KR102223640B1 - Cloud-based personalized contents subscription service providing system and method thereof - Google Patents

Abstract

The present invention relates to a system and method for providing a cloud-based personalized content subscription service. In providing a content subscription service, the present invention provides short-term intensive content including lottery recommendation content of interest to various users with high reliability and provides suitable content as a subscription service after identifying users′ personal content preferences through a deep learning analysis to encourage continued participation in the content subscription services. The present invention induces repeated interest in regularly provided subscription content by providing lottery recommendation information providing content, which has a high degree of interest when providing subscription content and is recognized as personalized information, in a more personalized and advanced quality on a basis that users can understand. The subscription service can be activated by providing content of interest to the level of user satisfaction by applying the deep reinforcement learning method that considers the user′s propensity, content use history, and social interest trend, but provides content selection suggestions in a direction that can increase users′ actual interest or obtain high content use revenue. In addition, the system comprises a content providing server and a service server.

Description

Cloud-based personalized contents subscription service providing system and method thereof}

The present invention relates to a system and method for providing a cloud-based personalized content subscription service. In particular, in providing a content subscription service, short-term intensive content including lottery recommended content of interest to various users is provided with high reliability, The present invention relates to a cloud-based personalized content subscription service providing system and method that induces continuous participation in content subscription services by identifying users' personal content preferences through deep learning analysis and providing appropriate content as a subscription service.

Different people are interested in a variety of topics, and these topics are generally constantly changing. However, there are topics that these people are interested in in common, and they correspond to topics that have a wide range of interests, such as lotteries and horoscopes.

Therefore, there is a high interest in a service that provides a recommendation number for a lottery ticket that includes a user's selection, such as a lotto, or provides a fortune according to the user's personal information, and can induce a lot of participation.

However, in the case of such lottery information or fortune information, it is easy to activate various content providing services based on the lottery information provision or fortune information provision service because interest rapidly decreases by checking for a short time as it corresponds to fragmentary information. not.

Therefore, in addition to providing lotto information or fortune information, it continuously induces user's interest by creating content that can induce interest for a long time by organizing various information desired by various users, and then providing it to subscribers periodically. There is a growing interest in subscription services that can be done.

Conventionally, there have been various subscription services provided to users by collecting and organizing information or contents on the user's interests. Basically, from the method of providing content to users who have subscribed to subscription through e-mail on a regular basis, to the method of selecting and providing content with high user interest through the dedicated application to the user who has installed a dedicated application, as well as a method that is configured to suit the user. There is also a method in which the content is configured as web content suitable for the user, and then the one-time access address information is delivered through a short message service message of the user terminal, a messenger, or SNS.

However, since all of these existing configurations are a method of directly providing content with information organized, or delivering a link to the access address of the web content that has organized the content and providing the content through the link, such received content is lost or links If the message containing is not found, it is difficult to check the content. Furthermore, when content is provided through a dedicated application, there is a limitation in that participation is lowered due to the inconvenience of installing a dedicated application.

In addition, depending on the type of subscription, the content on the subject of interest set in advance by the user was selected and provided, or the subject with high interest was analyzed and provided based on the user's response to the content delivered to the user. As a result, it is difficult for users to experience the actual level of personalization or to trust that the content is personalized according to themselves, so there is a limit that satisfaction with the subscription content is not high.

In the end, when a user uses a subscription service, the effort required to check the content is reduced, and even if the message containing the content access information is lost, the subscribing user can easily check his or her content, and substantially change the user. After analyzing the interests of interest by taking into account the behavior of the user's continuous participation in the subscription content, the user's tendency and external environment changes, the analyzed content is provided that the interest is high at that time, and lottery numbers that show high interest are usually recommended. There is a need for a new way to increase the reliability of reflecting personalization in the overall subscription content by making it clear that information related to or fortune-telling is personalized according to one's own needs.

Korean Patent Registration No. 10-1967696, [Subscription service provision method and apparatus] Korean Patent Application Publication No. 10-2015-0065365, [Recommended content providing method, recommended content display method, computing device and computer-readable medium]

The purpose of the embodiments of the present invention to improve the above-described problem is to provide lottery recommendation information that has a high degree of interest when providing subscription content and that is recognized as personalized information, and provides content with advanced quality tailored to individuals based on a standard that users can understand. In addition to inducing repetitive interest in subscription content provided periodically, a content selection proposal is made in the direction of increasing the user's actual interest or obtaining high content usage revenue while taking into account the user's disposition, content use history, and social interest trends. It is to provide a cloud-based personalized content subscription service providing system and method that enables users to provide content of interest to a level of satisfaction by applying the provided deep reinforcement learning method.

Another object of the present invention is to provide a unique access address determined at the time of membership registration for a corresponding user when providing subscription content to a user terminal, and to continuously access the subscription content through a general purpose application without installing a separate dedicated application through the access address. It is to provide a cloud-based personalized content subscription service providing system and method that improves the convenience of accessing subscription content for users by enabling confirmation.

Another object of the present invention is to generate a neural network for selecting a winning number based on accumulated winning number information or past winning information of a corresponding user when providing lotto recommendation information based on a Generative Adversarial Network (GAN). Short-term interest by making it possible to increase the likelihood of winning the expected winning number by creating a number with a high probability of winning based on a random number while searching for invisible and complex rules for winning, rather than a classifier based on simple neural network analysis. It is to provide a cloud-based personalized content subscription service providing system and method that enhances the reliability of the content so that it can obtain repetitive usability for the entire subscription content.

Another object of the present invention is a fortune-telling algorithm with a high hit rate among multiple fortune-telling algorithms obtained based on user personal information based on a past winning number, a user's winning history, or a user's investment result when providing content for lottery recommendation information or fortune information. It is to provide a cloud-based personalized content subscription service providing system and method that allows users to generate lottery recommendation information by reflecting the corresponding fortune after notifying the user that the selected fortune algorithm has a high hit rate after selecting the selected fortune-telling algorithm.

Another object of the present invention is to select content that reflects the user's personal interest in order to convert the interest induced by the short-term interest content into interest in the entire subscription content, but determine the user's content usage rate and user lifetime value (LTV). As control information, content selection information that obtains reward information as a standard and uses characteristic information on the content to be provided and external trend information (for example, classification information on major portal search terms for a week) as status information, and allows the corresponding reward to be the greatest. Deep reinforcement learning predicted, in particular, the A3C (Asynchronous Advantage Actor-Critic) method, which is known to have the best performance, is modified and applied according to the configuration of the present invention, thereby maintaining the uniqueness of a specific group and providing a high hit rate of interest. It is to provide a system and method for providing a base personalized content subscription service.

The cloud-based personalized content subscription service providing system according to an embodiment of the present invention to achieve the above object is a mixture of short-term interest content including lottery recommendation information content and continuous interest content selected according to the user's personal interest level. As a cloud-based personalized content subscription service providing system that provides personalized subscription content to user terminals, a content providing server that generates content that can be selected as subscription content or information necessary for content creation together with characteristic information according to predefined criteria Wow, it processes member sign-up, including collection of personal information about users, creates and provides a personal access address for confirming subscription content for members, and predicts lottery winning through neural network analysis based on accumulated lottery winning information. Generates a number, generates selected lottery recommendation content by reflecting the user's personal fortune among the generated lottery winning numbers, and collects user use information for the content exposed through the personal access address, and the content providing server Based on the characteristic information of the generated content, personalized content with high user interest is selected through neural network analysis, and then the subscription content created by combining with the lottery recommended content is registered in the personal access address, but configured in a cloud-based manner. It includes a service server that manages all information about users online based on their personal access address.

