KR20210150803A - Electronic system apparatus for recommending par value of discount coupon based on artificial intelligence and operating method thereof - Google Patents

Abstract

Disclosed are an electronic system apparatus for recommending a par value of a discount coupon based on artificial intelligence and an operating method thereof. The present invention relates to the electronic system apparatus and the operating method thereof, which can generate a recommending model that can recommend the par value of the discount coupon to be issued by a businessperson customer based on artificial intelligence.

Description

ELECTRONIC SYSTEM APPARATUS FOR RECOMMENDING PAR VALUE OF DISCOUNT COUPON BASED ON ARTIFICIAL INTELLIGENCE AND OPERATING METHOD THEREOF

The present invention relates to an electronic system device for recommending a face value of a discount coupon based on artificial intelligence and an operating method thereof.

Recently, business operators are using various marketing methods to maximize profits. One of these marketing methods is to issue a discount coupon. Discount coupons are a kind of securities that discount a certain amount from the price of goods.

As the types of discount coupons, there are a coupon to which a predetermined discount par value is applied, which applies a discount by a certain amount from the price of goods, and a coupon to which a discount is applied by a certain percentage from the price of the product.

In the case of a discount coupon to which a predetermined discount par value is applied, the minimum purchase price condition for applying the discount is often applied. For example, in the case of a discount coupon to which a discount par value of '5,000 won' is applied, conditions are often set so that it can be used only when purchasing an item of 'at least 50,000 won'.

As mentioned above, the purpose of issuance of discount coupons is to induce consumers to purchase and create profits. Therefore, considering the business type of business, the region where the business is located, the sales of the business, the scale of issuance of discount coupons, etc. It is necessary to issue an appropriate type of discount coupon to achieve the target rate of return.

In particular, in the case of issuing a discount coupon that applies a discount to a predetermined discount face value from the product price, it is issued in order to achieve a target rate of return in consideration of the business type, region, sales, and scale of issuance of the discount coupon mentioned above. It is necessary to properly determine the face value of the discount coupon you want to use.

For example, if the business operator is a business that mainly sells low-priced goods, and the region where the business operator is located is a region where there are many consumers who want to purchase low-priced goods, a discount coupon with a lower face value is issued rather than a discount coupon with a high face value. Since this can be beneficial to consumers' consumption inducement, business operators will be able to pursue high yield by issuing discount coupons with low face value.

In order to determine the par value of such a discount coupon, a method may be considered by referring to the industry, region, sales, par value information of the issued discount coupon, yield, etc. of other operators who have already issued discount coupons. That is, by referring to examples of business operators that have a history of issuing discount coupons, it will be possible to determine whether it is appropriate to issue a discount coupon with a face value.

However, general operators are not marketing experts, and it is difficult to access data such as industry, region, sales, and scale of issuance of discount coupons of other operators who have already issued discount coupons, so discount coupons with a certain face value are issued. It can be difficult to decide whether it is appropriate to do so.

In this regard, if a service that can recommend the par value of the discount coupon to be issued is provided to the business operator by complexly analyzing the business type, region, sales, scale of issuance of discount coupons, etc. By issuing discount coupons with reference to information on

On the other hand, recently, a machine learning-based artificial intelligence technology that can create a recommendation model for recommending a predetermined result based on some sample data has emerged. In this regard, even in recommending the face value of a discount coupon, a recommendation model that can recommend the face value of a discount coupon can be created by performing machine learning based on the information of other operators with a history of issuing discount coupons. .

Therefore, it is necessary to create an artificial intelligence-based recommendation model for recommending the face value of a discount coupon to business customers, and research on a technology that can recommend the face value of a discount coupon to business customers based on this is required.

The present invention generates a recommendation model capable of recommending the par value of a discount coupon that a business customer wants to issue based on artificial intelligence, and based on this, an electronic system device capable of recommending the face value of a discount coupon that the business customer wants to issue And by presenting its operation method, it is intended to support the recommendation of information on the face value of a discount coupon more suitable for business customers.

The electronic system device for recommending the face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention includes N (N is a natural number greater than or equal to 3) different discount coupon face values - Each discount coupon is used for each of the discount coupon face values Different predetermined minimum purchase amounts are specified for is a natural number of 2 or more) business customers - The M business customers are business operators who actually issued a discount coupon according to the face value of any one of the N discount coupon face values - Industry name information collected from each, area name Information, the sales amount for the first predetermined period, the total issuance amount of discount coupons, and input set information consisting of a fixed rate of return according to the issuance of discount coupons, and the M business operators as correct information for the input set information of each of the M business operators and customers M training sets in which the par value of the discount coupon actually issued by each customer - the par value of the discount coupon actually issued by each of the M business customers is one of the face values of the N discount coupons - is matched one by one is stored A training set storage unit, a plurality of predetermined industry name information and a predetermined different industry specific number for each of the plurality of industry name information is stored corresponding to the industry name information database, a plurality of predetermined area names In the business name information of each of the M business operators, with reference to the region name information database, in which the information and each of the plurality of region name information are stored in correspondence with each other, the region name information database, the business name information database An information confirmation unit that confirms a corresponding industry-specific number, and confirms a region-specific number corresponding to the region name information of each of the M business operator customers with reference to the region name information database; With reference to the input set information of each of the M business operators included in each of the M training sets, for each of the M business customers, each business operator customer's industry-specific number, region-specific number, sales amount, and discount coupon An input vector generator that generates a five-dimensional input vector corresponding to each of the M business customers by constructing a five-dimensional input vector with the total issuance amount and the fixed rate of return as components, and corresponding to each of the M business customers When the five-dimensional input vector is generated, by referring to the discount coupon par value information storage unit, the N-dimensional one-hot vector corresponding to the face value of the discount coupon actually issued by each of the M business customers is identified by checking the M A deep neural network composed of two or more weight matrices for a one-hot vector designation unit that specifies an N-dimensional one-hot vector corresponding to each of the business customers, and a 5-dimensional input vector corresponding to each of the M business customers. Two or more weight matrices constituting the neural network are matrices configured so that the dimension of a vector calculated as an output of the deep neural network becomes N-dimensional - by applying it as an input to generate an N-dimensional output vector corresponding to each of the M business operator customers When an N-dimensional output vector corresponding to each of the M business operator customers is generated, an N-dimensional output vector corresponding to each of the M business operator customers is used as a predetermined activation function - the activation function. The function is a function that converts the components constituting the N-dimensional output vector corresponding to each of the M business customers into values of 0 or more and 1 or less. An operation vector generating unit that generates an operation vector and an N-dimensional operation vector corresponding to each of the M operator customers and an N-dimensional one-hot vector corresponding to each of the M operator customers are compared with each other for the deep neural network. By performing machine learning, discount coupons for business customers and a recommendation model generator that generates a recommendation model for recommending a face value.

