KR20230058998A - Outdoor advertising media recommendation method using artificial intelligence algorithm and service server for the same - Google Patents

Abstract

The present invention relates to an out-of-home advertisement media recommendation method and service server using an artificial intelligence algorithm. Specifically, the present invention provides a method and a service server for inquiring past advertisement history information similar to an advertiser or an advertisement condition desired by an advertiser and recommending an optimal outdoor advertisement medium to the advertiser when there is a request for out-of-home advertisement. It is about.

Description

Outdoor advertising medium recommendation method using artificial intelligence algorithm and service server for it

The present invention relates to an out-of-home advertisement media recommendation method and service server using an artificial intelligence algorithm. Specifically, the present invention provides a method and a service server for inquiring past advertisement history information similar to an advertiser or an advertisement condition desired by an advertiser and recommending an optimal outdoor advertisement medium to the advertiser when there is a request for out-of-home advertisement. It is about.

Out-of-home advertising media are things that can be seen in places where the public can freely pass through at all times or continuously for a certain period of time, such as signboards, digital advertisements, standing billboards, banners, posters, flyers, and other similar things. included

On the other hand, since there are many types of outdoor advertising media and the installation locations are distributed nationwide, advertisers need to know where they are installed when they want to expose a specific product or service through outdoor advertising media. It is difficult to know what kind of media to choose.

As a way to solve this problem, advertisers are getting recommendations for out-of-home advertising media through agencies that manage and sell out-of-home advertising products. Difficulty in product purchase due to failure to accurately deliver desired advertising conditions remains a problem in this field.

The present invention has been proposed by focusing on these problems in the outdoor advertising market. A method for recommending an outdoor advertising medium optimized for an advertiser or an advertising condition desired by an advertiser using an artificial intelligence algorithm, and a service server that executes the method it's about

Republic of Korea Patent Publication No. 10-2017-0049059 (published on May 10, 2017)

An object of the present invention is to enable an advertiser to recommend an optimal outdoor advertising medium through an artificial intelligence algorithm. In particular, it aims to build a healthy outdoor advertising market by allowing people to receive reliable media recommendations based on objective evidence even if they lack knowledge in the field of advertising.

In addition, an object of the present invention is to enable anyone who can access a network to easily use an out-of-home advertising media recommendation service by enabling implementation as a system capable of providing online services.

In addition, an object of the present invention is to greatly reduce commission compared to purchasing outdoor advertising products through an advertising agency.

On the other hand, the technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

The present invention is intended to solve the above problems, and a method for recommending outdoor advertising media according to the present invention includes the steps of (a) receiving a request for advertising media recommendation; (b) pre-processing advertisement condition data included in the advertisement medium recommendation request; (c) searching for advertisement history information having a similarity to the advertisement condition data and a reference value or higher by referring to at least a portion of the advertisement condition data; and (d) generating a recommended media list including a plurality of candidate media by referring to the advertisement history information.

Further, in the method of recommending outdoor advertising media, the step (b) of searching for the advertisement history information may include the step of (b1) searching for similar advertisers by performing an operation using the advertisement condition data as an input; and (b2) searching for advertisement history information matched to the similar advertiser.

At this time, the step (b) of pre-processing the advertisement condition data further includes converting at least some of the items of the advertisement condition data into a first vector, wherein the first vector is selected from the step (b1) It can be characterized in that it is input to an algorithm used to search for similar advertisers in

Further, the algorithm in step (b1) may further input a second vector in addition to the first vector, and the second vector may correspond to any advertisement history information accumulated and stored from the past.

In addition, in the outdoor advertisement media recommendation method, the step (b1) includes calculating a similarity between the first vector and the second vector, and the advertiser corresponding to the second vector having the highest similarity is designated as a similar advertiser. It can be characterized as exploring.

In addition, in the outdoor advertising medium recommendation method, the step (b) of pre-processing the advertisement condition data further comprises extracting only valid items to be used in the recommendation operation from among the advertisement condition data included in the advertisement medium recommendation request. can include

In addition, the outdoor advertising media recommendation method may further include a step of verifying data for the advertising conditions after step (a) and before step (b).

In addition, in the method of recommending outdoor advertising media, the step (d) of generating the recommended media list may include: selecting advertisement history information having a higher similarity to the advertisement condition data than a reference value among the advertisement history information; It may include generating a list of recommended media to which an advertisement condition serving as a selection criterion and the selected advertisement history information are mapped.

In addition, in the outdoor advertising media recommendation method, the advertising condition may include at least one of an advertising target industry code, an advertising target attribute, a preferred region, a target gender, a target age, a medium attribute, an advertising budget, and an advertising period. there is.

