Method and apparatus for managing multiple services related to online shopping center using neural network

Abstract

Methods and apparatus are provided for managing a plurality of services related to an online shopping center using a neural network. Basic information of a user terminal and online activity history information of the user terminal are obtained according to the condition that the user terminal accesses a first webpage, and preference information of the user terminal is determined according to the online activity history of the user terminal. The method comprises the following steps: sending an attraction message to a user terminal according to historical information, including attraction information to the user terminal, of the user terminal after a preset time from the time when the user terminal accesses a first webpage, and requesting the user terminal to access a second webpage linked with the attraction message at a preset position on the webpage where the user terminal is located, therefore, the information of the second webpage is sent to the user terminal, namely the second webpage. Comprising product information recommended to the user terminal, and the recommendation of the product included in the second webpage by the user terminal can comprise the step of sending information about at least one matched product to the user terminal based on the searched product page.

Description

TECHNICAL FIELD

[0001] Embodiments of the disclosure relate to a method and apparatus for managing a plurality of services related to an online shopping mall using a neural network. BACKGROUND

[0002] In the modern digital age, online shopping provides consumers with easier and more convenient access to a variety of products and services. However, it is difficult to quickly find products that users like by category due to information overload.

[0003] To meet these requirements, existing online shopping malls provide customized content by researching users' preferences in advance, mainly through surveys. However, it is inconvenient to ask users to answer surveys, and there are limitations in determining users' exact preferences and reflecting users' preferences in real time only through surveys.

[0004] Therefore, there is a need for a user-centered customized shopping experience, and in particular, there is a need to utilize a neural network to provide a customized web page reflecting big data such as each user's taste, preference, purchase history, etc.

[0005] Therefore, there is a need for a method and apparatus for managing a plurality of services related to an online shopping mall by learning and analyzing users' behavior patterns, search history, purchase history, and various features using a neural network. SUMMARY

[0006] Problems to be Solved by the Invention

[0007] Embodiments of the disclosure can provide a method and apparatus for managing a plurality of services related to an online shopping mall using a neural network.

[0008] Technical challenges to be achieved in embodiments are not limited to the above and other technical challenges not mentioned can be considered by those skilled in the art from various embodiments described below.

[0009] Method for Solving the Problem

[0010] The server according to the embodiment provides information on at least one matching product and a recommended product to the user terminal using a neural network. The method for recommending a product based on the user terminal accessing a first web page to acquire basic information of the user terminal. And the online activity history information of the user terminal, determine the preference information of the user terminal according to the online activity history information of the user terminal, and send the bait message containing the attraction information of the user terminal to the user terminal after a preset time from the time of accessing the first web page, and send the bait message to the preset position of the web page where the user terminal is located, and based on the user terminal. Access to the second web page linked to the bait message sends information about the second web page to the user terminal, and the second web page page includes product information recommended for the user terminal, and based on the user terminal browsing the product page can include a step of providing the user terminal with information about at least one matching product included in the recommended product in the second web page. For example, determine the recommended product of the user terminal according to the preference information of the user terminal, the attraction information of the user terminal includes the basic information of the user terminal, the preference information of the user terminal, and the preference information of the user terminal. Based on the recommended product information generated, the at least one matching product is a product matched with the recommended product, and the at least one matching product includes the basic information of the user terminal, the preference information of the user terminal, and the user-determined product information recommended for the terminal based on the product matching model based on the first neural network, and at least one matching product is selected on the product page according to the matching relationship between the recommended product and the at least one matching product. The position of the display image can be determined.

[0011] Inventive Effects

[0012] As described above, the present application has the following effects:

[0013] In the embodiment of the present application, the server learns and analyzes the purchase history of the user terminal through the neural network, and generates the web page in real time, thereby improving the user experience and providing more personalized services for the user terminal.