In addition, the operating method of the electronic system device for recommending the face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention includes N (N is a natural number greater than or equal to 3) different discount coupon face values - Each of the discount coupon face values includes Different predetermined minimum purchase amounts for using each discount coupon are designated - maintaining a discount coupon face value information storage unit in which different predetermined N-dimensional one-hot vectors for each are stored, M(M is a natural number of 2 or more) business customers - The M business customers are business operators who actually issued a discount coupon according to the face value of any one of the N discount coupon face values - Industry name information collected from each, area name Information, the sales amount for the first predetermined period, the total issuance amount of discount coupons, and input set information consisting of a fixed rate of return according to the issuance of discount coupons, and the M business operators as correct information for the input set information of each of the M business operators and customers M training sets in which the par value of the discount coupon actually issued by each customer - the par value of the discount coupon actually issued by each of the M business customers is one of the face values of the N discount coupons - is matched one by one is stored Maintaining the training set storage unit, a plurality of predetermined industry name information and maintaining the industry name information database that is stored in correspondence with each of the plurality of industry name information and a predetermined different industry unique number for each , maintaining a region name information database in which a plurality of predetermined region name information and predetermined different region identification numbers for each of the plurality of region name information are stored in correspondence with each other, with reference to the industry name information database Check the industry unique number corresponding to the industry name information of each of the M number of business customers, and refer to the region name information database to refer to the region corresponding to the local name information of each of the M business customers Confirming a unique number, with reference to the input set information of each of the M number of business customers included in each of the M training sets, for each of the M business customers, the industry-specific number of each business operator, the region Generating a five-dimensional input vector corresponding to each of the M business operators customers by constructing a five-dimensional input vector having a unique number, sales amount, total issuance amount of discount coupons, and a fixed rate of return as components; When a five-dimensional input vector corresponding to each customer is generated, an N-dimensional one-hot vector corresponding to the face value of the discount coupon actually issued by each of the M business operators by referring to the discount coupon par value information storage unit is checked By doing so, specifying an N-dimensional one-hot vector corresponding to each of the M business customers, a deep neural network composed of two or more weight matrices using a 5-dimensional input vector corresponding to each of the M business customers. Two or more weight matrices constituting the neural network are matrices configured so that the dimension of a vector calculated as an output of the deep neural network becomes N-dimensional - by applying it as an input to generate an N-dimensional output vector corresponding to each of the M business operator customers When an N-dimensional output vector corresponding to each of the M operator customers is generated, an N-dimensional output vector corresponding to each of the M operator customers is set to a predetermined activation function - the activation function is the M operator customers It is a function that transforms the components constituting the N-dimensional output vector corresponding to each of the values of 0 or more and 1 or less - generating an N-dimensional operation vector corresponding to each of the M business operator customers by applying it as an input; By performing machine learning on the deep neural network by comparing an N-dimensional operation vector corresponding to each of the M business customers and an N-dimensional one-hot vector corresponding to each of the M business customers to each other, discount of business customers Including the step of creating a recommendation model for recommending the coupon face value do.

The present invention generates a recommendation model capable of recommending the par value of a discount coupon that a business customer wants to issue based on artificial intelligence, and based on this, an electronic system device capable of recommending the face value of a discount coupon that the business customer wants to issue And by presenting its operation method, it is possible to support the recommendation of information on the face value of a discount coupon more suitable for business customers.

1 is a diagram illustrating a structure of an electronic system device for recommending a face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention.
2 is a flowchart illustrating an operation method of an electronic system device for recommending a face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. These descriptions are not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. While describing each drawing, like reference numerals are used for similar components, and unless otherwise defined, all terms used in this specification, including technical or scientific terms, refer to those of ordinary skill in the art to which the present invention belongs. It has the same meaning as is commonly understood by those who have it.

In this document, when a part "includes" a certain component, it means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, in various embodiments of the present invention, each of the components, functional blocks or means may be composed of one or more sub-components, and the electrical, electronic, and mechanical functions performed by each component are electronic. A circuit, an integrated circuit, an ASIC (Application Specific Integrated Circuit), etc. may be implemented as various well-known devices or mechanical elements, and may be implemented separately or two or more may be integrated into one.

On the other hand, the blocks in the accompanying block diagram or steps in the flowchart are computer program instructions that are loaded in a processor or memory of equipment capable of data processing, such as a general-purpose computer, a special-purpose computer, a portable notebook computer, and a network computer, and perform specified functions. can be interpreted as meaning Since these computer program instructions may be stored in a memory provided in a computer device or in a computer-readable memory, the functions described in the blocks of the block diagram or the steps of the flowchart are produced as articles of manufacture containing instruction means for performing the same. it might be In addition, each block or each step may represent a module, segment, or portion of code comprising one or more executable instructions for executing the specified logical function(s). It should also be noted that, in some alternative embodiments, it is also possible for the functions recited in blocks or steps to be executed out of the prescribed order. For example, two blocks or steps shown one after another may be performed substantially simultaneously or in the reverse order, and in some cases, some blocks or steps may be omitted.

1 is a diagram illustrating a structure of an electronic system device for recommending a face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention.

Referring to FIG. 1 , the electronic system device 110 according to the present invention includes a discount coupon face value information storage unit 111 , a training set storage unit 112 , an industry name information database 113 , and an area name information database 114 . , the information confirmation unit 115 , the input vector generation unit 116 , the one-hot vector designation unit 117 , the output vector generation unit 118 , the operation vector generation unit 119 , and the recommended model generation unit 120 . include

In the discount coupon denomination information storage unit 111, different N-dimensional one-hot vectors are stored for each of N (N is a natural number equal to or greater than 3) different discount coupon denominations.

Here, different predetermined minimum purchase amounts for using each discount coupon are designated for each of the discount coupon face values. For example, in the case of a discount coupon to which a face value of '10,000 won' is applied, a minimum purchase amount that can be used only when purchasing an item of 'at least 50,000 won' may be designated.

At this time, according to an embodiment of the present invention, when 'N=5', the discount coupon face value information storage unit 111 may store information as shown in Table 1 below.

5 discount coupons face value 5D One-Hot Vector Discount coupon 10,000 won (for purchases over 50,000 won) [1 0 0 0 0] Discount coupon of KRW 20,000 (when purchasing more than KRW 100,000) [0 1 0 0 0] Discount coupon 30,000 won (for purchases over 150,000 won) [0 0 1 0 0] Discount coupon KRW 40,000 (when purchasing more than KRW 200,000) [0 0 0 1 0] Discount coupon KRW 50,000 (when purchasing more than KRW 250,000) [0 0 0 0 1]

In the training set storage unit 112, M (M is a natural number greater than or equal to 2) business customers (the M business customers are the business operators who actually issued discount coupons according to the par value of any one of the N discount coupon face values) ) input set information consisting of industry name information collected from each, region name information, sales amount for the first predetermined period, total discount coupon issuance amount, and fixed rate of return according to the discount coupon issuance, and input of each of the M business customers The face value of the discount coupon actually issued by each of the M business customers as correct answer information for the set information (the par value for the discount coupon actually issued by each of the M business customers is one of the face values of the N discount coupons) M training sets in which are matched one by one are stored.

Here, the industry name information means information on the name of the industry operated by each of the M business customers such as PC rooms, travel agencies, and clothing stores, and the region name information means the M business customers such as Seoul, Busan and Jeju Island. It means information on the name of the region where each business site is located.

For example, when 'M=5' and the first period is 'one month', information may be stored in the training set storage unit 112 as shown in Table 2 below.