As an example, the service server performs member registration by collecting member subscription information including personal information from the connected user terminal, and collects actual investment information including the purchase of lottery tickets from the user terminal, and the member registered user. A personalized URL (Uniform Resource Locator) management unit that creates a personal access address to check the subscription content and provides a web hosting service to check the subscription content through the personal access address, and a neural network based on accumulated lottery winning information. A lottery content provider that generates an estimated lottery winning number through analysis, and generates selected lottery recommendation contents by reflecting the user's personal fortune among the generated lottery winning expected numbers, and a user for the content exposed through a personal access address A content usage monitoring unit that collects the usage information of the content, a content selection unit that collects content or information necessary for content creation and characteristic information for each content or information generated through the content providing server, and a portal with a preset level of social interest. External trend information collection unit received as trend information obtained through big data analysis of the site, selectable content obtained through the content selection unit or characteristic information of information necessary for content creation, and trend information obtained through external trend information collection unit. As a standard, the optimal content selection unit for each individual provides content selection information through neural network analysis so that the usage information obtained through the content usage monitoring unit is improved, and the content providing server is configured according to the content selection information of the optimal content selection unit for each individual. It may include a personal content generator that selects the content generated through the content or information necessary for content creation to generate individual content, and provides the personalized URL management unit as a subscription content by adding the lottery recommended content provided by the lottery content providing unit. have.

As an example, the lottery content generation unit includes a winning number analysis unit, and the winning number analysis unit is configured to predict the number through a generative adversarial network (GAN) based on accumulated previous winning number information. You can update your own rules by reflecting the winning information for each round.

As an example, the lottery content generation unit includes a winning information management unit that collects cumulative winning information of the last round and winning information of a lottery ticket purchased by a user, a store information management unit that manages the location and winning history of a registered store, and cumulative winning information. Based on a neural network analysis, a winning number analysis unit that generates an expected lottery winning number and a user's personal fortune among the lottery winning prediction numbers generated through the winning number analysis unit are selected as a preset weight to select a personalized expected winning number. , Lottery ticket that generates lottery contents including the optimal fortune reflecting unit that selects the purchase time and store by reflecting the user's personal fortune with a preset weight, and the lottery number selected through the optimal fortune reflecting unit and the purchase time and store recommendation information It may include a content generation unit.

As an example, the individual optimal content selection unit calculates the level of interest that calculates the level of interest in the characteristic information of the content according to the collection history of the previous content use history collection unit, which collects the previous content use history from the content use monitoring unit, and the previous content use history collection unit. Reward by converting the result of using the content desired by the service provider into a numerical value by reflecting a preset weight in the results of the LTV (LifeTime Value) calculation unit, which calculates the wealth and profit value according to the content use, and the interest level calculation unit and the LTV calculation unit. Reward calculation unit provided as information, provided content characteristic information classification unit that collects the characteristic information of selectable content or information obtained through the content selection unit to create new subscription contents and provides it as status information, and obtained through the external trend information collection unit. An external interest information conversion unit that corresponds trend information to content feature information and provides status information, and a past content selection history and result accumulating unit having past content selection information and a usage history of the subscription content generated accordingly as a result. , A deep reinforcement learning unit that outputs content feature information that may have the largest reward information as control information by inputting status information based on the learning contents through the information of the past content selection history and result accumulator, and control information of the deep reinforcement learning unit. It may include a content selection information providing unit for receiving as the content selection information selectable content or information having content feature information for generating a new subscription content.

A method for providing a cloud-based personalized content subscription service according to another embodiment of the present invention provides a personalized subscription content in which short-term interest content including lottery recommendation information content and continuous interest content selected according to the user's personal interest level are mixed. As a method of providing a cloud-based personalized content subscription service using a cloud-based personalized content subscription service providing system provided to a terminal, the service server handles membership registration including collection of personal information for the connecting user, and checks the subscription content for the member. The step of generating and providing a personal access address for the service, and the service server generates an estimated lottery winning number through neural network analysis based on the accumulated lottery winning information, and selected by reflecting the user's personal fortune among the generated lottery winning prediction numbers. Generating recommended lottery content; generating content that can be selected as a subscription content by the content providing server or information necessary for content creation together with characteristic information according to a predefined criterion; and the personal access by the service server Collects user usage information for content exposed through the address, selects personalized content with high personal interest in the user based on the characteristic information of the content generated by the content providing server through neural network analysis, and then combines it with the recommended lottery content And generating a subscription content, and registering, by a service server, the generated subscription content in the personal access address.

The system and method for providing a cloud-based personalized content subscription service according to an embodiment of the present invention have a high degree of interest when providing subscription content, and the content for providing lottery recommendation information recognized as personalized information is further tailored to the individual as a standard that the user can understand. In addition to inducing repetitive interest in subscription content that is provided periodically by providing it in a quality that is provided, it takes into account the user's disposition, content use history, and social interest trends, but aims to increase the user's practical interest or obtain high content usage revenue. By applying a deep reinforcement learning method that provides a content selection proposal, there is an effect of activating a subscription service by providing content of interest to a level that satisfies the user.

In addition, subscription content is provided to a unique access address determined at the time of membership registration, and subscription content can be continuously checked through a general-purpose application without installing a separate application through the access address, thereby enhancing the convenience of accessing subscription content for users. There is an effect of increasing the user use rate for the content.

In addition, when lottery recommendation information is provided as a subscription content, a simple neural network analysis is performed by constructing a neural network for selecting winning numbers based on the accumulated winning number information or the actual user's past winning information based on a generative hostile neural network (GAN). Reliability of short-term interest content can be increased by making it possible to increase the likelihood of winning the expected winning number by creating a number with a high probability of winning based on a random number while searching for invisible and complex rules for winning, not a classifier according to It has the effect of increasing the repetitive usability of the entire subscription content.

Furthermore, when providing content for lottery recommendation information or fortune information, after selecting a fortune algorithm with a high hit rate among multiple fortune-telling algorithms obtained based on user personal information based on the past winning number, the user's winning history, or the user's investment result. After notifying the user that the selected fortune-telling algorithm has a high hit rate, it is possible to generate lottery recommendation information by reflecting the fortune-telling, thereby increasing the psychological satisfaction of the user.

In addition, content selection information is used as control information to obtain compensation information based on the user's content usage rate and user lifetime value, and to maximize the compensation by using characteristic information and external trend information on the content to be provided as status information. By applying the transformation of predicted deep reinforcement learning, it has the effect of providing a high hit rate with high interest while maintaining the uniqueness of each group.

1 is an exemplary view showing the configuration of a cloud-based personalized content subscription service providing system according to an embodiment of the present invention.
2 is a block diagram showing the configuration of a service server of a cloud-based personalized content subscription service providing system according to an embodiment of the present invention.
3 is a block diagram showing the configuration of a member management unit according to an embodiment of the present invention.
4 is a block diagram showing the configuration of a lottery content providing unit according to an embodiment of the present invention.
5 is a conceptual diagram illustrating an operation method of the optimal fortune reflecting unit shown in FIG. 4.
6 is a block diagram showing a configuration of an optimal content selection unit for each individual according to an embodiment of the present invention.
7 shows an extended configuration of a deep reinforcement learning unit according to an embodiment of the present invention.
8 is a flow chart showing an operation process according to an embodiment of the present invention.

The present invention as described above will be described in detail through the accompanying drawings and embodiments.