[0014] According to the embodiment, the server determines the recommended product based on the preference information of the user terminal analyzed using the neural network, and generates a phrase for inducing the purchase of the recommended product according to the preference information, thereby giving preference to the user terminal. You can improve the purchase rate of the product while providing the product.

[0015] According to the embodiment, the server classifies user terminals with similar tastes and purchase histories of recommended products for the user terminal, and selects a product with a higher purchase rate among the classified user terminals as a matching product of the recommended product. By determining, the matching product that the user terminal is likely to purchase can be provided together with the recommended product.

[0016] Effects obtainable from the embodiments are not limited to what has been described herein above, and other effects that are not described can be also derived from combinations of technical characteristics of the embodiments, by a person skilled in the art based on the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A method according to an embodiment is shown, in which a server provides information about at least one matching product to a user terminal using a neural network.

[0018] Figure 2 is a diagram showing an example of a preference prediction model according to an embodiment.

[0019] Figure 3 A flowchart of a method according to an embodiment is shown, in which a server provides information about at least one matching product to a user terminal using a neural network.

[0020] Figure 4 is a block diagram showing a configuration of a server according to an embodiment. DETAILED DESCRIPTION

[0021] BRIEF DESCRIPTION OF DRAWINGS Figure 1 is a diagram showing a structure of an electronic device to which embodiments are applicable.

[0022] Figure 1 is a block diagram of an electronic device in a network environment according to various embodiments. Referring to Figure 1 In the network environment, the electronic device communicates with another electronic device through a first network (e.g., a short-range wireless communication network), or a second network (e.g., a long-range wireless communication network). The long-range wireless communication network can communicate with at least one of them. According to an embodiment, the electronic device can communicate with another electronic device through the server. According to an embodiment, the electronic device includes a processor, a memory, an input module, an audio output module, a display module, an audio module, a sensor module, an interface, a connection terminal, a haptic module, a camera module, a power management module, a battery, a communication module, which can include a recognition module or an antenna module. In some embodiments, at least one of these components (e.g., the connection terminal) can be omitted, or one or more other components can be added to the electronic device. In some embodiments, some of these components (e.g., the sensor module, the camera module, or the antenna module) can be integrated into one component (e.g., the display module). The electronic device can also be referred to as a client, terminal, or peer.

[0023] For example, the processor can execute software (e.g., a program) to control at least one other component (e.g., a hardware or software component) of an electronic device connected to the processor and perform various data processing or operations. According to one embodiment, as at least a part of the data processing or computation, the processor stores a command or data received from another component (e.g., a sensor module or a communication module) in a volatile memory and processes the command or data stored in the volatile memory. It can be processed and the result data can be stored in a non-volatile memory. According to one embodiment, the processor is a main processor (e.g., a central processing unit or an application processor) or an auxiliary processor (e.g., a graphics processing unit, a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor) that can operate independently or together. For example, when the electronic device includes a main processor and an auxiliary processor, the auxiliary processor can be configured to use less power than the main processor or to be dedicated to a specified function. The co-processor can be implemented separately from the main processor or as part of the main processor.

[0024] The co-processor can act on behalf of the main processor, for example, when the main processor is in an inactive (e.g., sleep) state, or act together with the main processor when the main processor is in an active (e.g., application execution) state, and the electronic device can control at least some functions or states related to at least one component (e.g., a display module, a sensor module, or a communication module). According to one embodiment, the co-processor (e.g., an image signal processor or a communication processor) can be implemented as part of another functionally related component (e.g., a camera module or a communication module). According to one embodiment, the auxiliary processor (e.g., a neural network processing unit) can include a hardware structure dedicated to processing an artificial intelligence model.

[0025] An artificial intelligence model can be created through machine learning. For example, such learning can be performed in the electronic device itself that executes the artificial intelligence model, or can be performed through a separate server. The learning algorithm can include, for example, and without limitation, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. The artificial intelligence model can include multiple artificial neural network layers. The artificial neural network includes a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a belief deep network (DBN), a bidirectional recurrent deep neural network (BRDNN), etc. It can be one of a deep Q network or a combination of two or more of the above-described networks, but is not limited to the above-described examples. In addition to the hardware structure, the artificial intelligence model can additionally or alternatively include a software structure.

