriate interest rate in consideration of interest rates formed in the market.

Referring to FIG. 1 , the loan interest rate calculator 130 calculates a loan interest rate for each credit rating output from the credit evaluation unit, which is lower than the loan business by using the interest rate formed in the market. The AI-based credit loan system according to the present invention is higher than the bank loan interest rate and if the interest rate is determined while maintaining the level of the savings bank loan interest rate, it can provide efficient interest rates to small business owners through lower interest rates than card companies, capital companies, and loan businesses. have.

9 shows an example of the interest rate setting process of the loan interest rate calculation unit 130, after obtaining the reference rate information, the bank note loan interest rate information, and the savings bank loan interest rate information in order to refer to the market interest rate information (S910) , to set the loan interest rate range, define the area for each medium interest rate section and define the loan interest rate compared to the market interest rate (S920). That is, the AI-based credit loan system according to the present invention is the middle The interest rate can be determined and, through this, the interest rate can be determined for each grade section on the AI-based credit loan system. Based on this, it is possible to adjust the interest rate by periodically using data such as the evaluation of loan income and provision for bad debts according to insolvency.

Next, in the loanable amount calculation step, the loanable amount can be considered from two aspects. One is when a customer who wants a loan directly applies for the loan amount, and the other is the loanable amount in the AI-based credit loan system. case is presented.

The loan service to which the present invention is applied secures sales, purchases, cash flow and profit and loss estimation information of small business customers who are the target of credit loan services. Using this information, it is possible to provide a service that periodically calculates the loanable amount for all subscribers and presents the available amount first. Therefore, based on the loanable amount, we provide a service that allows customers to borrow an amount within the range they want to borrow.

Referring to FIG. 1 , the loan amount calculation unit 140 calculates the loan amount based on the future cash flow information of the small business owner. The loan amount calculation unit 140 generates the net profit information and future cash flow information of the small business using the loan service DB 150, and calculates the loan evaluation of the previous borrower again with reference to the future cash flow information to obtain the loan information. can be calculated.

FIG. 10 shows an example of the loanable amount calculation process of the loan amount calculation unit 140, and calculates future cash flows using sales/purchase information and fixed assets/loan information (S1010), the current loan It is possible to calculate the possible amount and calculate the maximum loanable amount by the maximum limit and future time point. (S1020) That is, to calculate the loanable amount, sales and purchase information, sales and purchase information, The loanable amount is calculated while generating future cash flows through fixed assets/loan information and evaluating the profit and loss. Under the premise that the main loan cause of the AI-based credit loan system is the lack of short-term liquid funds of the self-employed, it is possible to present the loanable amount by identifying the funds flowing in the mid- to long-term through future cash flows that are insufficient in the short term.

Through these four steps, a system and methodology for making and evaluating AI small business loans will be equipped, and a feedback system will be equipped to periodically adjust the additional loan amount and interest rate according to the loan according to the periodic evaluation after the loan. .

Finally, in the step of updating the information of small business owners after the loan, the interest rate can be adjusted periodically (mainly quarterly) through the feedback system for such borrowers, and a service can be provided that also suggests the amount of additional loan available. 11 shows a process for providing feedback after lending, it is possible to determine whether to adjust an interest rate and make an additional loan. And Figure 12 shows an example of the feedback detailed calculation process after the loan. Referring to FIG. 11 , by collecting sales/net profit information and other financial institution usage information after the loan (S1110), it is possible to determine whether to adjust the interest rate and whether to make an additional loan (S1120). That is, the actual sales and net profit information after the loan Accordingly, future cash flows are updated on the profit and loss book, and loan evaluations for borrowers are recalculated, enabling interest rate adjustment and additional loan information to be calculated.