5
training sets 5 business customers Input Set Information The face value of the discount coupons issued by each of the five business customers Sectors
designation
information Area name information sales amount
(1 month) discount coupon total
issue amount confirmed
yield training set 1 business customer 1 Industry name information 5 Area Name Information 7 five thousand
Ten thousand won 10 million won 10% 20,000 won discount coupon training set 2 business customer 2 Industry name information 4 Area name information 3 ten thousand
Ten thousand won 8 million won 25% 10,000 won discount coupon training set 3 business customer 3 Industry name information 6 Area name information 4 twenty thousand
Ten thousand won 20 million won 20% 50,000 won discount coupon training set 4 business customer 4 Industry name information 3 Area name information 5 eight thousand
Ten thousand won 32 million won 5% 40,000 won discount coupon training set 5 Business customer 5 Industry name information 1 Area Name Information 8 fifteen thousand
Ten thousand won 60 million won 50% 30,000 won discount coupon

Industry name information database 113 is stored in correspondence with a plurality of predetermined industry name information and predetermined different industry specific numbers for each of the plurality of industry name information.

For example, the industry name information database 113 may store information as shown in Table 3 below.

Multiple industry name information Industry unique number Industry name information 1 One Industry name information 2 3 Industry name information 3 5 Industry name information 4 7 Industry name information 5 9 Industry name information 6 11 Industry name information 7 13 Industry name information 8 15 Industry name information 9 17 Industry name information 10 19 ... ...

In the region name information database 114, a plurality of predetermined region name information and different predetermined regional unique numbers for each of the plurality of region name information are stored in correspondence with each other.

For example, information may be stored in the area name information database 114 as shown in Table 4 below.

Multiple area name information Regional code Area Name Information 1 2 Area name information 2 4 Area name information 3 6 Area name information 4 8 Area name information 5 10 Area Name Information 6 12 Area Name Information 7 14 Area Name Information 8 16 Area Name Information 9 18 Area Name Information 10 20 ... ...

The information confirmation unit 115 refers to the industry name information database 113 to confirm the industry unique number corresponding to the industry name information of each of the M number of business operators, and refers to the area name information database 114 for the M name Check the region identification number corresponding to the region name information of each business operator.

The input vector generator 116 refers to the input set information of each of the M number of business customers included in each of the M training sets, and for each of the M business customers, the industry-specific number of each business operator, A five-dimensional input vector corresponding to each of the M business operator customers is generated by constructing a five-dimensional input vector including a region-specific number, a sales amount, a total issuance amount of a discount coupon, and a fixed rate of return.

When a five-dimensional input vector corresponding to each of the M business operators is generated, the one-hot vector designation unit 117 refers to the discount coupon denomination information storage unit 111 to determine the number of customers actually issued by each of the M business operators. By identifying the N-dimensional one-hot vector corresponding to the face value of the discount coupon, the N-dimensional one-hot vector corresponding to each of the M business operator customers is designated.

The output vector generator 118 applies a 5-dimensional input vector corresponding to each of the M business operators as an input to a deep neural network composed of two or more weight matrices, and outputs an N-dimensional output corresponding to each of the M business operators customers. create a vector

Here, the two or more weight matrices constituting the deep neural network are matrices configured so that the dimension of a vector calculated as an output of the deep neural network becomes N-dimensional.

For example, it is assumed that 'M=5' and 'N=5' as in the example described above. At this time, the information confirmation unit 115 refers to the training set storage unit 112 as shown in Table 2 and the industry name information database 113 as shown in Table 3 above, the five business customers, 'business customer 1, business customer. 2, business customer 3, business customer 4, business customer 5' Industry corresponding to 'industry name information 5, industry name information 4, industry name information 6, industry name information 3, and industry name information 1', which are industry name information, respectively You can check the unique number '9, 7, 11, 5, 1'.

After that, the information confirmation unit 115 refers to the training set storage unit 112 as shown in Table 2 and the region name information database 114 as shown in Table 4 above to read 'Business customer 1, business customer 2, and business customer 3 , '14, which is a region-specific number corresponding to 'region name information 7, region name information 3, region name information 4, region name information 5, region name information 8', which are region name information of each of , business customer 4, business customer 5' , 6, 8, 10, 16'.

Then, the input vector generating unit 116 'Training Set 1, Training Set 2, Training Set 3, Training Set 4, Referring to the input set information of each of 'business customer 1, business customer 2, business customer 3, business customer 4, business customer 5' included in each training set 5, 'business customer 1, business customer 2, business customer 3, for each business customer 4 and business customer 5', by constructing a five-dimensional input vector containing the industry-specific number, region-specific number, sales amount, total issuance amount of discount coupons, and fixed rate of return of each business customer, A 5-dimensional input vector corresponding to each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' may be generated. In this regard, for 'Business Customer 1', the industry-specific number, region-specific number, sales amount, total issuance amount of discount coupons, and fixed rate of return are '9', '14', '5 (10 million won)' and '10 ( 10 million won)' and '10(%)', the input vector generator 116 may generate '[9 14 5 10 10]' as a five-dimensional input vector for 'business customer 1'. In this way, the input vector generation unit 116 refers to the training set storage unit 112 as shown in Table 2 above, to each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5'. '[9 14 5 10 10], [7 6 10 8 25], [11 8 20 20 20], [5 10 8 32 5], [1 16 15 6 50]' as the corresponding 5-dimensional input vector can create

In this way, the five-dimensional input vector corresponding to each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' by the input vector generator 116 is '[9 14 5 10 10] , [7 6 10 8 25], [11 8 20 20 20], [5 10 8 32 5], [1 16 15 6 50]', the one-hot vector designator 117 is With reference to the discount coupon par value information storage unit 111 as shown in 1, the 'discount coupon' which is the face value of the discount coupon actually issued by 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' '[0 1 0 0 0], [1 0 0 0 0], a 5-dimensional one-hot vector corresponding to 20,000 won, discount coupon 10,000 won, discount coupon 50,000 won, discount coupon 40,000 won, discount coupon 30,000 won , [0 0 0 0 1], [0 0 0 1 0], [0 0 1 0 0]' A 5-dimensional one-hot vector corresponding to each is '[0 1 0 0 0], [1 0 0 0 0], [0 0 0 0 1], [0 0 0 1 0], [0 0 1 0 0]'.

After that, the output vector generating unit 118 is a 5-dimensional input vector '[9 14 5 10 10] corresponding to each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5'. , [7 6 10 8 25], [11 8 20 20 20], [5 10 8 32 5], [1 16 15 6 50]' are applied as inputs to a deep neural network composed of two or more weight matrices, and The five-dimensional output vector corresponding to each of customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' is '[a ₁ a ₂ a ₃ a ₄ a ₅ ], [b ₁ b ₂ b ₃ b ₄ b ₅ ], [c ₁ c ₂ c ₃ c ₄ c ₅ ], [d ₁ d ₂ d ₃ d ₄ d ₅ ], [e ₁ e ₂ e ₃ e ₄ e ₅ ]' have.

When the N-dimensional output vector corresponding to each of the M business operator customers is generated, the operation vector generator 119 generates an N-dimensional output vector corresponding to each of the M business operator customers as a predetermined activation function (Activation Function) ( The activation function is a function that converts the components constituting the N-dimensional output vector corresponding to each of the M number of business customers into values of 0 or more and 1 or less) as input to N corresponding to each of the M business customers. Creates an operation vector of dimensions.

In this case, according to an embodiment of the present invention, the activation function may be a softmax function as shown in Equation 1 below.