In addition, in the outdoor advertising media recommendation method, after the step (d) of generating the recommendation media list, scores are calculated for candidate media included in the recommendation media list, and the candidate media listed according to the score ranking are included. It may further include; (e) step of generating media score information to.

Meanwhile, a method for learning an algorithm for recommending an outdoor advertising medium according to another embodiment of the present invention includes the steps of designing and simulating an algorithm for recommending an outdoor advertising medium using first data as an input; When the accuracy of the simulation results does not meet the reference value, adding an arbitrary item of the second data as an input or deleting an arbitrary item from the input and then performing simulation again.

In addition, in the method for learning the algorithm, the first data includes information obtainable from advertisement history information, and the second data is different from the first data and includes regional information or market information. can be done with

In addition, in the method of learning the algorithm, if the simulation result accuracy does not meet the reference value again, any item among the third data besides the first data and the second data is added as an input or an arbitrary item is deleted from the input. and then simulating again; may be further included.

Also, in the method for learning the algorithm, the third data may include information derived from at least one of the first data and the second data.

Meanwhile, a service server according to another embodiment of the present invention includes a pre-processing unit for pre-processing advertisement condition data included in an advertisement medium recommendation request; a calculation unit for searching for advertisement history information having a similarity with the advertisement condition data or higher than a reference value by referring to at least a portion of the advertisement condition data, and generating a list of recommended media including a plurality of candidate media by referring to the advertisement history information; can include

Further, in the service server, the calculation unit may search for similar advertisers by performing calculations using the advertisement condition data as an input, and search for advertisement history information matched with the similar advertisers.

Also, at this time, advertisement condition data used as an input to the algorithm for searching for similar advertisers may be converted into vectors by the pre-processing unit.

In addition, in the service server, the calculation unit selects advertisement history information having a similarity with the advertisement condition data higher than a reference value among the advertisement history information, and the advertisement condition serving as a selection criterion and the selected advertisement history information are mapped. It may be characterized by generating a list of recommended media.

In addition, in the service server, the calculation unit may calculate scores for candidate media included in the recommendation media list and generate recommendation information including selected media listed according to the score order.

According to the present invention, there is an effect of being able to recommend an optimal medium to an advertiser who wants an out-of-home advertisement.

In addition, according to the present invention, since the recommendation of outdoor advertising media is made by an objective means such as an artificial intelligence algorithm, the reliability of the final recommended result is high, and commissions can be greatly reduced in terms of cost, so that economic effects can be obtained.

In addition, since the present invention recommends out-of-home advertising media based on past advertisement history information compared to the present time, the recommendation performance improves as data is accumulated, that is, as the service is maintained for a long time.

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

1 is a conceptual diagram for easily understanding the basic concept of an out-of-home advertisement media recommendation service according to the present invention.
2 shows a method for recommending outdoor advertising media according to the present invention in sequence.
3 shows the step of pre-processing the advertisement condition data by dividing it into detailed steps.
4 shows a step of searching for advertisement history information similar to advertisement condition data by dividing it into detailed steps.
5 shows a step of generating a list of recommended media divided into detailed steps.
6 shows an example of a recommendation media list including a plurality of candidate media.
7 illustrates an example of recommendation information finally provided to an advertiser.
8 illustrates a method of learning an algorithm for out-of-home advertising medium recommendation.
9 schematically illustrates the configuration of a service server according to the present invention.

Objects and technical configurations of the present invention and details of the operational effects thereof will be more clearly understood by the following detailed description based on the accompanying drawings in the specification of the present invention. An embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

The embodiments disclosed herein should not be construed or used as limiting the scope of the present invention. It goes without saying that the description, including the embodiments herein, has a variety of applications for those skilled in the art. Therefore, any embodiments described in the detailed description of the present invention are illustrative for better explaining the present invention, and the scope of the present invention is not intended to be limited to the examples.

The functional blocks shown in the drawings and described below are only examples of possible implementations. Other functional blocks may be used in other implementations without departing from the spirit and scope of the detailed description. Also, while one or more functional blocks of the present invention are represented as separate blocks, one or more of the functional blocks of the present invention may be a combination of various hardware and software configurations that perform the same function.

In addition, the expression that certain components are included simply refers to the existence of the corresponding components as an expression of “open type”, and should not be understood as excluding additional components.

Furthermore, it should be understood that when a component is referred to as being “connected” or “connected” to another component, it may be directly connected or connected to the other component, but other components may exist in the middle. do.