Referring to FIG. 12, a new credit rating rating is calculated while looking at the cash flow at each future point according to the flowchart, and through comparison between the loan amount and the loanable interest rate, it is determined whether additional adjustment should be made and a service to inform the borrower. can For example, after calculating net profit for 3 months, 6 months, and 1 year in the future (S1210) and obtaining current loan information (S1220), credit ratings for 3 months, 6 months, and 1 year may be calculated. (S1230) Then, the loanable amount is calculated for each 3 months, 6 months, and 1 year according to the credit rating (S1240) and the current loan amount and the future loan amount are compared. (S1250) Also, 3 months according to the credit rating grade, By calculating the loanable interest rate by 6 months and 1 year (S1260), it is possible to compare the current loan interest rate and the loan interest rate at each time in advance (S1270).

13 shows the integrated data processing flow of each stage of the AI-based small business credit loan system according to the present invention, and after calculating the credit rating and credit rating of the small business through the small business credit evaluation module through the A! technique (S1310), the small business alternative information is reflected. (S1320) Then, the loan amount and the loan interest rate are determined according to the evaluation score. (S1340) And after the loan, by updating the information of the small business owner, it is possible to adjust the loan interest rate and determine whether to make an additional loan. (S1350)

On the other hand, Figure 14 is a flowchart showing the artificial intelligence-based small business credit loan method according to the present invention. Referring to FIG. 22 , financial and non-financial data of at least small business owners including business sales, business history, and borrowings from financial institutions are collected through the data collection unit 210. (Step S1410) The data verification unit 220 Through group classification, chi-square verification, and t-test, a statistically significant independent variable is generated through the model calculation unit 230 to construct a credit evaluation model. (Step S1420)

Then, through the alternative information generating unit 120, SNS (Social Network Service) activity information including the reputation and awareness of the sales items and services of the small business owner and the online site information advertising the items of the small business owner are evaluated, and the credit evaluation The credit rating and credit rating of the small business are calculated by reflecting the credit rating according to the model. (Step S1430) Here, the credit evaluation model continuously records the financial and non-financial data of the small business owner receiving the loan service even after the credit evaluation model is built. It can be built as an AI learning-based credit evaluation model that is continuously upgraded with the parameters of the credit evaluation model using the collected data whenever new data is collected and learned as a model suitable for the loan service.

When the credit rating and credit rating for the small business are calculated, the loan interest rate calculation unit 130 calculates the loan interest rate for each credit rating at a lower interest rate than the loan business based on the interest rate formed in the market. (Step S1440) Also, the loan amount is calculated The unit 140 calculates the loan amount based on the future cash flow information of the business place of the small business and the credit rating. (Step S1450)

15 is a flowchart illustrating calculation of a loan amount after loan. Referring to FIG. 15 , for the loan small business, financial and non-financial information of the small business including business sales information, purchase information, fixed assets, loan information, and sales and administrative expense information after the loan is collected. (Step S1510) The collection Generates future cash flow information using the data obtained. (Step S1520) Using the future cash flow information, it is possible to calculate the additional loan amount of the pre-loan small business owner. (Step S1530)

Although the present invention has been described with reference to the embodiment shown in the drawings, which is merely exemplary, those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Accordingly, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

110: Credit evaluation unit 120: Alternative information generation unit
130: loan interest rate calculation unit 140: loan amount calculation unit
150: loan service DB 210: data collection unit
212: credit rating agency 214: financial institution
216: delivery agency service company 218: POS system
220: data verification unit 230: model calculation unit
610: SNS information generation unit 620: subscription information generation unit

Claims (9)