는 상기 M명의 사업자 고객들 각각에 대응되는 N차원의 출력 벡터를 구성하는 성분들 중 i번째 성분에 대한 연산 값으로,

는 상기 M명의 사업자 고객들 각각에 대응되는 N차원의 출력 벡터를 구성하는 성분들 중 i번째 성분의 값,

는 자연상수를 의미한다.here,

is an operation value for the i-th component among components constituting an N-dimensional output vector corresponding to each of the M business operator customers,

is the value of the i-th component among the components constituting the N-dimensional output vector corresponding to each of the M business operators,

is a natural constant.

The recommendation model generating unit 120 compares the N-dimensional operation vector corresponding to each of the M business operator customers and the N-dimensional one-hot vector corresponding to each of the M business operator customers to machine learning for the deep neural network. By performing , a recommendation model for recommending the face value of the discount coupon of the business operator is generated.

For example, as in the example described above, 'M=5', 'N=5', and 'business customer 1, business customer 2, business customer 3, business customer 4, business customer 5, which are the five business customers ' The corresponding 5-dimensional output vector is '[a ₁ a ₂ a ₃ a ₄ a ₅ ], [b ₁ b ₂ b ₃ b ₄ b ₅ ], [c ₁ c ₂ c ₃ c ₄ c ₅ ] , [d ₁ d ₂ d ₃ d ₄ d ₅ ], [e ₁ e ₂ e ₃ e ₄ e ₅ ]', the operation vector generating unit 119 is 'operator customer 1, business customer 2 , a 5-dimensional output vector '[a ₁ a ₂ a ₃ a ₄ a ₅ ], [b ₁ b ₂ b ₃ b ₄ b ₅ ], [c ₁ c ₂ c ₃ c ₄ c ₅ ], [d ₁ d ₂ d ₃ d ₄ d ₅ ], [e ₁ e ₂ e ₃ e ₄ e ₅ ]' into the activation function as in Equation 1 above. By applying as input, a five-dimensional operation vector corresponding to each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' is '[a ₍₁₎ a ₍₂₎ a ₍₃₎ a ₍₄₎ a ₍₅₎ ], [b ₍₁₎ b ₍₂₎ b ₍₃₎ b ₍₄₎ b ₍₅₎ ], [c ₍₁₎ c ₍₂₎ c ₍₃₎ c ₍₄₎ c ₍₅₎ ], [d ₍₁₎ d ₍₂₎ d ₍₃₎ d ₍₄₎ d ₍₅₎ ], [e ₍₁₎ e ₍₂₎ e ₍₃₎ e ₍₄₎ e ₍₅₎ ]' can be created.

After that, the recommendation model generation unit 120 is a 5-dimensional operation vector corresponding to each of 'business customer 1, business customer 2, business customer 3, business customer 4, business customer 5', '[a ₍₁₎ a _{( 2)} a ₍₃₎ a ₍₄₎ a ₍₅₎ ], [b ₍₁₎ b ₍₂₎ b ₍₃₎ b ₍₄₎ b ₍₅₎ ], [c ₍₁₎ c ₍₂₎ c _{( 3)} c ₍₄₎ c ₍₅₎ ], [d ₍₁₎ d ₍₂₎ d ₍₃₎ d ₍₄₎ d ₍₅₎ ], [e ₍₁₎ e ₍₂₎ e ₍₃₎ e _{( 4)} e ₍₅₎ ]' and '[0 1 0 0 0], a 5-dimensional one-hot vector corresponding to 'Business customer 1, Business customer 2, Business customer 3, Business customer 4, and Business customer 5' , [1 0 0 0 0], [0 0 0 1], [0 0 0 1 0], [0 0 1 0 0]' by performing machine learning on the deep neural network by comparing It is possible to create a recommendation model for recommending the face value of a discount coupon.

In this case, according to an embodiment of the present invention, the recommendation model generator 120 may include a machine learning performer 121 .

The machine learning performing unit 121 is a machine for the deep neural network such that an N-dimensional operation vector corresponding to each of the M business customers is closest to an N-dimensional one-hot vector corresponding to each of the M business customers. carry out learning

At this time, according to an embodiment of the present invention, the machine learning performing unit 121 includes components constituting an N-dimensional operation vector corresponding to each of the M number of business customers and an N-dimensionality corresponding to each of the M number of business customers. By calculating a loss value based on a loss function according to Equation 2 below based on the components constituting the one-hot vector of Machine learning may be performed on the deep neural network so that the average of the loss values for each of the M business customers is minimized.

은 상기 손실 값으로,

는 상기 M명의 사업자 고객들 각각에 대응되는 N차원의 원-핫 벡터를 구성하는 성분들 중 i번째 성분의 값,

는 상기 M명의 사업자 고객들 각각에 대응되는 N차원의 연산 벡터를 구성하는 성분들 중 i번째 성분의 값을 의미한다.here,

is the loss value,

is the value of the i-th component among the components constituting the N-dimensional one-hot vector corresponding to each of the M business customers,

denotes the value of the i-th component among the components constituting the N-dimensional operation vector corresponding to each of the M business operator customers.

For example, as in the example described above, 'M=5', 'N=5', and 'business customer 1, business customer 2, business customer 3, business customer 4, business customer 5, which are the five business customers , the operation vector of the fifth dimension corresponding to each _{'[a (1) a (} 2) a (3) a (4) a (5)], [b (1) b (2) b (3) b ( ₄₎ b ₍₅₎ ], [c ₍₁₎ c ₍₂₎ c ₍₃₎ c ₍₄₎ c ₍₅₎ ], [d ₍₁₎ d ₍₂₎ d ₍₃₎ d ₍₄₎ d _{( 5)} ], [e ₍₁₎ e ₍₂₎ e ₍₃₎ e ₍₄₎ e ₍₅₎ ]', and 'Business customer 1, business customer 2, business customer 3, business customer 4, business customer 5 ' Five-dimensional one-hot vectors corresponding to each are '[0 1 0 0 0], [1 0 0 0 0], [0 0 0 1], [0 0 0 1 0], [0 0 1 0 0]'.

At this time, the machine learning performing unit 121 is a five-dimensional operation vector '[a ₍₁₎ a _{(2) )} a ₍₃₎ a ₍₄₎ a ₍₅₎ ], [b ₍₁₎ b ₍₂₎ b ₍₃₎ b ₍₄₎ b ₍₅₎ ], [c ₍₁₎ c ₍₂₎ c _{(3) )} c ₍₄₎ c ₍₅₎ ], [d ₍₁₎ d ₍₂₎ d ₍₃₎ d ₍₄₎ d ₍₅₎ ], [e ₍₁₎ e ₍₂₎ e ₍₃₎ e _{(4) )} e ₍₅₎ ]' and the five-dimensional one-hot vector '[0 1 0 0 0], [1 0 0 0 0], [0 0 0 0 1], [0 0 0 1 0], [0 0 1 0 0]' by calculating the loss value based on the loss of function of, the loss value for each "business customer 1, the carrier customer 2, carrier customer 3, carrier customer 4, the carrier customer 5 ', L _1, L _2, L _3, It can be calculated as L ₄ , L _{5 '. Then, the machine learning performing unit 121 'L 1} , L ₂ , L ₃ , L ₄ L, which is a loss value for each of 'business customer 1, business customer 2, business customer 3, business customer 4, and business customer 5' Machine learning may be performed on the deep neural network so that 'L _{m ', which is an average of 5 ', is minimized.}

At this time, according to an embodiment of the present invention, the machine learning performing unit 121 performs a backpropagation process so that the average of the loss values for each of the M business operator customers is minimized, thereby machine learning can be performed.