[0026] The memory can store various data used by at least one component (e.g., a processor or a sensor module) of an electronic device, for example, input data or output data, and instructions for the same. The memory can include a volatile memory or a non-volatile memory.

[0027] The program can be stored in the memory as software and can include, for example, a kernel, middleware, or an application program.

[0028] The input module can receive a command or data to be used by at least one component (e.g., a processor) of the electronic device, from the outside (e.g., a user) of the electronic device. The input module can include, for example, a microphone, a mouse, a keyboard, a key (e.g., a button), or a digital pen (e.g., a stylus pen).

[0029] The sound output module can output sound signals to the outside of the electronic device. The sound output module can include, for example, a speaker or a receiver. The speaker can be used for general purposes, such as multimedia play or recording play. The receiver can be used to receive calls. According to an embodiment, the receiver can be implemented as separate from, or as part of the speaker.

[0030] The display module can visually provide information to the outside (e.g., to the user) of the electronic device. The display module can include, for example, a display, a hologram device, or a projector and a control circuit for controlling the same. According to an embodiment, the display module can include a touch sensor configured to detect a touch, or a pressure sensor configured to measure the intensity of force incurred by the touch.

[0031] The audio module can convert a sound into an electrical signal or vice versa. According to an embodiment, the audio module can obtain sound through a microphone, the sound output module, or an external electronic device (e.g., an electronic device) connected to the electronic device directly or wirelessly (e.g., a speaker or a headphone).

[0032] The sensor module can detect an operational state (e.g., power or temperature) or an external environmental state (e.g., a user state) of the electronic device, and generate an electrical signal or data value corresponding to the detected state. According to an embodiment, the sensor module can include, for example, a motion sensor, a gyro sensor, a barometric sensor, a magnetic sensor, an acceleration sensor, a grip sensor, a proximity sensor, a color sensor, an infrared sensor, a biometric sensor, an acoustic sensor, a temperature sensor, a humidity sensor, or a light sensor.

[0033] The interface can support one or more designated protocols for enabling the electronic device to directly or wirelessly connect to the external electronic device (e.g., electronic device). According to an embodiment, the interface can include, for example, a high definition multimedia interface (HDMI), a universal serial bus (USB) interface, an SD card interface, or an audio interface.

[0034] The connection terminal can include a connector through which the electronic device can be physically connected to the external electronic device (e.g., electronic device). According to an embodiment, the connection terminal can include, for example, a HDMI connector, a USB connector, an SD card connector, or an audio connector (e.g., a headphone connector).

[0035] The haptic module can convert electrical signals into mechanical stimuli (e.g., vibration or movement) or electrical stimuli that a user can feel through a tactile sensation or a kinesthetic sensation. According to an embodiment, the haptic module can include, for example, a motor, a piezoelectric element, or an electrical stimulation device.

[0036] The camera module can capture still images and videos. According to an embodiment, the camera module can include one or more lenses, image sensors, image signal processors, or flashes.

[0037] The power management module can manage power supplied to the electronic device. According to an embodiment, the power management module can be implemented as at least a part of, for example, a power management integrated circuit (PMIC).

[0038] The battery can supply power to at least one component of the electronic device. According to an embodiment, the battery can include, for example, a primary cell which is not rechargeable, a secondary cell which is rechargeable, or a fuel cell.