1 is for explaining the basic concept of an out-of-home advertising medium recommendation service according to the present invention. Referring to this, in the out-of-home advertising medium recommendation service, when there is a request for recommendation from an advertiser terminal 100, the service server 200 responds to the request. Providing recommendation information as a response is a basic service structure, and at this time, the service server 200 may generate recommendation information by utilizing its own calculation models, that is, algorithms.

Looking briefly at the principle of the out-of-home advertising medium recommendation service according to the present invention, the service server 200 searches for past advertisement history information by referring to the advertisement conditions of the advertiser who previously requested the recommendation. In particular, the advertisement conditions similar to those of the advertiser. Based on the data, a past advertiser (hereinafter referred to as a "similar advertiser") who has conducted out-of-home advertisements is searched for, and the advertising history information of the similar advertiser is searched. In this case, the similar advertiser may be a specific person or may be a plurality of people. Thereafter, the service server 200 selects out-of-home advertising media used by similar advertisers or out-of-home advertising media highly preferred by similar advertisers, assigns scores and ranks to each media, and finally generates recommendation information. , The recommendation information generated in this way is provided to the advertiser terminal 100 .

1 is a simplified illustration of this series of processes so that it can be easily understood. For example, when advertiser A wants to advertise a newly compiled mathematics teaching material, she uses her advertiser terminal 100 to provide a service server After accessing (200), you can request a recommendation. At this time, the recommendation request includes personal information about the advertiser, information about the mathematics textbook to be advertised, the area the advertiser is considering, budget, target customer group, type of advertisement, period, Alternatively, advertisement condition data such as daily advertisement exposure time may be included together.

On the other hand, the service server 200 receiving such a recommendation request refers to the advertisement conditions of the above group, that is, by referring to the information about the advertiser, the information about the advertisement target, and the information about other advertisement conditions, It is possible to search whether there were other similar advertisers who had similar advertising conditions, and which media the similar advertisers used for out-of-home advertising. can create

In the process of providing a series of services as described above, the service server 200 utilizes the calculation models that are individually equipped. Here, as shown in FIG. 1, a similar advertiser search model, a product recommendation model, or a ranking model are used. can be included More detailed calculation processes performed in the service server 200 will be described again below.

Referring to FIG. 1, the outline of the out-of-home advertising media recommendation service that can be provided according to the present invention has been reviewed.

2 is a sequential order of a method for recommending outdoor advertising media according to the present invention. For reference, the method mentioned in this detailed description provides various information necessary for the service server 200 to perform calculations, for example, past advertisement sales history, region by age group, gender traffic volume, and region where outdoor advertising media are installed. , it is premised that the type of outdoor advertising media and the advertising fee for each outdoor advertising medium are collected and databased in advance. The database creation process may include a data pre-processing process of separately extracting only items necessary for out-of-home advertising media recommendation among the collected information. Data pre-processing in the process of collecting data in advance and creating a database may go through a process similar to the pre-processing step of step S20 to be described later, so a detailed description thereof will be omitted here. Meanwhile, the above databased information may be stored in a storage means provided in the service server 200, or may be stored in an external cloud server accessible through a network by the service server 200.

Referring to FIG. 2 , the method first starts with the service server 200 receiving a recommendation request (S10). The service server 200 may receive a request from the advertiser terminal 100 to use an out-of-home advertising media recommendation service, that is, a recommendation request. Step S10 can be implemented in various ways. For example, after an advertiser inputs membership registration through an application installed on a smartphone, inputs information about an advertisement target, and inputs other advertisement conditions, the service is used. By pressing the start button, the advertiser's recommendation request may be transmitted to the service server 200 . Alternatively, after the advertiser makes a wire call to the counselor, the information transmitted in the conversation process is input by the counselor, and the inputted information is transmitted and stored electronically in the service server 200, so that the advertiser's recommendation request is sent to the service server 200. may be conveyed. As such, in step S10, there is no limit to the actual implementation method of this step, as long as the information necessary for the advertiser to receive the out-of-home advertising medium recommendation is included, and the information reflecting the advertiser's intention to receive the out-of-home advertising medium recommendation exists. will be.

For reference, in step S10, the recommendation request received by the service server 200 from the advertiser may include at least one of information about the advertiser himself, information about an advertisement target, or information about other advertisement conditions. Such information transmitted together with the recommendation request will be referred to as advertisement condition data for convenience.

After step S10, the service server 200 may perform step S20 of pre-processing the received advertisement condition data. The pre-processing step can be understood as a process of making advertisement condition data received from the outside available for use in the recommendation operation process. steps can be included.