According to an embodiment of the present invention, the electronic system device 110 includes a recommendation information confirmation unit 122 , a recommendation input vector generation unit 123 , a recommendation output vector generation unit 124 , and a recommendation operation vector generation unit. It may further include a unit 125 , a similarity calculating unit 126 , and a discount coupon face value recommendation unit 127 .

After the creation of the recommendation model is completed, the recommendation information confirmation unit 122 receives the business name information of the first business customer 130 from the customer terminal 140 of the first business customer 130 who wants to issue a discount coupon; As input set information consisting of region name information, sales amount for the first period, total expected issuance amount of discount coupons, and expected yield according to the issuance of discount coupons is received, the first business customer 130 provides a discount coupon to be issued. When a face value recommendation request is received, the industry name information database 113 is referred to to check the industry unique number corresponding to the industry name information of the first business operator customer 130, and the first business type name information database 114 is referred to. 1 Check the region-specific number corresponding to the region name information of the business customer 130 .

When the recommendation input vector generation unit 123 is confirmed the industry-specific number and region-specific number of the first business customer 130, the industry-specific number, region-specific number, sales amount, discount coupon of the first business customer 130 A five-dimensional input vector for recommendation is generated with the total expected issuance amount and expected return as components.

When the five-dimensional recommendation input vector is generated, the recommendation output vector generator 124 applies the five-dimensional recommendation input vector as an input to the deep neural network where machine learning has been completed, and the five-dimensional recommendation input vector is generated. An N-dimensional output vector for recommendation corresponding to the vector is generated.

When the N-dimensional output vector for recommendation is generated, the recommendation operation vector generation unit 125 applies the N-dimensional output vector for recommendation to the activation function as an input to correspond to the N-dimensional output vector for recommendation. Generates an N-dimensional operation vector for recommendation.

For example, as in the example described above, the first period is 'one month', and the first business customer 130 is operated from the customer terminal 140 of the first business customer 130 who wants to issue a discount coupon. 'Industry name information 5' as the industry name information of the industry in which the first business customer 130 is located, 'region name information 8' as the region name information of the region where the business place of the first business customer 130 is located, and '1 of the first business customer 130 '6 (10 million won)' as the sales amount for 'month', '8 (10 million won)' as the total expected issuance amount of discount coupons that the first business customer 130 wants to issue, and the first business customer 130 discount coupon When it is said that '15 (%)' is received as input set information for the first operator customer 130 as the desired expected return by issuing with reference to the industry name information database 113 as shown in Table 3 above, and confirms the industry unique number '9' corresponding to the industry name information of the first business customer 130 'industry name information 5', and Referring to the region name information database 114 as shown in FIG. 4 , '16', which is a region-specific code corresponding to 'region name information 8', which is region name information of the first business operator 130 , may be identified.

In this way, when the industry-specific number and the region-specific number of the first business customer 130 are identified as '9' and '16' by the recommendation information confirmation unit 122, the recommendation input vector generation unit 123 is 1 Business customer 130's industry-specific number '9', region-specific code '16', sales amount '6 (10 million won)', the total expected issuance amount of discount coupons '8 (10 million won)', and expected profit A 5-dimensional input vector for recommendation with '15(%)' as a component can be generated as '[9 16 6 8 15]'.

At this time, when the five-dimensional input vector for recommendation is generated as '[9 16 6 8 15]' by the recommendation input vector generator 123, the recommendation output vector generator 124 generates the five-dimensional recommendation '[9 16 6 8 15]', which is an input vector for recommendation, is applied as an input to the deep neural network where machine learning has been completed, An output vector for recommendation can be generated as _{'[f 1} f ₂ f ₃ f ₄ f _{5 ]'.}

In this way, when the 5-dimensional output vector for recommendation is generated as '[f ₁ f ₂ f ₃ f ₄ f ₅ ]' by the output vector generator for recommendation 124, the operation vector generator 125 for recommendation generates The 5-dimensional output vector for recommendation '[f ₁ f ₂ f ₃ f ₄ f ₅ ]' is applied as an input to the activation function as in Equation 1, and the 5-dimensional output vector for recommendation '[f _{1 ]' is applied as an input.} A five-dimensional recommendation operation vector corresponding to f ₂ f ₃ f ₄ f _{5 ] can be created as '[f (1)} f ₍₂₎ f ₍₃₎ f ₍₄₎ f ₍₅₎ ]' have.

When the N-dimensional operation vector for recommendation is generated, the similarity calculating unit 126 calculates a similarity between the N-dimensional operation vector for recommendation and the N-dimensional one-hot vector for each of the face values of the N discount coupons.

At this time, according to an embodiment of the present invention, the similarity between the N-dimensional operation vector for recommendation and the N-dimensional one-hot vector for each of the N discount coupon face values is the Euclidean Distance or the cosine similarity ( Vector similarity such as cosine similarity) may be utilized.

The discount coupon denomination recommendation unit 127 is configured to calculate the N-dimensional first one-hot vector in which the similarity with the N-dimensional recommendation operation vector is calculated at the maximum among the N-dimensional one-hot vectors for each of the N discount coupon denominations. After checking the hot vector, the first discount coupon par value corresponding to the N-dimensional first one-hot vector is checked from the discount coupon par value information storage unit 111, and the identified first discount coupon face value is set to the first business customer After determining the face value of the recommended discount coupon of 130 , information on the face value of the recommended discount coupon is transmitted to the customer terminal 140 of the first business operator customer 130 .

For example, as in the example described above, it is assumed that 'N=5', and the five-dimensional operation vector for recommendation is '[f ₍₁₎ f ₍₂₎ f ₍₃₎ f ₍₄₎ f ₍₅₎ ]', the similarity calculating unit 126 calculates the five-dimensional recommendation operation vector '[f ₍₁₎ f ₍₂₎ f ₍₃₎ f ₍₄₎ f ₍₅₎ ]' and the five discount coupon face values of 'discount coupon 10,000 won, discount coupon 20,000 won, discount coupon 30,000 won, discount coupon 40,000 won, discount coupon 50,000 won' are 5-dimensional one-hot vectors, respectively. The similarity between '[1 0 0 0 0], [0 1 0 0 0], [0 0 1 0 0], [0 0 1 0], and [0 0 0 1]' can be calculated.

_{If '[f (1)} f ₍₂₎ f ₍₃₎ f ₍₄₎ f ₍₅₎ ]' which is the five-dimensional operation vector for recommendation by the similarity calculating unit 126 and the face values of the five discount coupons, '[1 0 0 0 0], [0 1 0 Among the similarities between 0 0], [0 0 1 0 0], [0 0 0 1 0], [0 0 0 0 1], the 5-dimensional one-hot vector '[0 If it is assumed that the similarity with '1 0 0 0]' is calculated at the maximum, the discount coupon par value recommendation unit 127 selects 'discount coupon 10,000 won, discount coupon 20,000 won, discount coupon 30,000 won, 5-dimensional one-hot vectors for each of the discount coupon KRW 40,000 and the discount coupon KRW 50,000 ‘[1 0 0 0 0], [0 1 0 0 0], [0 0 1 0 0], [0 0 0 1 0] and [0 0 0 0 1]', '[0 1 0 0 0]' may be identified as the 5-dimensional first one-hot vector in which the similarity is calculated to the maximum.