It should be noted that the technical terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. In addition, the technical terms used in the present invention should be interpreted as generally understood by those of ordinary skill in the technical field to which the present invention belongs, unless otherwise defined in the present invention, and is excessively comprehensive. It should not be construed as a human meaning or an excessively reduced meaning. In addition, when a technical term used in the present invention is an incorrect technical term that does not accurately express the spirit of the present invention, it should be replaced with a technical term that can be correctly understood by those skilled in the art. In addition, general terms used in the present invention should be interpreted as defined in the dictionary or according to the context before and after, and should not be interpreted as an excessively reduced meaning.

In addition, the singular expression used in the present invention includes a plurality of expressions unless the context clearly indicates otherwise. In the present invention, terms such as “consisting of” or “comprising” should not be construed as necessarily including all of the various elements or various steps described in the invention, and some of the elements or some steps are included. It should be interpreted that it may not be, or may further include additional components or steps.

In addition, terms including ordinal numbers such as first and second used in the present invention may be used to describe the elements, but the elements should not be limited by the terms. The terms are used only to distinguish one component from another. For example, without departing from the scope of the present invention, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, but the same or similar components are assigned the same reference numerals regardless of the reference numerals, and redundant descriptions thereof will be omitted.

In addition, in describing the present invention, when it is determined that a detailed description of a related known technology may obscure the subject matter of the present invention, a detailed description thereof will be omitted. In addition, it should be noted that the accompanying drawings are only for easily understanding the spirit of the present invention and should not be construed as limiting the spirit of the present invention by the accompanying drawings.

In particular, the user terminal of the present invention encompasses terminal devices capable of various types of online access, and includes smart phones, portable terminals, mobile terminals, and personal information with communication functions. Various terminals such as personal digital assistant (PDA), portable multimedia player (PMP) terminal, personal computer, notebook computer, slate PC, tablet PC, ultrabook, etc. The communication network may include various wired/wireless communication methods that are widely known.

In addition, the service server or the content providing server may be configured with various types of servers including a single computer, a parallel computer, and a distributed computer. In particular, the service server does not store separate applications or information in the user terminal through a cloud configuration. The server has a cloud configuration that stores information and provides all functions, and a plurality of physical or logical servers may be included for the cloud configuration.

A configuration other than this embodiment of the present invention will be described with reference to FIGS. 1 to 8.

1 is an exemplary view showing the configuration of a cloud-based personalized content subscription service providing system according to an embodiment of the present invention.

As shown, the service server 100 that substantially selects content to generate subscription content and provides personalized subscription content to the user terminal 10 to which a member has subscribed has a configuration 100c of a cloud server.

The concept and scope of cloud is not clearly defined, but it is an environment in which users can use desired services anytime, anywhere without worrying about information technology resources, and by integrating various computing resources existing in different physical locations. It can also be viewed as a technology that allows users to use them as if they were directly using them through a network connection.

Therefore, in the cloud configuration 100c, the user terminal 10 connected by integrating a plurality of servers that can be physically distinguished, continuously provides online subscription content in a personalized environment, as if using an application installed in the user terminal 10. It provides a function of the service server 100 to be used as a service server, and may be composed of a plurality of physically or logically divided servers. Furthermore, although these divided servers each manage the user terminals 10, it may appear that the user terminals 10 are commonly connected to an integrated service server 100 from the outside.

Cloud-based personalized content subscription service according to an embodiment of the present invention is a service that provides personalized subscription content in which short-term interest content including lottery recommendation information content and persistent interest content selected according to the user's personal interest level are mixed to a user terminal As a result, short-term interest content that needs to generate personalized content, such as personalized lotto recommendation information content, is generated by the service server 100, and is necessary to generate a number of pre-generated content or specific content such as other continuous interest content. Information (content components (articles, photos, images, data, statistical information, etc.) that enable the creation of content by combining such information) are generated in a separate content providing server 200 and then stored in the database 300. Save it.

The service server 100 selects the content generated through the content providing server 200 or information for content creation according to the user's individual preference, then integrates it to create a subscription content, and provides it to the user terminal 10. .

Here, the content providing server 200 generates content that can be selected as a subscription content or information necessary for content creation together with characteristic information according to a predefined standard, and in fact, the content is created by a separate content producer. Collected content or information for content creation, but it is possible to automatically or manually classify and match the characteristic information of each content or information for content creation.

Such characteristic information is the creation date of individual content (information for content creation), keywords included, and related categories (living, food, sports, automobile, fashion, health, investment, travel, webtoon, politics, economy, law, games, etc.) It may be fixed content characteristic information such as, and may be a result of using a neural network classifier capable of classifying into a set detailed feature classification by learning a context or a composition word constituting the corresponding content.

The characteristic information of such content is intended to be used to select content suitable for personal taste, so the more it is subdivided or advanced through neural network learning, the more clearly the characteristics of the content can be distinguished. Therefore, when selecting the content suitable for the user, satisfaction can be increased. have.

The illustrated user terminal 10 can be any type of terminal as long as the content in the form of a web page can be checked through an online connection, and in the embodiment of the present invention, the user terminal 10 is connected to an access address according to a user input. This is to check the subscription content through a general-purpose web browser, and there is no need to install a separate dedicated application.

The service server 100 receives member subscription information from the user terminal 10 and registers as a member, generating access information (URL:Uniform Resource Locator) unique to the member user to create a personal access address (for example,'lottocloud. .io/7693'). Thereafter, the subscription content is provided so that it can be checked through the corresponding personal access address.

In this way, the service server 100 collects personal information from the accessing user terminal 10 to process membership registration, creates and provides a personal access address for confirming the subscription content for the member, and recommends lottery content (for example, , Lottery recommendation information content), selects content that suits the user's taste by using the content characteristic information, creates a subscription content, and registers it in the personal access address.

To this end, the service server 100 generates an estimated lottery winning number based on the accumulated lottery winning information through neural network analysis, and selects a lottery ticket specialized for the user by reflecting the user's personal fortune among the generated lottery winning expected numbers. Create recommended content. In addition, after collecting user usage information for content exposed through personal access addresses, the user's taste is determined, and then personalized content with high personal interest of the user is selected based on the characteristic information of the content generated by the content providing server.

Furthermore, the service server 100 is configured on a cloud basis and manages all information about users online based on their personal access address, and divides them into a plurality of groups according to the user's preferences as needed, and then divides them into different servers. Can be managed.

On the other hand, as shown, in providing personalized content to users, the service server 100 reflects current social interest trends, rather than simply using fixed propensity information or personal past use history. The user's personalized content can be selected in consideration of social trends, and social interest trends can be used by collecting information about classification of major portal search terms or trend information of social network services, for example, for a week.

As described above, an embodiment of the present invention improves the convenience of checking the user's subscription content, and provides lottery number recommendation content showing high interest as short-term interest content, and a user who accesses to check such lottery number recommendation content. By providing personalized, sustained-use contents with high interest in users, users can increase the utilization of subscription contents and increase service satisfaction.

To this end, user satisfaction with lottery number recommendation content must be increased, and subscription content must be created by selecting content that can attract the user's interest. In an embodiment of the present invention, a user's psychological satisfaction with respect to content recommended for lottery numbers is clearly increased, and it is possible to advance the selection of content of interest.

2 is a block diagram showing the configuration of the service server 100 of the cloud-based personalized content subscription service providing system according to an embodiment of the present invention.