a credit evaluation unit that collects financial and non-financial data of small business owners, including at least the small business's business sales, business history, and borrowings from financial institutions, to build a credit evaluation model, and calculates the credit rating of the small business using the constructed credit evaluation model;
An alternative information generating unit that evaluates SNS (Social Network Service) activity information that significantly affects the sales of small businesses and online site information that advertises items of small businesses and reflects them in the credit rating according to the credit evaluation model;
a loan interest rate calculator that calculates a loan interest rate for each credit rating output from the credit evaluation unit, which is lower than the loan business by using the interest rate established in the market; and
Artificial intelligence-based small business credit loan system, including a loan amount calculation unit that calculates the loan amount based on the future cash flow information of the small business owner. According to claim 1, wherein the credit evaluation unit
A data collection unit that collects financial and non-financial data of small business owners, including business place sales, business history, borrowings from financial institutions, monthly rent or rental deposits of business sites, and asset information outside the business place;
Filtering out data that does not fit the credit rating model through data verification including missing value processing, outlier value processing, and data standardization processing for the collected small business financial and non-financial data a data verification unit that performs; and
Model calculation that calculates a credit rating model by selecting independent variable candidates of the credit rating model from the verified data and performing group classification, chi-square test, and t-test on the independent variable candidates to generate statistically significant independent variables Artificial intelligence-based small business credit loan system, characterized in that it includes wealth. The method of claim 2, wherein the model calculation unit
Group classification, chi-square verification, and t-test are performed on the independent variable candidates to generate a primary independent variable, and a stepwise selection method and multicollinearity are checked for the primary independent variable to select a final independent variable,
The stepwise selection method includes constructing a credit evaluation model through a mechanical variable addition operation based on the p-value while including the variables one by one in the model through a logistic regression model, an artificial intelligence-based small business credit loan system . According to claim 1, wherein the credit evaluation unit
Even after the credit evaluation model is established, financial and non-financial data of small business owners receiving loan services are continuously collected, and whenever data is newly collected, the parameters of the credit evaluation model are continuously upgraded using the collected data to match the loan service. Artificial intelligence-based small business credit loan system, characterized in that it builds an AI learning-based credit evaluation model that is learned by the model. The method of claim 1, wherein the alternative information generating unit
SNS information that collects and scores SNS (Social Network Service) activity information that significantly affects the sales of small business owners, including reputation and awareness (number of posts, number of followers, etc.) generator; and
Artificial intelligence-based small business credit loan system, characterized in that it comprises a subscription information generator that evaluates and scores online site information (number of subscribers, number of visitors, number of posts, etc.) that introduces and advertises sales items and services of the small business. According to claim 1,
Loan service DB for storing the financial and non-financial information of the small business, including sales information, purchase information, fixed assets, loan information, and sales and administrative expenses information of the small business establishment for the small business that has signed up for the loan service;
The loan amount calculation unit
Artificial intelligence-based small business, characterized in that by using the loan service DB to generate net profit information and future cash flow information of the small business, and recalculating the loan evaluation of the previous borrower with reference to the future cash flow information, to calculate the loan information credit loan system. Establish a credit evaluation model by generating statistically significant independent variables by collecting financial and non-financial data of small business owners, including at least their business place sales, business history, and borrowings from financial institutions, and performing group classification, chi-square test, and t-test to do;
Social Network Service (SNS) activity information including social network service (SNS) reputation and awareness of sales items and services of small business owners and online site information that advertises small business items are evaluated and reflected in the credit rating according to the credit rating model. calculating a credit rating;
calculating a loan interest rate for each credit rating at a lower interest rate than the loan business based on the interest rate formed in the market; and
An artificial intelligence-based small business credit loan method, comprising the step of calculating a loan amount based on the credit rating and future cash flow information of the small business establishment. The method of claim 7, wherein the credit rating model is
Even after the credit evaluation model is established, financial and non-financial data of small business owners receiving loan services are continuously collected, and whenever data is newly collected, the parameters of the credit evaluation model are continuously upgraded using the collected data to match the loan service. Artificial intelligence-based small business credit loan method, characterized in that it is an AI learning-based credit evaluation model that is learned by the model. The method of claim 8, wherein the calculation of the loan amount
Collecting financial and non-financial information of small business owners, including business place sales information, purchase information, fixed assets, loan information, and sales and administrative expenses information after loan, for pre-loan small business;
generating future cash flow information using the collected data; and
Artificial intelligence-based small business credit loan method, characterized in that it further comprises the step of calculating the additional loan amount of the pre-loan small business using the future cash flow information.

Images