After that, the discount coupon face value recommendation unit 127 corresponds to the 5-dimensional first one-hot vector '[0 1 0 0 0]' from the discount coupon face value information storage unit 111 as shown in Table 1 above. After confirming the face value of the first discount coupon, 'discount coupon 20,000 won,' and determining the 'discount coupon par value of 20,000 won', which is the first discount coupon face value, as the face value of the recommended discount coupon of the first business customer 130, the recommended discount Information on 'discount coupon 20,000 won,' which is a coupon face value, may be transmitted to the customer terminal 140 of the first business operator customer 130 .

That is, the industry name information of the first business customer 130 is 'industry name information 5', the region name information of the region where the first business customer 130 is located is 'region name information 8', and the first business operator In a situation where the sales of the customer 130 for one month is '60 million won,' and the total expected issuance amount of the discount coupon to be issued by the first business customer 130 is '80 million won', the first business customer 130 If an expected rate of return of '15%' is desired through issuance of this discount coupon, the electronic system device 110 recommends a discount coupon face value of 'discount coupon 20,000 won' to the first business customer 130 through a recommendation model. By issuing a discount coupon having a face value according to the recommended 'discount coupon 20,000 won', the first business customer 130 is optimal for achieving a profit of '15%', which is the expected rate of return that he wanted. You can issue a discount coupon with a face value of .

2 is a flowchart illustrating an operation method of an electronic system device for recommending a face value of a discount coupon based on artificial intelligence according to an embodiment of the present invention.

In step S210, different N (N is a natural number greater than or equal to 3) different discount coupon face values (each of the discount coupon face values has a different predetermined minimum purchase amount for using each discount coupon is designated) for each A discount coupon face value information storage unit in which predetermined different N-dimensional one-hot vectors are stored is maintained.

In step S220, from each of M (M is a natural number greater than or equal to 2) business customers (the M business customers are businesses that actually issued a discount coupon according to the par value of any one of the N discount coupon face values) In the input set information consisting of the collected industry name information, regional name information, sales amount for the first predetermined period, the total issuance amount of discount coupons, and the fixed rate of return according to the issuance of discount coupons, and the input set information of each of the M business customers As the correct answer information, the face value of the discount coupon actually issued by each of the M business customers (the par value for the discount coupon actually issued by each of the M business customers is any one of the face values of the N discount coupons) is matched one by one A training set storage unit in which M training sets are stored is maintained.

In step (S230), a plurality of predetermined industry name information and a predetermined different industry specific number for each of the plurality of industry name information is stored in correspondence to maintain the industry name information database.

In step S240, a region name information database is maintained in which a plurality of predetermined region name information and predetermined different region-specific numbers for each of the plurality of region name information are stored in correspondence with each other.

In step S250, with reference to the industry name information database, check the industry number corresponding to the industry name information of each of the M business operator customers, and refer to the regional name information database to refer to the regional name of each of the M business operator customers Check the region-specific code corresponding to the information.

In step S260, with reference to the input set information of each of the M number of business customers included in each of the M training sets, for each of the M business customers, the industry-specific number of each business operator, the region-specific number , sales amount, total issuance amount of discount coupons, and a five-dimensional input vector corresponding to each of the M business operator customers are generated by constructing a five-dimensional input vector as components.

In step S270, when a five-dimensional input vector corresponding to each of the M business customers is generated, referring to the discount coupon par value information storage unit, each of the M business customers corresponds to the face value of the discount coupon actually issued. By identifying the N-dimensional one-hot vector, the N-dimensional one-hot vector corresponding to each of the M business operators is designated.

In step S280, a deep neural network composed of two or more weight matrices with a five-dimensional input vector corresponding to each of the M business operators customers (two or more weight matrices constituting the deep neural network are vectors calculated as outputs of the deep neural network) N-dimensional output vectors corresponding to each of the M business operators are generated by applying them as inputs to matrices configured so that the dimension of N is N-dimensional.

In step S290, when an N-dimensional output vector corresponding to each of the M business operator customers is generated, an N-dimensional output vector corresponding to each of the M business operator customers is generated using a predetermined activation function (the activation function is the M number of business customers). It is a function that converts the components constituting the N-dimensional output vector corresponding to each of the business operators into values between 0 and 1) as input to generate an N-dimensional operation vector corresponding to each of the M business customers. .

In step S300, by performing machine learning on the deep neural network by comparing an N-dimensional operation vector corresponding to each of the M business customers and an N-dimensional one-hot vector corresponding to each of the M business customers. , create a recommendation model for recommending the face value of the discount coupon of the business operator.

In this case, according to an embodiment of the present invention, the activation function may be a softmax function as in Equation 1 above.

In addition, according to an embodiment of the present invention, in step S300 , the N-dimensional operation vector corresponding to each of the M business operator customers is maximized to the N-dimensional one-hot vector corresponding to each of the M business operator customers. It may include performing machine learning on the deep neural network to be close.

In this case, according to an embodiment of the present invention, the performing of the machine learning includes components constituting an N-dimensional operation vector corresponding to each of the M number of business customers and N-dimensional corresponding to each of the M number of business customers. By calculating a loss value based on the loss function according to Equation 2 based on the components constituting the one-hot vector of Machine learning may be performed on the deep neural network so that the average of the loss values for each is minimized.

In addition, according to an embodiment of the present invention, in the method of operating the electronic system device, after the generation of the recommendation model is completed, the industry name of the first operator customer from the customer terminal of the first operator customer who wants to issue a discount coupon As input set information consisting of information, region name information, sales amount during the first period, the total expected issuance amount of discount coupons, and expected yields according to the issuance of discount coupons is received, the discount coupons that the customer of the first business wants to issue When a face value recommendation request is received, the industry name information database is referred to, and the industry unique number corresponding to the industry name information of the first operator customer is identified, and the region of the first operator customer is referenced to the area name information database. Checking the region-specific number corresponding to the name information, when the industry-specific number and region-specific number of the first business customer are confirmed, the first business operator customer's sector-specific number, region-specific number, sales amount, discount coupon total Generating a five-dimensional input vector for recommendation with the expected issuance amount and expected return as components; when the five-dimensional input vector for recommendation is generated, the five-dimensional input vector for recommendation is converted into the deep neural network in which machine learning is completed generating an N-dimensional output vector for recommendation corresponding to the 5-dimensional input vector for recommendation by applying as an input to generating an N-dimensional recommendation operation vector corresponding to the N-dimensional recommendation output vector by applying it as an input to an activation function; when the N-dimensional recommendation operation vector is generated, the N-dimensional recommendation operation vector and calculating a similarity between the N-dimensional one-hot vectors for each of the N discount coupon face values, and the N-dimensional recommendation operation among the N-dimensional one-hot vectors for each of the N discount coupon face values. After confirming the N-dimensional first one-hot vector in which the degree of similarity with the vector is maximally calculated, the N-dimensional first circle is retrieved from the discount coupon denomination information storage unit. -Check the face value of the first discount coupon corresponding to the hot vector, determine the identified face value of the first discount coupon as the face value of the recommended discount coupon of the customer of the first business operator, and then provide information on the face value of the recommended discount coupon to the first business operator It may further include the step of transmitting to the customer terminal of the customer.