As shown, the service server 100 collects member subscription information including personal information from the accessing user terminal 10 to perform member registration, and collects actual investment information including the purchase of lottery tickets from the user terminal 10. A personalized URL (Uniform Resource Locator) that provides a web hosting service to create a member management unit 100 to check the subscription content for member-registered users, and to check the subscribed content through the personal access address. ) The management unit 150, based on the accumulated lottery winning information, generates an expected lottery number through neural network analysis, and generates selected lottery recommended content by reflecting the user's personal fortune among the generated lottery winning prediction numbers. The content providing unit 140, the content usage monitoring unit 160 that collects the user's use information for the content exposed through the personal access address, and the content generated through the content providing server 200 or necessary for content creation. A content selection unit 120 that collects information and characteristic information for each content or information, an external trend information collection unit 170 that receives social interest as trend information obtained through big data analysis of a preset portal site, Based on the characteristic information of selectable content obtained through the content selection unit 120 or information necessary for content creation, and the trend information obtained through the external trend information collection unit 170, the content usage monitoring unit 160 Content generated through a content providing server according to the content selection information of the individual optimal content selection unit 180 that provides content selection information through neural network analysis so that the usage information is improved B. Includes a personal content generation unit 130 that selects information necessary for content creation to generate individual content, adds lottery recommendation content provided by the lottery content providing unit 140, and provides it to the personalized URL management unit as a subscription content. do.

Each configuration of the service server 100 will be described in more detail with reference to the configurations of FIGS. 3 to 6.

3 is a block diagram showing the configuration of the member management unit 100 according to an embodiment of the present invention.

As shown, the member management unit 100 checks the connection according to the service server address set in advance or the personal access address issued to the member through the user terminal 10, and in the case of access to the personal access address, pre-registered access authentication Member terminal interlocking unit 111 that checks the number and provides subscription content, and personal information including the region and date of birth from the user terminal 10 accessed to the address of the service server through the member terminal interlocking unit 111 A member subscription processing unit 113 that collects as subscription information and performs member registration, creates an account for a registered user, registers a personal access address issued to the user, and provides information from the user terminal 10 B. A member information management unit 112 that stores the usage information obtained through the content usage monitoring unit, and a member investment information collection management unit 114 that collects investment history including the user's lottery purchase history from the user terminal 10 registered as a member. ), and a subscription information management unit 115 that determines the type of subscription content to be provided according to the member's level, and delivers the subscription content delivery cycle and the generation notification for the generated subscription content to the registered user terminal 10 when registering as a member. do.

The member registration processing unit 113 collects personal information such as date of birth, address of residence, and area where lottery tickets can be purchased as member registration information in order to check personalized fortunes in the lottery content provider 140. Content preference information can also be collected.

The member terminal interworking unit 111 can separately use authentication information to verify the proper access of the user terminal 10 accessed through the personal access address, and such authentication information is issued to the member upon registration by the member registration processing unit 113. It may be one authentication information, and it may be information entered by directly specifying authentication information, as if a user registers a password. Therefore, after accessing the personal access address through the user terminal 10 at any time, the subscription content can be used through verification of authentication information.

The member information management unit 112 provides information on a user who accesses through the user terminal 10 in order for the service server 200 to operate as a cloud (member information for the user, user preference information, content read by the user). By managing all of the information, information on the content being read, information on the result of using the user content, etc.), it is possible to provide a continuous user environment as if using a local application configured in the user terminal 10 with only access.

Member investment information collection management unit 114 collects the lottery purchase history (purchase place, purchase date, number, whether or not to win) from the user terminal 10 registered as a member. A method of reading the printed code (using a web application) can be used. Such information may be stored in the member information management unit 112 and may be utilized when the lottery content providing unit 140 provides personalized lottery content.

4 is a block diagram showing the configuration of a lottery content providing unit 140 according to an embodiment of the present invention.

Since the illustrated lottery content providing unit 140 is a content for inducing repetitive interest from a user for a subscription content, its accuracy must be high, and at least the user must be able to psychologically recognize its reliability. Accordingly, in the embodiment of the present invention, two new methods (winning number analysis, optimal fortune-reflecting) are applied to enhance the performance of lottery number prediction and greatly increase the psychological satisfaction of the user.

First, the illustrated lottery content providing unit 140 includes a winning information management unit 141 that collects accumulated winning information of the last round and winning information of a lottery ticket purchased by a user, and a store that manages the location and winning history of the registered store. Users among the information management unit 142, the winning number analysis unit 143 that generates an expected lottery winning number through neural network analysis based on the accumulated winning information, and the lottery winning prediction number generated through the winning number analysis unit 143 The optimal fortune reflecting unit 144, which selects the purchase time and store by reflecting the personal fortune of the user with a preset weight, selects a personalized predicted winning number, and reflects the user’s personal fortune with a preset weight, and an optimal fortune reflecting unit It includes a lottery content generation unit 145 for generating lottery content including the lottery number selected through 144, purchase time, and store recommendation information.

In this configuration, the winning information management unit 141 inputs the accumulated winning history previously won, that is, the number of the 1st and 2nd prizes of the past round, the number of winners, the amount of the winning, and the information of the winning lottery store. The number of the lottery ticket and the purchase store information may be collected based on the information included in the lottery for the actual purchased lottery information provided by the user through the member investment information collection management unit 114.

The store information management unit 142 manages the geographical location of each store, a history of past winning lottery tickets, and the like, and, if necessary, may generate a store score such as a recent trend of winning lottery tickets. If there are many users who have entered the lottery information, the sales quantity information for each store included in the lottery information (for lotto lottery tickets (receipts), purchase date, lottery date, game selection number, store information, total sales order, and corresponding store) The number of lottery tickets sold at the corresponding retailer can be predicted based on the information such as the order of sales of the corresponding day at (in), and the degree of winning compared to the number of lottery tickets sold can be calculated.

As a new method applied according to an exemplary embodiment of the present invention, in the exemplary embodiment of the present invention, a winning number analyzing unit 143 is used to analyze the winning number in a new method.

There are various statistical analysis methods as a general lotto number analysis method, and there is a neural network analysis method that has recently been in the spotlight. In the case of such a neural network analysis method, there are a high-performance classification method that predicts the probability of winning each number this time by simply learning the past winning numbers, or a method that analyzes the past winning numbers after imaged them.

However, the winning number analysis unit 143 according to an embodiment of the present invention is configured to predict the number through a generative adversarial network (GAN) based on the accumulated previous winning number information. The rules are configured to update themselves by reflecting the winning information for each round. The accumulated information on the winning number of the previous round is not only for 45 numbers, but various related information such as the date of each round, weather, sales volume, number of winners, winning amount, and winning number. Try to generate numbers with high probability.

In general, generative adversarial neural networks generate virtual fake information that meets specific criteria, and it is a method of upgrading such fake information to the extent that it is difficult to distinguish it from the real one. It is a way of learning to be hostile to each other to provide realistic results.

Since tens of thousands of training data are required to train a general neural network classifier, it is not possible to analyze a meaningful neural network by learning the winning numbers of less than 1000 until now.

Therefore, in the embodiment of the present invention, not'supervised learning' that provides confirmed learning information such as such simple winning information, but for hundreds of previous rounds, correlation between changes in winning numbers between each round and correlation between changes in winning numbers between 10 rounds. Relationship, correlation between changes in winning numbers for 20 times, correlation between changes in winning numbers of the last 100, changes between these correlations, relationship with weather, relationship with sales volume, relationship with number of winners, etc. A generative adversarial neural network that can derive an object and derives the relevance of hundreds of winning information that cannot be individually designated by a person, that is, a regularity that is difficult for a person to check within randomness, through'unsupervised learning'. It allows you to get a predicted number that fits the rules you find yourself.