Figure 3 shows the detailed steps of the preprocessing step (S20). Referring to the figure, the preprocessing step in the present invention includes a valid item extraction step (S201), a data verification step (S203), and a data conversion step (S205). can include

The effective item extraction step (S201) is performed to exclude data that is not necessary when the service server 200 executes the calculation model (algorithm) for media recommendation. This is a step to increase the efficiency of the recommendation operation by extracting only the items related to the operation and to increase the accuracy of the final result, the recommendation information. Valid items extracted by the effective item extraction step (S201) include, for example, the advertisement target industry code, advertisement preference region, target customer age and gender, advertisement medium type, advertisement budget (desired price), or advertisement period. can Meanwhile, the above valid items may be information directly extracted from the advertisement condition data, or may be information newly generated by performing a separate search process based on the advertisement condition data.

On the other hand, since the advertisement target industry codes are databased in advance, an interface can be provided so that advertisers can easily select them on the screen when first inputting advertisement conditions. However, if the advertiser is using this service for the first time, past information about the target product and target service to be advertised may not exist, and at the same time, the industry code for it may not exist. In this case, the target of advertising A step of finding the most similar similar advertiser in the past by referring to the properties of the product or service to be provided may be performed.

In addition, in the valid item extraction step (S201), unnecessary resource consumption in the recommendation calculation process can be reduced by excluding unique names such as contract names and media names. Whether or not it is a unique name may be determined by an algorithm used by the service server 200, and in this case, an algorithm for extracting a valid item may be an artificial intelligence algorithm whose performance is improved by accumulated learning.

Next, the pre-processing step (S20) may also include a step (S203) of verifying data. For example, data consistency can be adjusted by deleting missing data, or for items that may degrade performance in the recommendation calculation process when deleting data, such as price (cost), refer to other advertisement condition data that can be referenced. Data consistency can be matched by calculating the average value and substituting the calculated value. That is, the data verification step ( S203 ) can be understood as a step of preventing performance degradation and efficiency degradation in the future recommendation operation process through deletion of data for the missing item or replacement input.

Next, the data conversion step (S205) is a step of converting the advertisement condition data received from the advertiser into an appropriate data format that the service server 200 can process. Preferably, this step may be implemented as a step of converting advertisement condition data into a vector. When converting to a vector, items with text are digitized into numbers, and items such as price (cost) are normalized to a value between 0 and 1, so they can be converted into vectors.

As such, the preprocessing step (S20) may include a valid item extraction step (S201), a data verification step (S203), and a data conversion step (S205). On the other hand, it is understood that the preprocessing step (S20) does not necessarily include all of the above steps, and if any one of the above steps is included, it can be regarded as a preprocessing step (S20). In addition, it is understood that some steps of the preprocessing step (S20) may be omitted as needed.

After step S20, the service server 200 performs a step (S30) of searching for advertisement history information. More precisely, based on the previously preprocessed advertisement condition data, the advertisement history information having a similarity with the advertisement condition data is greater than or equal to a reference value. steps to explore them. Among the contents described with reference to FIG. 1, it has been mentioned that the essence of the present invention is to recommend the most suitable outdoor advertising medium to the advertiser requesting the current recommendation by referring to past advertisement history information. This corresponds to the process of searching for advertisement history information.

4 shows a state in which step S30 is divided into detailed steps. Referring to FIG. 4, step S30 includes a step (S301) of searching for similar advertisers by using the previously converted advertisement condition data into a vector format as an input of the first algorithm (S301). A step (S303) of retrieving advertisement history information matched with the similar advertiser may be included. First, the first algorithm in step S301 is an algorithm that can be named a 'similar advertiser search model', and compares the advertisement condition data converted into a vector format with other advertisement condition data in the past stored in the database to determine which similar condition. It is to find out whether advertisers or advertisers have a history of selecting out-of-home advertising media in the past. At this time, advertisers under similar conditions naturally include advertisers currently using the service.

The first algorithm calculates the degree of similarity by comparing at least two input vectors. At this time, a formula such as cosine similarity may be used for comparison between vectors.

On the other hand, before the step S301 is executed, there may be a further step of extracting the input of the first algorithm (S300). At least one of the two inputs is the advertising condition data received from the advertiser in a vector format. The converted one (this is referred to as a first vector for convenience), and the other will be any one of previously stored advertisement condition data converted into a vector format (this is referred to as a second vector for convenience). In the step of extracting the input value (or input vector) of the first algorithm, a specific item among several items of advertisement condition data may be utilized. For example, the advertisement target industry code, that is, the advertisement target product or service The industry code becomes one criterion, and vectors having the same industry code can be extracted for input to the first algorithm.