The operating method of the electronic system device according to an embodiment of the present invention has been described above with reference to FIG. 2 . Here, since the operation method of the electronic system device according to an embodiment of the present invention may correspond to the configuration of the operation of the electronic system device 110 described with reference to FIG. 1 , a more detailed description thereof will be omitted. .

The method of operating an electronic system device according to an embodiment of the present invention may be implemented as a computer program stored in a storage medium for execution through combination with a computer.

In addition, the method of operating an electronic system device according to an embodiment 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, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

As described above, in the present invention, specific matters such as specific components, etc., and limited embodiments and drawings have been described, but these are only provided to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , various modifications and variations are possible from these descriptions by those of ordinary skill in the art to which the present invention pertains.

Therefore, the spirit of the present invention should not be limited to the described embodiments, and not only the claims described below, but also all of the claims and all equivalents or equivalent modifications to the claims will be said to belong to the scope of the spirit of the present invention. .

110: Electronic system device for recommending face value of discount coupon based on artificial intelligence
111: discount coupon face value information storage unit 112: training set storage unit
113: industry name information database 114: region name information database
115: information confirmation unit 116: input vector generation unit
117: one-hot vector designator 118: output vector generator
119: operation vector generation unit 120: recommended model generation unit
121: machine learning execution unit 122: information confirmation unit for recommendation
123: input vector generator for recommendation 124: output vector generator for recommendation
125: calculation vector generation unit for recommendation 126: similarity calculation unit
127: discount coupon par value recommendation unit 130: first business customer
140: customer terminal of the customer of the first operator

Claims (12)

Different N (N is a natural number greater than or equal to 3) discount coupon face values - Each of the discount coupon face values is assigned a different predetermined minimum purchase amount for using each discount coupon - Predetermined for each a discount coupon face value information storage unit in which different N-dimensional One-Hot vectors are stored;
M (M is a natural number greater than or equal to 2) number of business customers - The M business customers are business operators who actually issued a discount coupon according to the face value of any one of the N discount coupon face values - Industry name information collected from each , region name information, sales amount for the first predetermined period, input set information consisting of the total issuance amount of discount coupons, and the fixed rate of return according to the issuance of discount coupons, and the correct information for the input set information of each of the M business customers. M trainings in which the face value of the discount coupon actually issued by each of the M business customers - the par value of the discount coupon actually issued by each of the M business customers is one of the face values of the N discount coupons - is matched one by one a training set storage in which sets are stored;
Industry name information database in which a plurality of predetermined industry name information and predetermined different industry specific numbers for each of the plurality of industry name information are stored in correspondence with each other;
a region name information database in which a plurality of predetermined region name information and predetermined different region identification numbers for each of the plurality of region name information are stored in correspondence with each other;
Check the industry-specific number corresponding to the industry name information of each of the M business operator customers with reference to the industry name information database, and a region corresponding to the regional name information of each of the M business operator customers with reference to the area name information database Information confirmation unit to check the unique number;
With reference to the input set information of each of the M business operators included in each of the M training sets, for each of the M business customers, each business operator's unique number, region-specific number, sales amount, discount an input vector generator for generating a five-dimensional input vector corresponding to each of the M business operators by constructing a five-dimensional input vector including the total coupon issuance amount and the fixed rate of return;
When a five-dimensional input vector corresponding to each of the M business customers is generated, the N-dimensional circle corresponding to the face value of the discount coupon actually issued by each of the M business customers by referring to the discount coupon par value information storage unit- a one-hot vector designator for designating an N-dimensional one-hot vector corresponding to each of the M business operator customers by identifying the hot vector;
A deep neural network composed of two or more weight matrices with a five-dimensional input vector corresponding to each of the M business operators customers. Two or more weight matrices constituting the deep neural network have an N-dimensional dimension of a vector calculated as an output of the deep neural network. an output vector generator that generates an N-dimensional output vector corresponding to each of the M business operator customers by applying as an input to the matrices configured to be ;
When an N-dimensional output vector corresponding to each of the M business operators customers is generated, an N-dimensional output vector corresponding to each of the M business operator customers is used as a predetermined activation function - The activation function is the M business operator It is a function that transforms the components constituting the N-dimensional output vector corresponding to each customer into values between 0 and 1 - Calculation that generates an N-dimensional operation vector corresponding to each of the M business operator customers by applying it as an input vector generator; and
By performing machine learning on the deep neural network by comparing an N-dimensional operation vector corresponding to each of the M business customers and an N-dimensional one-hot vector corresponding to each of the M business customers to each other, discount of business customers A recommendation model generation unit that generates a recommendation model for recommending the coupon face value
An electronic system device for recommending the face value of a discount coupon based on artificial intelligence, comprising: According to claim 1,
The activation function is an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that it is a softmax function as shown in Equation 1 below.
[Equation 1]

here,

is an operation value for the i-th component among components constituting an N-dimensional output vector corresponding to each of the M business operator customers,

is the value of the i-th component among the components constituting the N-dimensional output vector corresponding to each of the M business operators,

is a natural constant. According to claim 1,
The recommendation model generation unit
A machine learning performing unit that performs machine learning on the deep neural network so that an N-dimensional operation vector corresponding to each of the M business customers is closest to an N-dimensional one-hot vector corresponding to each of the M business customers.
An electronic system device for recommending the face value of a discount coupon based on artificial intelligence, comprising: 4. The method of claim 3,
The machine learning performing unit
Based on the components constituting the N-dimensional operation vector corresponding to each of the M business customers and the components constituting the N-dimensional one-hot vector corresponding to each of the M business customers, according to Equation 2 below By calculating a loss value based on a loss function, a loss value for each of the M business customers is calculated, and the average of the loss values for each of the M business customers is minimized in the deep neural network. An electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that it performs machine learning.
[Equation 2]

here,

is the loss value,

is the value of the i-th component among the components constituting the N-dimensional one-hot vector corresponding to each of the M business customers,