Currently, these generative adversarial neural networks are used to create new images that do not exist after learning pictures of human faces but are judged as'persons', or learn Gogh's work and create new images as if they were drawn by Gogh. It is used to provide results such as conversion into works. In other words, the administrator does not directly learn the numerous features and rules that can be judged by the human face, but by identifying the regularity by itself and generating noise with rules that match the face of the person, the result of combining the noise is converted into a person. It is a method of providing an image that can be judged.

Therefore, unlike the conventional simple neural network learning method in which a person finds a rule by statistical analysis through various types of analysis methods or learns a sequence of hundreds of times to predict the relationship between them, it is applied in the embodiment of the present invention. The generative adversarial neural network (GAN) constructs itself a complex predictive model that cannot be simply defined because a number of rules are mixed in the process of advancing the countless relationships between hundreds of winning numbers as the classification model and the generated model compete with each other. . Through this, the regularity of number extraction using specific machines in a specific environment can be estimated more accurately than a simple prediction model.

If this winning number analysis method is to secure reliability through the combination of the latest information and communication technology and artificial intelligence technology, another novel method applied according to an embodiment of the present invention is an optimal fortune-telling reflector using a fortune-telling that is regarded as unscientific ( 144).

From a statistical point of view, the probability of winning a lottery is the same, but in reality, when analyzing the winning history of such a lottery, the probability of winning for each number is not the same, and there are various variables, including the characteristics of the lottery lottery device and the lottery environment. In addition, there are actually stores where the lottery is sold out of statistical figures, and there are stores where the winnings are repeated, and there is a phenomenon in which buyers are concentrated by considering such stores as so-called'Myeongdang'. Furthermore, scientific proof is difficult but can not be completely ignored, such as winning because of good fortune, dreaming well, or winning a lottery ticket purchased at a sales office that wins well according to the Feng Shui geography. It is often occurring in the complex real world.

Therefore, there have been services that recommend lottery numbers by reflecting the user's fortune or to recommend the sales date and direction of the store.

However, if you want to apply additional suggested methods based on various Eastern and Western philosophy, such as using keys, horoscopes, constellations, or feng shui geography, the results of each method may be different or even conflict with each other. In addition, since it is not possible to determine which of these methods is best to reflect first, only psychological stability is provided at present, and the application of these methods has not been explained to increase the reliability of users.

In order to solve this problem, the illustrated optimal fortune-telling reflecting unit 144 does not specify the accuracy of the individual philosophy-related analysis content itself or a specific methodology for the individual analysis method, but considers that it can affect the winning of the lottery. The philosophical analysis methods of various individuals are applied, but through historical analysis, which of these various analysis methods and targets (detailed items) that are difficult to determine priorities and that conflict with each other or provide different results, is suitable for the individual. Identify and apply a composition that reflects the method and target with a high hit rate.

To this end, the optimal fortune reflecting unit 144 checks the purchase time, store, and number of the winning lottery ticket of each round or the user's previous lottery ticket through the accumulated information on the winning of a plurality of previous rounds or the user's previous lottery purchase history. After calculating the user's fortune at the point of time using multiple fortune determination methods, select the method with a high hit rate for the lucky number, purchase date, and direction of the member location-based store, and then select the weight for the fortune determination method to be applied to the user. It includes a fortune-based history analysis unit (144a) to reflect.

In addition, based on the personal information of the member user, a number of lucky numbers are obtained by performing multiple methods among four weeks, fortunes, feng shui geography, and constellations, and a lucky number weighting part that increases the weight for the lucky number, and the lucky date and time and at the corresponding lucky date and time. A personal fortune-telling reflecting unit (144d) including a fortune temporary weighting unit, a fortune-direction weighting unit that increases the weight for the direction of fortune based on the user's residence or current location (location of the member terminal), and a personal fortune-telling reflecting unit. A store selection unit (144b) for selecting stores with a high score among the stores corresponding to the corresponding direction among the sales stores in the location of the member or the set region by reflecting the direction weight of the fortune direction weighting unit of (144d), and a personal fortune reflecting unit Based on the information and weight obtained in (144d) and the weight for each method obtained in the fortune-based history analysis unit 144a and the store information obtained in the store selection unit 144b, the information to be reflected in the actual purchase proposal setting and a weight for calculating the weight It includes a reflecting portion (144c).

For example, the fortune-based history analysis unit 144a uses the personal fortune reflecting unit 144d based on the lottery number of the previous round and the winning lottery store information, and the lucky number according to the personal fortune at the right time for each round, After obtaining the lucky date and the lucky direction, among these various methods, select the methods with a high hit rate, such as the same as the lottery number of the previous round or the direction of the store, or specific details (number, date and time, direction, etc.) for each method and weight them. Can be set. For example, when analyzing the past winning lottery information for a specific user, A school's Myeongrihak key-based luck date and time calculation has the highest hit rate, B school feng shui-based luck direction calculation has the highest hit rate, and luck number is C school Since the protagonist-based lucky number calculation has the highest hit rate, a weight may be generated based on the fortune with the highest hit rate in the past and included in the weight reflecting unit 144c.

Thereafter, the lottery contents generation unit 145 selects and combines the numbers generated through the winning number analysis unit 143 by applying the weight of the weight reflecting unit 144c, and generates lottery contents including the recommended purchase date and direction. I can.

5 is a conceptual diagram illustrating another operation method of the optimal fortune reflecting unit 144 shown in FIG. 4.

The operation method described as an example in FIG. 4 above is based on past winning information (winning information of the entire lottery or winning information on the user's purchase history), and then selecting a target with a high hit rate among various philosophical methods, and then hitting the target with a high hit rate. It is a philosophical selection, and it may not be difficult to specify such a hit rate, or there may be no significant difference.

Therefore, according to the fortune-based history analysis unit 144a, only for obtaining calculation information on objects (luck number, luck date, luck direction, etc.) according to all kinds of philosophical methods, and which of these are to be reflected with more weight. You can also use past hit rates.

For example, the personal fortune reflector 144d calculates the user's lucky number according to various philosophical methods according to the date and time when the lottery ticket can be purchased for the corresponding round and sets the corresponding number information and weight, and sets the lucky date and time at the corresponding date and time. You can set the weight for the luck direction, and you can set the information and weight for the lucky direction. Each of these pieces of information and weights are provided to the weight reflecting unit 144c.

Furthermore, the store selection unit 144b selects stores corresponding to a specific region and direction based on the location of the current member or the residence set by the member by using the direction information of the luck direction weighting unit of the personal fortune reflecting unit 144d. After selecting the store in consideration of the score, the fitness level can be determined, and the result can be provided to the weight reflecting unit 144c, and the fortune-based history analysis unit 144a can provide information on the past member's lottery purchase history or past round lottery winning information. To select a philosophical method with a high hit rate for the member.

Through this, one of various philosophical methods itself may be selected, but each detailed target for the philosophical method may be selected, or each hit rate may be reflected in the weight reflecting unit 144c as a weight according to a probability value. In the illustrated case, a method of selecting one by one and providing a result of increasing the weight for a corresponding detailed object, but is not limited thereto.

In this way, the lottery content providing unit 140 according to an embodiment of the present invention basically generates lottery recommendation information content (lottery content) included in the subscription content, but induces member users to access each time the subscription content is updated. The latest artificial intelligence function and a philosophical analysis function that gives a sense of psychological stability are applied together to become a factor, thereby increasing the reliability and psychological satisfaction of users.