On the other hand, as a result of calculating the degree of similarity between vectors using the first algorithm, if second vectors whose similarity to the first vector is equal to or higher than the reference value are searched, the advertisers corresponding to the respective second vectors are soon classified as similar advertiser(s). can be determined

After the similar advertiser is determined, advertisement history information matched with the similar advertiser, that is, advertisement history information about which outdoor advertising medium the similar advertiser used in the past to advertise products/services can be inquired (S303).

Meanwhile, after step S30, the service server 200 may create a list of recommended media including a plurality of candidate media by referring to previously inquired advertisement history information (S40). 5 shows the recommended media list creation step by dividing it into detailed steps. Referring to the drawing, step S40 selects advertisement history information having a higher similarity to the advertisement condition data of the current advertiser than a reference value among the advertisement history information. It may include a step (S401), and a step (S403) of generating a list of recommended media in a state in which an advertisement condition serving as a selection criterion and the selected advertisement history information are mapped.

In step S401, a step of selecting advertisement history information having a higher similarity to the advertisement condition data of the current advertiser than a reference value among advertisement history information matched with similar advertisers is performed. At this time, if advertiser A is determined to be a similar advertiser, among the advertisement history information of advertiser A, advertisement history information having a higher similarity to the advertiser's advertisement condition data than a reference value or advertisement history information having the highest similarity may be selected.

In step S403, a recommended media list is created. At this time, the recommended media list can be created so that all advertisement conditions and selected advertisement history information are displayed in the recommended media list.

For example, when advertiser A is a pseudo-advertiser, past advertisement history information of advertiser A may include advertisement history in the same area as the advertiser's regional condition. In this case, the advertisement history information may appear in the list. At this time, the list of recommended media includes 'advertising area', which is the advertisement condition that was the criteria for selection, and selected advertisement history information, that is, advertisement target attributes (product name, etc.), target gender, target age, media attributes (what kind of outdoor advertisement media), advertising budget or advertising period, etc. may be displayed together.

FIG. 6 shows an exemplary list of recommended media generated in step S40. In FIG. 6 (a), when the similar advertiser is one advertiser A, the advertisement area among past advertisement history information of advertiser A is "Seoul Gangnam". It is to select and list those that contain. On the other hand, in (b) of FIG. 6, when there are three similar advertisers, such as advertiser A, advertiser B, and advertiser C, the target gender/age among their advertisement history information includes "female/20s to 70s", and advertisements The target service is to select and list items that include “luxury shops for women”. As can be seen in FIG. 6, the similar advertisers searched by the service server 200 may not necessarily be one but several, and also when selecting those with similar advertising conditions among the advertisement history information of similar advertisers, based on several items selection can be made.

Meanwhile, after step S40, the service server 200 may calculate scores for candidate media included in the recommended media list, and generate media score information including candidate media listed according to the score ranking (S50). can In simple words, this step is understood as a step of performing score calculation to determine which outdoor advertising medium is suitable for the current advertiser, targeting the list of recommended media calculated in step S40, and generating a list of candidate media and scores for each candidate media. It can be.

Operation S50 may be performed by a second algorithm, and for convenience, in this detailed description, the second algorithm that performs a score calculation for the list of recommended media will be referred to as a 'product recommendation model'.

Finally, after step S50, by executing the third algorithm for the media score information generated in the previous step, media currently on sale are selected from candidate media listed by score, and only high-scoring media by a predetermined number are filtered. Finally, a step of generating recommendation information (S60) may proceed. For reference, in this detailed description, for convenience, the third algorithm will be referred to as a 'ranking/filtering model'.

FIG. 7 illustrates an example of a result obtained after calculation using a ranking/filtering model or recommendation information that can be finally provided to an advertiser. As can be seen in the drawing, the recommendation information can include only outdoor advertising media that can be sold at present, and the number of selected media included in the recommendation information is also limited to a predetermined number, for example, 3 or 5, so that advertisers can use outdoor advertising media. , and by displaying the score together, advertisers can numerically understand the suitability of the media. In addition, as shown in the figure, the recommendation information may further include images of selected media stored in the database, thereby supporting the advertiser's selection. As another example, the selection media included in the recommendation information may be helpful when the advertiser tries to select a specific item among the advertisement conditions by weighting the media with the highest score based on each item of the advertisement condition. there is.

On the other hand, the ranking/filtering model used in step S60 may calculate scores based on how similar the advertising conditions of the current advertiser requesting recommendation are when calculating scores for each candidate media. Elements may be further reflected. For example, when each candidate medium was used as an actual advertising medium in the past, the advertising effect score, such as how much the product sales rate increased during the same period, how much the number of customer calls increased, or how much the number of web page views increased during the same period More may be included in the calculation. In addition, when calculating the score, the priorities input from the advertiser in advance may be further reflected. If the advertisement condition data entered in the third priority is input to the service server 200, the service server 200 multiplies the weights for each of the above priority items when calculating the score in step S60, thereby giving priority when calculating the final score. can be reflected.