[0039] The communication module can support establishing a direct (e.g., wired) or wireless communication channel between the electronic device and an external electronic device (e.g., an electronic device or a server), and performing communication through the established communication channel. The communication module operates independently of the processor (e.g., an application processor) and can include one or more communication processors supporting direct (e.g., wired) communication or wireless communication. According to an embodiment, the communication module is a wireless communication module (e.g., a cellular communication module, a short-range wireless communication module, or a global navigation satellite system (GNSS) communication module) or a wired communication module (e.g., a local area network (LAN) communication module) or a power line communication module). Among these communication modules, the corresponding communication module is a first network (e.g., a short-range communication network such as Bluetooth, wireless-fidelity (Wi-Fi) direct, or infrared data association (IrDA)) or a second network (e.g., a cellular network, a 5G network, which can communicate with an external electronic device through a next-generation communication network, the Internet, or a computer network (e.g., a LAN or a WAN) and the like. These various types of communication modules can be integrated into one component (e.g., a single chip) or can be implemented as a plurality of separate components (e.g., a plurality of chips). The wireless communication module can use subscriber information (e.g., an international mobile subscriber identifier (IMSI)) stored in a subscriber identification module to identify or authenticate the electronic device within the communication network (e.g., the first network or the second network).

[0040] The wireless communication module can support a 5G network after a 4G network and next-generation communication technology such as an NR access technology (New Radio Access Technology). The NR access technology provides high-speed transmission of large-capacity data (eMBB (enhanced mobile broadband)), minimization of terminal power consumption and multiple terminal access (mMTC (massive machine type communication)), or high reliability and low latency (URLLC (URLLC) can support ultra-reliability and low latency). The wireless communication module 192 can support, for example, a high frequency band (e.g., a millimeter wave band) to achieve a high data rate. The wireless communication module uses various technologies to secure the performance of the high frequency band, such as beamforming, massive MIMO (Multiple Input Multiple Output), and full-dimensional multi-input / output (FD) technologies. Multi-dimensional MIMO (MIMO), array antenna, analog beamforming, or massive antenna. The wireless communication module can support various requirements specified in the electronic device, the external electronic device, or the network system (e.g., a secondary network). According to an embodiment, the wireless communication module has a peak data rate (e.g., 20 Gbps or more) for implementing eMBB, a loss coverage range (e.g., 164 dB or less) for implementing mMTC, or a U-plane delay (e.g., downtime) for implementing URLLC. Link (DL) and uplink (UL) each 0.5 milliseconds or less, or 1 millisecond or less for round trip) can be supported.

[0041] The antenna module can transmit or receive a signal or power to or from the outside (e.g., an external electronic device). According to an embodiment, the antenna module can include an antenna including a radiator made of a conductor or a conductive pattern formed on a substrate (e.g., a PCB). According to an embodiment, the antenna module can include a plurality of antennas (e.g., array antennas). In this case, at least one antenna suitable for a communication method used in a communication network such as a first network or a second network can be selected from the plurality of antennas, for example, by the communication module. A signal or power can be transmitted or received between the communication module and the external electronic device through the selected at least one antenna. According to some embodiments, in addition to the radiator, other components (e.g., a radio frequency integrated circuit (RFIC)) can be additionally formed as part of the antenna module.

[0042] According to various embodiments, the antenna module can form a millimeter wave antenna module. According to an embodiment, the millimeter wave antenna module includes a printed circuit board, an RFIC disposed on or adjacent to a first side (e.g., a bottom side) of the printed circuit board and capable of supporting a designated high frequency band (e.g., a millimeter wave band), and a plurality of antennas (e.g., array antennas) can be disposed on or adjacent to a second side (e.g., a top or a side) of the printed circuit board and capable of transmitting or receiving a signal in the designated high frequency band.

[0043] At least some of the components are connected to each other by a communication method between a peripheral device (e.g., a bus, a general purpose input and output (GPIO), a serial peripheral interface (SPI), or a mobile industry processor interface (MIPI)) and a signal. ((e.g., commands or data) can be exchanged with each other.