Now, it is necessary to effectively select and provide the content of continuous interest selected according to the user's personal interest. In the embodiment of the present invention, by applying the deep reinforcement learning method, it is possible to select the content that can induce the user's participation somewhat aggressively. To be there.

The deep reinforcement learning method is a game artificial intelligence, typically a more improved reinforcement learning method applied to AlphaGo. The reinforcement learning method is to learn how to perform control so that the reward increases through the reward obtained by repeating the experience of performing the action in an unknown environment. It is a way to improve one's own control performance after being evaluated for one action, and it is a way to learn artificial intelligence in a way that a person becomes familiar with the operation method of the game while playing the game and becomes good at the game.

To this end, if control information (e.g. game character operation information) is generated based on state information about the game environment (e.g., degree of manipulation at a time, operation content, relative information, target information, etc.), the expected reward (e.g., score By creating a function that provides) with a neural network, you learn how to generate control information to get better rewards. Here, a deep neural network is applied so that the state information can be learned even without specific refinement of the environment information by receiving the original information on the game screen as it is, which is referred to as the cardiovascular reinforcement learning method.

By applying such a deep reinforcement learning method, even if somewhat ambiguous state information is used, if the reward can be predicted according to the learning history, an artificial intelligence control unit can be configured that can generate control information in the direction in which the actual high reward is obtained. Deep reinforcement learning based on A3C (Asynchronous Advantage Actor-Critic) also appeared, which improved the learning speed and solved the problem of bias caused by limited learning information.

In the case of the A3C-based deep reinforcement learning method, a plurality of sub-A3C learning units and a global or main A3C learning unit that manages them in an integrated manner uses an asynchronous method. In the embodiment of the present invention, such learning information is used to reflect each user's taste. The synchronization method of is modified and applied differently from the existing one, which will be described in more detail with reference to FIG. 7.

6 is a block diagram showing a configuration of a personal optimal content selection unit according to an embodiment of the present invention, and FIG. 7 is a diagram showing an expanded configuration of a deep reinforcement learning unit according to an embodiment of the present invention.

First, the deep reinforcement learning unit shown as showing the configuration of the optimal content selection unit 180 for each individual according to an embodiment of the present invention is preferably A3C-based, but since the configuration of FIG. 6 can operate sufficiently, it is based on A3C. It may not be limited.

As shown, the individual optimal content selection unit 180 includes a previous content usage history collection unit 181 that collects previous content usage history from the content usage monitoring unit 160 and the previous content usage history collection unit 181. An interest level calculation unit 182 that calculates the level of interest in the characteristic information of the content according to the collection history, an LTV (LifeTime Value) calculation unit 183 that calculates the profit value according to the use of the content, and the interest level calculation unit 182 A reward calculation unit 184 that converts the result of using the content desired by the service provider into a numerical value by reflecting a preset weight in the result of the LTV calculation unit 183 and provides it as compensation information, and a content selection unit ( 120), the content feature information classification unit 185, which collects the selectable content or the characteristic information of the information obtained through, and provides it as status information, and the trend information obtained through the external trend information collection unit 170, to the content feature information. An external interest information conversion unit 186 correspondingly provided as status information, and a past content selection history and result accumulating unit 188 having past content selection information and a usage history of the subscription content generated accordingly, as a result, A deep reinforcement learning unit 187 that inputs the status information based on the learning contents through the information of the past content selection history and the result accumulator 188, and outputs the content characteristic information of which the compensation information is the largest as control information. And a content selection information providing unit 189 that receives control information of the deep reinforcement learning unit 187 and provides selectable content or information having content characteristic information for generating a new subscription content as content selection information.

The illustrated in-depth reinforcement learning unit 187 basically generates control information in a direction in which the compensation increases based on the status information and the compensation information, but refers to past histories, and learns the status information therein as control information. An actor to which an optimal control policy neural network model is applied that outputs the fitness according to the content characteristic information to be provided as a probability value, and content characteristic information preset in the environment according to the state information by inputting the same state information as the execution unit. A value estimation neural network model that converts the expected content usage result into a numerical value and outputs it as reward information when a subscription content is provided that includes information (e.g., information selected according to previous control information or control information to be output from this execution unit). This includes the applied critic. That is, the criticism unit predicts the compensation information provided by the reward operation unit 184 through a neural network, so that the control information can provide better results.

To this end, the deep reinforcement learning unit 187 must accurately calculate compensation information, which is a criterion for providing control information. Such compensation information substantially means the result of using the content desired by the service provider, and simply, if the user wants to increase the user use time, the user use time may be the compensation information. However, in the case of service users, since it is intended to obtain substantial profits by providing the corresponding subscription service, the purpose may be profits such as advertising subscriptions and purchase of paid contents, rather than an increase in content usage time itself. have. Therefore, the interest level calculation unit 182 uses the history provided by the previous content use history collection unit 181 to a preset algorithm (repeated access count, content use time, content satisfaction level expression (selection of preference)), and investment according to content use. The degree of interest obtained by comprehensive analysis of information generation, etc.) can be calculated as numerical information. Furthermore, the LTV calculation unit 183 may calculate a LifeTime Value (LTV) through a profit conversion rate calculation for a corresponding user or a profit value calculation according to content use through a previous content use history for the corresponding user. The reward calculation unit 184 calculates the content use result as a numerical value by reflecting a preset weight desired by the service provider in the results of the interest level calculation unit 182 and the LTV calculation unit 183, and uses the reward information as a deep reinforcement learning unit ( 187).

The state information largely includes environmental information on which content can be selected and information to reflect issues of high social interest.

First, the provided content characteristic information classification unit 185 is for accurately providing characteristic information about content or information for content creation as status information, for example, by integrating and grouping characteristic information or subdividing characteristic information as necessary. Thus, normalized information for grasping the content to be provided this time or information for content creation on a uniform basis (the overall characteristic information distribution information for the content to be provided this time or information for content creation, content with predetermined characteristic information (information ), the distribution state of characteristic information of individual contents, etc.) may be generated and provided to the deep reinforcement learning unit 187 as part of the state information.

The external interest information conversion unit 186 receives trend information through the external trend information collection unit 170, for example, the number of keywords with constant trend information (for example, keywords for 100 frequent keywords searched on the portal for one week and If it is a frequency value), it is converted into normalized information (classification of the characteristic information to which keywords belong, information on the importance distribution of characteristic information according to the frequency of keywords, etc.) for classification of the content characteristic information, and the deep reinforcement learning unit 187 Can be provided to.

The deep reinforcement learning unit 187 may generate, as probability information, information on the selection criteria for selection content that is predicted to have a higher reward by referring to such status information, compensation information, and past history. It should be larger than the predicted compensation in the case of selecting as control information.

This control information is a selection criteria for content selection, and is information on a criterion that it is desirable to select what kind of content feature information (probability information on whether it is desirable to select content having a certain distribution of characteristic information or information for content generation). I can.

According to the control information, the content selection information providing unit 189 selects or at least selects selectable content or information with content characteristic information for creating a new subscription content from among the content or content generation information generated by the content providing server 200. Provides information to choose from.

In the case of such deep reinforcement learning, the accuracy may increase as it is repeated. To reduce this learning period, a plurality of deep reinforcement learning units are operated in parallel in a plurality of similar learning environments, and such a plurality of deep reinforcement learning units learn. An A3C-based deep reinforcement learning unit can be configured that can improve performance by managing one learning data in an integrated manner and synchronizing it with individual deep reinforcement learning units.