Referring to FIGS. 2 to 7 , the outdoor advertising media recommendation method according to the present invention has been reviewed.

On the other hand, the series of methods described above utilize artificial intelligence algorithms, and artificial intelligence algorithms are essentially a means that requires learning. Hereinafter, referring to FIG. Let's take a brief look at the learning process of the algorithm.

To conceptually understand the learning process, when defining an artificial intelligence algorithm throughout steps S20 to S60 of the method for recommending outdoor advertising media, the above algorithm is the final output when certain data are included as variables in the recommendation operation. It can be understood that learning is achieved by monitoring whether the accuracy is higher than the standard value. In other words, the above algorithm is repeatedly simulated from beginning to end, but the accuracy of the output is checked by changing the type of input variables, and if the accuracy does not meet the standard, the simulation is performed by adding new variables or removing existing variables. You can learn by repetition.

Referring to FIG. 8 , the learning method may first start with the step of primarily designing an artificial intelligence algorithm using a first data group (S1000). At this time, the first data group includes previously collected information, for example, information about advertisers accumulated from the past, advertisement history information of each advertiser, regional information, account information, or account amount information. can do.

On the other hand, the artificial intelligence algorithm designed based on the first data group above is verified (S1010) how high the accuracy of the final result calculated from the input is after executing the data group / data and AI modeling, that is, the performance. , If the performance satisfies the criterion set by the manager, the learning process may be terminated. If not, the process proceeds to applying the second data group as an input (S1020). That is, the step S1010 can also be understood as a step of simulating while changing and applying data modeling and algorithm modeling in an improved direction. It can be understood as a process of proceeding to the next second data group application step (S1020). The second data group includes information different from the first data group, and the second data group is, for example, the current status of the market related to the product/service to be advertised, such as industry information, floating population, and lifestyle by region. It may be information about, or information about the market status of the region.

Thereafter, the performance is monitored (S1030) while repeating the data group/data and AI modeling with the second data group applied, and if the performance meets the manager's criteria, the learning process ends as it is, and if not, the third It proceeds to the step of applying the data group as an input (S1040). For reference, the fact that the data group/data and AI modeling are repeated while the second data group is applied means that various types of information included in the second data group are added as inputs to the artificial intelligence algorithm, or It means repeating the process of removing information and monitoring the simulation output according to each addition/removal of information.

Meanwhile, the third data group applied to the input in step S1040 can be viewed as a group of information that can be derived by using the first data group and the second data group, including, for example, the floating population by region and the monthly floating population. Population, number of advertisers by industry, etc. may be included. The learning process may proceed in the order of repeating the data group/data and AI modeling in the same way even after the third data group is applied and determining whether the performance of the output satisfies the administrator's criteria. If the performance of the output does not meet the manager's criteria, information optimized for outdoor advertising media recommendation by returning to the application stage of the second data group or the application stage of the third data group and adding or removing different information again. combinations can be found.

On the other hand, although not shown separately in FIG. 8, negative sampling is used in the above learning method, so that data of negative attributes that are difficult for the manager to collect can be secured, and in this way, the scale of the data previously collected by the manager can be greatly increased so that the artificial intelligence algorithm can experience sufficient training data.

On the other hand, the training data required to train the above artificial intelligence algorithm is the step of extracting valid items in the data preprocessing step (S20) mentioned in the description of FIG. 2 (S201), verifying the data (S203), And / or it can be secured through the step of converting data (S205). All of the detailed steps of the data preprocessing step do not necessarily have to be included, and it will be sufficient if vector format learning data can be generated based on different types of advertisement condition data.

Finally, the service server 200, which is the main body of the outdoor advertising media recommendation method according to the present invention, may include a communication unit 201, a pre-processing unit 202, a calculation unit 203, and a control unit 205, A database 204 may be further included as needed.

The communication unit 201 is a component in charge of network communication between the service server 200 and other terminals, for example, a function of receiving various data including a request for outdoor advertisement recommendation from the advertiser terminal 100, and the service server 200 may be responsible for transmitting the calculated recommendation information to the advertiser terminal 100.

The pre-processing unit 202 serves to pre-process the advertising condition data included in the advertising medium recommendation request, and can execute the processes in the data pre-processing step (S20) previously described with reference to FIG. 2 .