denotes the value of the i-th component among the components constituting the N-dimensional operation vector corresponding to each of the M business operator customers. 4. The method of claim 3,
After the creation of the recommendation model is completed, from the customer terminal of the first business customer who wants to issue a discount coupon, the business name information of the first business customer, region name information, the sales amount during the first period, and the total issuance of discount coupons When input set information composed of an expected amount and expected rate of return according to issuance of a discount coupon is received, and a request for recommending a face value for a discount coupon that the customer of the first business wants to issue is received, the first business name information database is referred to. a recommendation information confirmation unit for confirming a specific industry number corresponding to the business type name information of the business operator, and confirming a region unique number corresponding to the region name information of the first business customer with reference to the region name information database;
When the industry-specific number and region-specific number of the first business operator are confirmed, the first business operator customer's industry-specific number, region-specific number, sales amount, the total expected issuance amount of the discount coupon, and the expected rate of return as components a recommendation input vector generator for generating a recommendation input vector;
When the five-dimensional input vector for recommendation is generated, the five-dimensional input vector for recommendation is applied as an input to the deep neural network where machine learning has been completed, and an N-dimensional output for recommendation corresponding to the five-dimensional input vector for recommendation is generated. a recommendation output vector generator that generates a vector;
When the N-dimensional output vector for recommendation is generated, the N-dimensional output vector for recommendation is applied as an input to the activation function to generate an N-dimensional recommendation operation vector corresponding to the N-dimensional output vector for recommendation arithmetic vector generator for recommendation;
a similarity calculating unit for calculating a similarity between the N-dimensional recommendation operation vector and the N-dimensional one-hot vector for each of the N discount coupon face values when the N-dimensional recommendation operation vector is generated; and
After confirming the N-dimensional first one-hot vector in which the similarity with the N-dimensional one-hot vector for recommendation is calculated to the maximum among the N-dimensional one-hot vectors for each of the face values of the N discount coupons, the face value of the discount coupon Check the face value of the first discount coupon corresponding to the N-dimensional first one-hot vector from the information storage unit, determine the face value of the identified first discount coupon as the face value of the recommended discount coupon of the customer of the first business operator, and then recommend the recommendation A discount coupon par value recommendation unit that transmits information on the face value of the discount coupon to the customer terminal of the customer of the first business operator
An electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that it further comprises. Different N (N is a natural number greater than or equal to 3) discount coupon face values - Each of the discount coupon face values is assigned a different predetermined minimum purchase amount for using each discount coupon - Predetermined for each maintaining a discount coupon face value information storage unit in which different N-dimensional One-Hot vectors are stored;
M (M is a natural number greater than or equal to 2) number of business customers - The M business customers are business operators who actually issued a discount coupon according to the par value of any one of the N discount coupon face values - Industry name information collected from each , region name information, sales amount for the first predetermined period, input set information consisting of the total issuance amount of discount coupons, and the fixed rate of return according to the issuance of discount coupons, and the correct information for the input set information of each of the M business customers. M trainings in which the face value of the discount coupon actually issued by each of the M business customers - the par value of the discount coupon actually issued by each of the M business customers is one of the face values of the N discount coupons - is matched one by one maintaining a training set storage in which sets are stored;
maintaining a industry name information database in which a plurality of predetermined industry name information and a predetermined different industry specific number for each of the plurality of industry name information are stored in correspondence with each other;
maintaining a region name information database in which a plurality of predetermined region name information and different predetermined region identification numbers for each of the plurality of region name information are stored in correspondence with each other;
Check the industry-specific number corresponding to the industry name information of each of the M business operator customers with reference to the industry name information database, and a region corresponding to the regional name information of each of the M business operator customers with reference to the area name information database confirming a unique number;
With reference to the input set information of each of the M business operators included in each of the M training sets, for each of the M business customers, each business operator's unique number, region-specific number, sales amount, discount generating a five-dimensional input vector corresponding to each of the M business operator customers by constructing a five-dimensional input vector including the total coupon issuance amount and the fixed rate of return;
When a five-dimensional input vector corresponding to each of the M business customers is generated, the N-dimensional circle corresponding to the face value of the discount coupon actually issued by each of the M business customers by referring to the discount coupon par value information storage unit- designating an N-dimensional one-hot vector corresponding to each of the M business operators by identifying the hot vector;
A deep neural network composed of two or more weight matrices with a five-dimensional input vector corresponding to each of the M business operators customers. Two or more weight matrices constituting the deep neural network have an N-dimensional dimension of a vector calculated as an output of the deep neural network. generating an N-dimensional output vector corresponding to each of the M business operator customers by applying as an input to the matrices configured to be ;
When an N-dimensional output vector corresponding to each of the M business operators customers is generated, an N-dimensional output vector corresponding to each of the M business operator customers is used as a predetermined activation function - The activation function is the M business operator It is a function that transforms the components constituting the N-dimensional output vector corresponding to each customer into values between 0 and 1 - generating an N-dimensional operation vector corresponding to each of the M business operator customers by applying it as an input ; and
By performing machine learning on the deep neural network by comparing an N-dimensional operation vector corresponding to each of the M business customers and an N-dimensional one-hot vector corresponding to each of the M business customers to each other, discount of business customers Step of creating a recommendation model for recommending coupon face value
An operating method of an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, comprising: 7. The method of claim 6,
The method of operating an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that the activation function is a Softmax function as shown in Equation 1 below.
[Equation 1]

here,

is an operation value for the i-th component among components constituting an N-dimensional output vector corresponding to each of the M business operator customers,

is the value of the i-th component among the components constituting the N-dimensional output vector corresponding to each of the M business operators,

is a natural constant. 7. The method of claim 6,
The step of generating the recommendation model is
performing machine learning on the deep neural network so that an N-dimensional operation vector corresponding to each of the M business customers is closest to an N-dimensional one-hot vector corresponding to each of the M business customers.
An operating method of an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, comprising: 9. The method of claim 8,
The step of performing the machine learning is
Based on the components constituting the N-dimensional operation vector corresponding to each of the M business customers and the components constituting the N-dimensional one-hot vector corresponding to each of the M business customers, according to Equation 2 below By calculating a loss value based on a loss function, a loss value for each of the M business customers is calculated, and the average of the loss values for each of the M business customers is minimized in the deep neural network. An operating method of an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that the machine learning is performed.
[Equation 2]

here,

is the loss value,

is the value of the i-th component among the components constituting the N-dimensional one-hot vector corresponding to each of the M business customers,

denotes the value of the i-th component among the components constituting the N-dimensional operation vector corresponding to each of the M business operator customers. 9. The method of claim 8,
After the creation of the recommendation model is completed, from the customer terminal of the first business customer who wants to issue a discount coupon, the business name information of the first business customer, region name information, the sales amount during the first period, and the total issuance of discount coupons When input set information composed of an expected amount and expected rate of return according to issuance of a discount coupon is received, and a request for recommending a face value for a discount coupon that the customer of the first business wants to issue is received, the first business name information database is referred to. confirming the industry-specific number corresponding to the industry name information of the business operator, and confirming the region-specific number corresponding to the region name information of the first operator customer with reference to the regional name information database;
When the industry-specific number and region-specific number of the first business operator are confirmed, the first business operator customer's industry-specific number, region-specific number, sales amount, the total expected issuance amount of the discount coupon, and the expected rate of return as components generating an input vector for recommendation;
When the five-dimensional input vector for recommendation is generated, the five-dimensional input vector for recommendation is applied as an input to the deep neural network where machine learning has been completed, and an N-dimensional output for recommendation corresponding to the five-dimensional input vector for recommendation is generated. generating a vector;
When the N-dimensional output vector for recommendation is generated, the N-dimensional output vector for recommendation is applied as an input to the activation function to generate an N-dimensional operation vector for recommendation corresponding to the N-dimensional output vector for recommendation step;
when the N-dimensional operation vector for recommendation is generated, calculating a similarity between the N-dimensional operation vector for recommendation and an N-dimensional one-hot vector for each of the N discount coupon face values; and
After checking the N-dimensional first one-hot vector in which the similarity with the N-dimensional one-hot vector for recommendation is calculated at the maximum among the N-dimensional one-hot vectors for each of the N discount coupon face values, the discount coupon face value Check the face value of the first discount coupon corresponding to the N-dimensional first one-hot vector from the information storage unit, determine the face value of the identified first discount coupon as the face value of the discount coupon recommended by the customer of the first business operator, and then recommend the recommendation Transmitting information on the face value of the discount coupon to the customer terminal of the customer of the first business operator
An operating method of an electronic system device for recommending the face value of a discount coupon based on artificial intelligence, characterized in that it further comprises. A computer-readable recording medium recording a computer program for executing the method of any one of claims 6 to 10 through combination with a computer. A computer program stored in a storage medium for executing the method of any one of claims 6 to 10 through combination with a computer.

Images