In particular, in the case of the present invention, since the service server 100 is configured on a cloud basis, a plurality of servers that are physically or logically separated can be operated, so that each server is provided with a deep reinforcement learning unit to provide the member users in charge of it. It is possible to independently perform the optimal content selection learning and manage each of these learning data in an integrated manner.

7 is an exemplary view showing such a configuration, a main deep reinforcement that collects and reflects the learning results learned by a plurality of deep reinforcement learning units (187_1 to 187_n) and synchronizes the neural network of the deep reinforcement learning unit with its own neural network. It may further include a learning unit (187_M).

The main deep reinforcement learning unit 187_M may have integrated learning data having unbiased accuracy by asynchronously or synchronously collecting learning results from deep reinforcement learning units 187_1 to 187_n provided in different cloud servers. Therefore, the deep neural network of the main A3C learning unit 187_M maintains the highest learning state.

Originally, in the case of the existing A3C-based deep reinforcement learning method for games, when each individual sub-A3C learning unit synchronizes the main A3C learning unit and real-time learning data, all of them have the same learning result, so each characteristic cannot be reflected. The main A3C learning unit and the sub A3C learning unit asynchronously synchronize one by one individually in an asynchronous manner. Therefore, the main A3C learning unit itself has integrated learning data, but each sub-A3C learning unit uses slightly different learning data (using different learning data due to the interval between asynchronous).

That is, the operation according to each learning data is slightly different, but overall, control information is provided similarly to the operation according to the integrated learning data. This corresponds to a way to effectively play the'same' game by collecting multiple experiences in the completely identical'same' game. The goal of deep reinforcement learning for such game AI is to finally learn one of the main A3C learning units best.

However, in the case of the present invention, since it is about selecting preferred contents for users with different personal tastes, it would be nice if an integrated deep reinforcement learning unit that can effectively classify all personal tastes could be constructed. There is a problem with this take. In other words, when a single artificial intelligence learns the tastes of all users, unlike games, the learning load increases due to differences in environments in which the tastes of individual users are too different.

However, in that they are all'people' and have a universal taste, learning time can be reduced by slightly modifying the existing A3C deep reinforcement learning method.

Since the present invention uses a cloud-type service server, a group having a specific preference is placed on a specific cloud server by classifying users based on preferences according to personal information collected when a user registers as a member, and the cloud server has a similar preference. Each in-depth reinforcement learning department performs the selection of optimal content for the group. In this case, a user account whose preference group is changed through the actual content use history may be transferred to a cloud server in charge of the changed preference group.

For example, as shown, the sub-A3C learning unit 1 (187_1) for a user group having a specific preference performs learning to select individual optimized content according to the characteristics of the preference group. In this way, each of the sub-A3C learning units 187_1 to The learning data of 187_n) is collected by the main A3C learning unit 187_M in a synchronous or asynchronous manner to form a learned deep neural network according to the integrated learning data.

However, the present invention does not aim to best learn the main A3C learning unit 187_M, but uses the main A3C learning unit 187_M as a kind of reference information to grasp the universal preference of'people', and each individual The group operates as a self-learned deep neural network. However, if the compensation information according to the control information of a specific sub A3C learning unit (187_1 to 187_n) provides a result below the standard (that is, when the performance is degraded by too skewed learning data), the corresponding sub A3C learning unit and By performing synchronization asynchronously, the judgment result of the sub-A3C learning unit is returned to the universal level.

Through this, it is possible to reflect the preference characteristics of each group and provide the desired level of preferred content selection information at a fast learning speed.

8 is a flow chart showing an operation process according to an embodiment of the present invention.

As shown in the figure, the service server processes membership registration including collection of personal information for the accessing user, and creates and provides a personal access address for confirming subscription content for the member.

Afterwards, the service server generates a store score based on the lottery winning information of the previous round in order to generate basic lottery contents, and generates an expected lottery winning number through neural network analysis (GAN) based on the accumulated lottery winning information.

Then, among the generated lottery winning predicted numbers, the selected lottery recommendation content is generated by reflecting the fortune with high accuracy among the user's personal fortune.

On the other hand, if the content providing server generates content that can be selected as a subscription content or information necessary for content creation together with characteristic information according to a predefined criterion, the corresponding characteristic information is collected, and after collecting external companion information, the user's Previous content usage history is analyzed through in-depth reinforcement learning to select content suitable for the user.

After creating subscription content by combining lottery recommended content and content selected according to the user, it is registered in a personal access address and provided to the user.

Thereafter, the user's subscription content usage history is monitored to collect the content usage history, the user's investment information is collected and analyzed, and the lottery-related content creation conditions are updated based on the usage history and actual investment information (user lottery purchase history). Lottery recommendation weight is changed according to), and the learning contents for selecting the optimal user contents are accumulated and updated.

Through this process, it is possible to increase the profitability of the service provider and increase the user's satisfaction by increasing the user's interest in the subscription content and inducing repeated access and continued interest.

The method according to the embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include physical hard disks, solid state drives (SSDs), web hard, large-capacity data storage means such as cloud storage, magnetic media such as floppy disks and magnetic tapes ( magnetic media, optical media such as CD-ROM, DVD, magneto-optical media such as floptical disk, and ROM, RAM, Hardware devices specially configured to store and execute program instructions such as flash memory and the like are included. Examples of program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

In the above, preferred embodiments according to the present invention have been illustrated and described. However, the present invention is not limited to the above-described embodiments, and any person with ordinary knowledge in the technical field to which the present invention pertains may implement various modifications without departing from the gist of the present invention appended in the claims. .

10: user terminal 100: service server
110: member management unit 120: content selection unit
130: individual content generation unit 140: lottery content providing unit
150: personalized URL management unit 160: content usage monitoring unit
170: external trend information collection unit 180: individual optimal content selection unit
200: content providing server 300: database

Claims (10)