Next, the calculation unit 203 searches for advertisement history information having a similarity to the advertisement condition data and a reference value or more by referring to at least a part of the advertisement condition data, and by referring to the advertisement history information, recommends a medium including a plurality of candidate media. It plays a role of creating a list, and also plays a role of calculating scores for candidate media included in the list of recommended media and generating recommendation information including selected media according to the score ranking. Since the function of the operation unit 230 is also substantially the same as the steps S30 to S60 described above with reference to FIG. 2 , redundant description will not be made here.

The database 204 has a function of storing advertisement condition data received from the advertiser terminal 100 and a function of storing advertisement history information of advertisers cumulatively from the past. However, since the database 204 may be a hardware device with a very large capacity, it may not necessarily be included in the service server 200, and may exist as a device in a form accessible to the service server 200. Alternatively, the service server 200 may be in the form of a cloud server accessible through an online network.

Finally, the service server 200 may include a control unit 205, which may actually correspond to a central processing unit. The central processing unit may also be called a controller, a microcontroller, a microprocessor, a microcomputer, and the like. In addition, the central processing unit may be implemented by hardware, firmware, software, or a combination thereof. When implemented using hardware, an ASIC (application specific integrated circuit) or DSP (digital signal processor) , DSPD (digital signal processing device), PLD (programmable logic device), FPGA (field programmable gate array), etc., when implemented using firmware or software, modules, procedures, or functions that perform the above functions or operations Firmware or software may be configured to include. In addition, the service server 200 may also include a memory as a matter of course. The memory includes read only memory (ROM), random access memory (RAM), erasable programmable read only memory (EPROM), and electrically erasable programmable read (EEPROM) memory. -Only Memory), flash memory, SRAM (Static RAM), HDD (Hard Disk Drive), SSD (Solid State Drive), etc.

On the other hand, the artificial intelligence algorithm mentioned in this detailed description is capable of performing a series of computational processes on a framework called a neural network. Here, the neural network uses artificial neurons that simplify the function of biological neurons, and the artificial neurons It may mean an environment interconnected through a connection line having a connection weight. The neural network may include a plurality of layers, and may include, for example, an input layer, a hidden layer, and an output layer. The input layer may receive an input for performing learning and pass it to the hidden layer, and the output layer may generate an output based on signals received from nodes of the hidden layer. The hidden layer is located between the input layer and the output layer, and can change the learning data transmitted through the input layer into a value that is easy to predict. Nodes included in the input layer and the hidden layer are connected to each other through a connection line having a connection weight, and nodes included in the hidden layer and the output layer may also be connected to each other through a connection line having a connection weight. The input layer, hidden layer, and output layer may include a plurality of nodes. Meanwhile, the neural network may include a plurality of hidden layers, which is referred to as a deep neural network, and learning such a neural network is referred to as deep learning. It can be understood that the artificial intelligence algorithm mentioned in this detailed description is an algorithm model capable of deep learning in such a neural network environment.

The product classification method and system therefor according to the present invention have been described above. On the other hand, the present invention is not limited to the specific embodiments and application examples described above, and various modifications may be carried out by those skilled in the art to which the present invention belongs without departing from the gist of the present invention claimed in the claims. Of course, these modifications should not be understood separately from the technical spirit or perspective of the present invention.

In particular, the components implementing the technical features of the present invention included in the block diagrams and flowcharts shown in the accompanying drawings refer to the logical boundaries between the components. However, according to an embodiment of software or hardware, the illustrated components and their functions are executed in the form of independent software modules, monolithic software structures, codes, services, and combinations thereof, and may execute stored program codes, instructions, and the like. Since functions may be implemented by being stored in a medium executable on a computer having a processor, all of these embodiments should also be regarded as falling within the scope of the present invention.

Accordingly, although the accompanying drawings and description thereof describe technical features of the present invention, they should not be simply inferred unless a specific arrangement of software for implementing these technical features is explicitly mentioned. That is, various embodiments described above may exist, and since such embodiments may be partially modified while retaining the same technical characteristics as the present invention, this should also be regarded as falling within the scope of the present invention.

In addition, in the case of a flow chart, although operations are depicted in a specific order, this is shown in order to obtain the most desirable results, and such operations must be executed in a specific sequence or sequential order shown, or all illustrated operations must be executed. It should not be understood as what it should be. In certain cases, multitasking and parallel processing can be advantageous. In addition, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and the described program components and systems are generally integrated together in a single software product or in multiple software products. It should be understood that it can be packaged.

100 advertiser terminal
200 service server
201 communication unit 202 pre-processing unit 203 calculation unit
204 Database 205 Control

Claims (19)

In the method of recommending outdoor advertising media,
(a) receiving a request for advertising media recommendation;
(b) pre-processing advertisement condition data included in the advertisement medium recommendation request;
(c) searching for advertisement history information having a degree of similarity with the advertisement condition data or higher than a reference value by referring to at least a portion of the advertisement condition data; and
(d) generating a recommended media list including a plurality of candidate media by referring to the advertisement history information;
including,
How to recommend out-of-home advertising media.
According to claim 1,
Step (b) of searching for the advertisement history information,
(b1) searching for similar advertisers by performing calculations using the advertisement condition data; and
(b2) searching for advertisement history information matched with the similar advertiser;
Characterized in that it includes,
How to recommend out-of-home advertising media.
According to claim 2,
In step (b) of pre-processing the advertisement condition data,
Further comprising converting at least some of the items of the advertisement condition data into a first vector;
Characterized in that the first vector is input to an algorithm used to search for similar advertisers in step (b1),
How to recommend out-of-home advertising media.
According to claim 3,
In addition to the first vector, a second vector is further input to the algorithm in step (b1),
Characterized in that the second vector corresponds to any advertisement history information accumulated and stored from the past.
How to recommend out-of-home advertising media.
According to claim 4,
In step (b1),
Calculating a similarity between the first vector and the second vector;
Characterized in that the advertiser corresponding to the second vector with the highest similarity is searched as a similar advertiser.
How to recommend out-of-home advertising media. According to claim 1,
In step (b) of pre-processing the advertisement condition data,
Characterized in that it further comprises; extracting only valid items to be used for recommendation operation from among advertisement condition data included in the advertising medium recommendation request.
How to recommend out-of-home advertising media.
According to claim 1,
After step (a) and before step (b),
Characterized in that it further comprises; performing data verification for the advertisement conditions.
How to recommend out-of-home advertising media.
According to claim 1,
Step (d) of generating the recommended media list,
selecting advertisement history information having a higher similarity to the advertisement condition data than a reference value among the advertisement history information;
generating a list of recommended media to which an advertisement condition serving as a selection criterion and the selected advertisement history information are mapped;
Characterized in that it includes,
How to recommend out-of-home advertising media.
According to claim 1,
The advertising conditions are
Characterized in that it includes at least one of the advertising target industry code, advertising target attribute, preferred region, target gender, target age, media attribute, advertising budget, or advertising period,
How to recommend out-of-home advertising media.
According to claim 1,
After the step (d) of generating the recommended media list,
(e) calculating scores for candidate media included in the recommended media list, and generating media score information including candidate media listed according to score ranking;
Including more,
How to recommend out-of-home advertising media.
A method for learning an algorithm for recommending outdoor advertising media,
designing and simulating an algorithm for out-of-home advertising medium recommendation using the first data as an input;
adding an arbitrary item of the second data as an input or deleting an arbitrary item from the input when the simulation result accuracy does not meet the reference value, and then performing the simulation again;
Characterized in that it includes,
A method for training an algorithm for out-of-home advertising medium recommendation.
According to claim 11,
The first data includes information obtainable from advertisement history information,
The second data is different from the first data and includes regional information or market information.
A method for training an algorithm for out-of-home advertising medium recommendation.
According to claim 12,
If the accuracy of the simulation result does not meet the reference value again, adding an arbitrary item of the third data in addition to the first data and the second data as an input or deleting an arbitrary item from the input and then performing the simulation again;
Characterized in that it further comprises,
A method for training an algorithm for out-of-home advertising medium recommendation.
According to claim 13,
Characterized in that the third data includes information derived from at least one of the first data and the second data,
A method for training an algorithm for out-of-home advertising medium recommendation.
In a service server providing an out-of-home advertising media recommendation service,
a pre-processing unit that pre-processes advertisement condition data included in the advertisement medium recommendation request;
a calculation unit that searches for advertisement history information having a similarity to the advertisement condition data or higher than a reference value by referring to at least a portion of the advertisement condition data, and generates a list of recommended media including a plurality of candidate media by referring to the advertisement history information;
Characterized in that it includes,
service server.
According to claim 15,
The calculation unit,
It is characterized in that similar advertisers are searched for by performing an operation using the advertisement condition data as input, and advertisement history information matched with the similar advertisers is retrieved.
service server.
According to claim 16,
The advertisement condition data used as an input to the algorithm for searching for similar advertisers is in a state of being converted into a vector by the pre-processing unit,
service server.
According to claim 17,
The calculation unit,
Selecting advertisement history information having a similarity with the advertisement condition data higher than a reference value among the advertisement history information, and generating a list of recommended media to which the advertisement condition as a selection criterion and the selected advertisement history information are mapped. ,
service server.
According to claim 18,
The calculation unit,
Characterized in that scores are calculated for the candidate media included in the recommendation media list, and recommendation information including selected media listed according to the score ranking is generated.
service server.

Images