As a cloud-based personalized content subscription service providing system that provides personalized subscription content that is a mixture of short-term interest content including lottery recommendation information content and persistent interest content selected according to the user's personal interest, to a user terminal,
A content providing server that generates content that can be selected as a subscription content or information necessary for content creation together with characteristic information according to a predefined criterion;
Process membership sign-up, including collection of personal information for access users, create and provide a personal access address for confirming subscription content for members, and predict lottery winning numbers through neural network analysis based on accumulated lottery winning information. Generates and generates selected lottery recommended content by reflecting the user's personal fortune among the generated lottery winning numbers, and collects user use information for the content exposed through the personal access address, and is generated by the content providing server. Based on the characteristic information of the content, the personalized content with high user interest is selected through neural network analysis, and then the subscription content created by combining with the recommended lottery content is registered in the personal access address, Includes a service server that manages all information about online based on personal access address,
Individual optimal content selection unit included in the service server,
A previous content usage history collection unit for collecting a previous content usage history from the content usage monitoring unit;
An interest level calculation unit for calculating an interest level for characteristic information of the content according to the collection history of the previous content use history collection unit;
An LTV (LifeTime Value) calculating unit that calculates a profit value according to content use;
A reward calculation unit that converts a result of using content desired by a service provider into a numerical value by reflecting a preset weight on the results of the interest level calculation unit and the LTV calculation unit and provides compensation information;
A provided content characteristic information classification unit for collecting selectable content or characteristic information of information obtained through a content selection unit to generate a new subscription content and providing it as status information;
An external interest information conversion unit for providing trend information obtained through the external trend information collection unit as status information by corresponding to the content feature information;
A past content selection history and result accumulating unit having past content selection information and a usage history of the subscription content generated accordingly;
An in-depth reinforcement learning unit for outputting, as control information, content characteristic information for which the compensation information may be the largest, based on the learning contents through the information of the past content selection history and the result accumulator;
A cloud-based personalized content subscription service, comprising: a content selection information providing unit that receives control information of the deep reinforcement learning unit and provides selectable content or information having content characteristic information for creating a new subscription content as content selection information. Delivery system.
The method of claim 1, wherein the service server
A member management unit that collects member subscription information including personal information from a connected user terminal, performs member registration, and collects actual investment information including purchase of a lottery ticket from the user terminal;
A personalized URL (Uniform Resource Locator) management unit that generates a personal access address for checking subscription content for a registered user and provides a web hosting service to check the subscription content through the personal access address;
A lottery content provider configured to generate a lottery-winning prediction number based on the accumulated lottery winning information through neural network analysis, and to generate selected lottery recommendation content by reflecting the user's personal fortune among the generated lottery-winning prediction numbers;
A content usage monitoring unit that collects user usage information for content exposed through a personal access address;
A content selection unit for collecting content generated through the content providing server or information necessary for content generation and characteristic information for each content or information;
An external trend information collection unit that receives social interest as trend information obtained through big data analysis of a preset portal site;
Neural network analysis is performed so that the use information obtained through the content use monitoring unit is improved based on the characteristic information of the selectable content obtained through the content selection unit or the information necessary for content generation, and the trend information obtained through the external trend information collection unit. A personal optimal content selection unit that provides content selection information through;
According to the content selection information of the individual optimal content selection unit, the content generated through the content providing server or information necessary for content generation is selected to generate individual content, and the lottery recommended content provided by the lottery content providing unit is added and subscribed. A cloud-based personalized content subscription service providing system, comprising: a personal content generation unit provided to the personalized URL management unit as content.
The method of claim 2, wherein the member management unit
Member terminal interlocking unit that checks the connection according to the service server address set in advance or the personal access address issued to the member through the user terminal, and in the case of access to the personal access address, checks the pre-registered access authentication number and provides subscription content Wow;
A member subscription processing unit that collects personal information including a region and date of birth as member subscription information from a user terminal connected to the address of the service server through the member terminal linkage unit to perform member registration;
A member information management unit for creating an account for a registered user, correspondingly registering a personal access address issued to the user, and storing information provided from a user terminal or usage information obtained through a content usage monitoring unit;
A member investment information collection management unit for collecting investment history including a history of purchasing a lottery ticket of a corresponding user from a member-registered user terminal;
Cloud-based personalized content, characterized in that it includes a subscription information management unit that determines the type of subscription content to be provided according to a member's level, and delivers a subscription content delivery cycle and a notification about the generated subscription content to a registered user terminal when registering as a member. Subscription service provision system.
The method of claim 2, wherein the lottery content providing unit
The winning number analysis unit is included, but the winning number analysis unit is configured to predict the number through a generative adversarial network (GAN) based on the accumulated previous winning number information. A cloud-based personalized content subscription service providing system, characterized in that it updates itself by reflecting information.
The method of claim 2, wherein the lottery content providing unit
A winning information management unit that collects accumulated winning information of the last round and winning information of the lottery ticket purchased by the user;
A store information management unit that manages the location and winning history of the registered store;
A winning number analysis unit for generating an expected lottery winning number through neural network analysis based on the accumulated winning information;
Among the predicted lottery winning numbers generated through the winning number analysis unit, the user's personal fortune is reflected as a preset weight to select a personalized predicted winning number, and the user's personal fortune is reflected as a preset weight to select the purchase time and store. An optimal fortune-telling reflector;
A cloud-based personalized content subscription service providing system, comprising: a lottery content generation unit that generates lottery content including lottery numbers selected through the optimal fortune reflecting unit, purchase time, and store recommendation information.
The method of claim 5, wherein the optimal fortune reflecting unit checks the purchase time, store, and number of the winning lottery ticket of each round or the user's previous lottery ticket through the accumulated information on the winning of a plurality of previous rounds or the user's previous lottery purchase history. After calculating the user's fortune by multiple fortune determination methods, select the method with high hit rate for the lucky number, purchase date, and member location-based store direction by type, and reflect the weight for the fortune determination method to be applied to the user. Cloud-based personalized content subscription service providing system, characterized in that it comprises a fortune-based history analysis unit.
delete The method according to claim 1, wherein the deep reinforcement learning unit
An actor to which an optimal control policy neural network model is applied, which learns the state information and outputs a fitness level according to the content feature information to be provided as control information as a probability value;
Value calculation that converts the expected content use result into a numerical value and outputs it as reward information when a subscription content with preset content characteristic information is provided in an environment according to the status information by inputting the same status information as the execution unit. A cloud-based personalized content subscription service providing system comprising a critic to which a neural network model is applied.
The method according to claim 8, further comprising a main deep reinforcement learning unit for collecting and reflecting the learning result learned by the deep reinforcement learning unit and synchronizing the neural network of the deep reinforcement learning unit with its own neural network, wherein the main deep reinforcement learning unit is Asynchronously or synchronously, each learning result is asynchronously or synchronously collected from deep reinforcement learning units provided in other cloud servers, and if compensation information according to the control information of a specific deep reinforcement learning unit provides a result below the standard, asynchronous with the corresponding deep reinforcement learning unit Cloud-based personalized content subscription service providing system, characterized in that the synchronization is performed individually.
Cloud-based personalized content subscription using a cloud-based personalized content subscription service providing system that provides personalized subscription content that is a mixture of short-term interest content including lottery recommendation information content and continuous interest content selected according to the user's personal interest level As a method of providing services,
Processing, by the service server, a member subscription including collection of personal information for the accessing user, and generating and providing a personal access address for confirming the subscription content for the member;
Generating, by the service server, an estimated lottery winning number through neural network analysis based on the accumulated lottery winning information, and generating recommended lottery content selected by reflecting the user's personal fortune among the generated lottery winning prediction numbers;
Generating, by the content providing server, content that can be selected as a subscription content or information necessary for content creation together with characteristic information according to a predefined criterion;
The service server collects user usage information for the content exposed through the personal access address, and selects personalized content with high user interest based on the characteristic information of the content generated by the content providing server through neural network analysis. And then generating a subscription content by combining it with the recommended lottery content;
And registering, by the service server, the generated subscription content to the personal access address,
Individual optimal content selection unit included in the service server,
A previous content usage history collection unit for collecting a previous content usage history from the content usage monitoring unit;
An interest level calculation unit for calculating an interest level for characteristic information of the content according to the collection history of the previous content use history collection unit;
An LTV (LifeTime Value) calculation unit that calculates a revenue value according to content use;
A reward calculating unit for converting a result of using content desired by a service provider into a numerical value by reflecting a preset weight on the results of the interest level calculating unit and the LTV calculating unit and providing compensation information;
A provided content characteristic information classification unit for collecting selectable content or characteristic information of information obtained through a content selection unit to generate a new subscription content and providing it as status information;
An external interest information conversion unit for providing trend information obtained through the external trend information collection unit as status information by matching the content feature information;
A past content selection history and result accumulating unit having past content selection information and a usage history of the subscription content generated accordingly;
An in-depth reinforcement learning unit for outputting, as control information, content characteristic information for which the compensation information is the largest, based on the learning contents through the information of the past content selection history and the result accumulating unit;
A cloud-based personalized content subscription service, comprising: a content selection information providing unit that receives control information of the deep reinforcement learning unit and provides selectable content or information having content characteristic information for creating a new subscription content as content selection information. Delivery method.

Images