[0044] According to an embodiment, commands or data can be transmitted or received between the electronic device and an external electronic device via a server connected to the second network. Each of the external electronic devices can be of a same or different type as the electronic device. According to an embodiment, all or a part of operations that are to be executed at the electronic device can be executed at one or more of the external electronic devices. For example, when the electronic device is required to perform a function or a service automatically, or in response to a request from a user or another device, the electronic device can perform, instead of, or in addition to, the function or the service, one or more functions or services that pertain to at least a part of the function or the service. The service itself can be requested to perform at least part of the function or the service. One or more external electronic devices that receive the request can execute at least a part of the requested function or service, or an additional function or service related to the request, and transmit the execution results to the electronic device. The electronic device can process the results as is, or additionally, and provide them as at least part of a response to the request. To this end, cloud computing, distributed computing, mobile edge computing (MEC), or client-server computing techniques can be used, for example. The electronic device can use distributed computing or mobile edge computing, etc., to provide an ultra-low-latency service. In another embodiment, the external electronic devices can include Internet of Things (IoT) devices. The server can be an intelligent server using machine learning and / or a neural network. According to an embodiment, the external electronic device or the server can be included in the second network. The electronic device can be applied to intelligent services (e.g., smart home, smart city, smart car, or health care) based on 5G communication technology and Internet of Things (IoT) related technology.

[0045] The server is connected to the electronic device and can provide a service to the connected electronic device. Also, the server can perform a membership registration process, store and manage various information of users who have registered as members, and provide various purchase and payment functions related to the service. Also, the server can share execution data of a service application running on each of a plurality of electronic devices in real time, so that the service can be shared among users. The server can have the same hardware configuration as a typical web server or service server. However, in terms of software, it can be implemented by any language such as C, C++, Java, Python, Golang, and Kotlin, and can include program modules that perform various functions. Also, the server is generally a computer system that is connected to an unspecified number of clients and / or other servers through an open computer network such as the Internet, receives a work performance request from the clients or other servers, derives and provides a work result, and it refers to computer software (server program) installed for this purpose. Also, the server includes a series of application programs running on the server, and in some cases, various databases (DB: Database, hereinafter referred to as "DB") built-in or built-out, in addition to the above-mentioned server program, and it should be understood as a broad concept as appropriate. Therefore, the server classifies and stores membership registration information and various information and data about the game in the DB, and manages the DB, which can be implemented inside or outside the server. Also, the server can be implemented using general server hardware and server programs that are provided in various ways according to operating systems (e.g., Windows, Linux, UNIX, and Macintosh), representative examples of which include a web service that can be implemented using IIS (Internet Information) used in a Windows environment, CERN, NCSA, APPACH, TOMCAT, etc. used in a Unix environment. Also, the server can be linked with an authentication system and a payment system for user authentication for the service or payment for purchase related to the service.

[0046] The first and second networks refer to a connection structure that allows exchange of information between each node (e.g., terminal and server), or a network that connects servers and electronic devices. The first and second networks are the Internet, LAN (Local Area Network), wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), 3G, 4G, LTE, 5G, Wi-Fi, etc., but are not limited thereto. The first and second networks can be a closed network such as a LAN or a WAN, but are preferably an open network such as the Internet. The Internet includes protocols such as TCP / IP protocol, TCP, and UDP (User Datagram Protocol), and various services existing at the upper layer, such as HTTP (Hypertext Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), SNMP (Simple Network Management Protocol), NFS (Network File System), and NIS (Network Information Service).

[0047] The database can have a general data structure implemented in the storage space (hard disk or memory) of the computer system using a database management program (DBMS). The database can have a data storage format that allows free search (extraction), deletion, editing, addition of data, etc. The database can be a relational database management system (RDBMS) such as Oracle, Infomix, Sybase, and DB2, or an object-oriented database management system such as Gemstone, Orion, and O2. This is implemented using systems (OODBMS) and XML native databases (e.g., Excelon, Tamino, and Sekaiju) and has its own functions to achieve this, and you can have appropriate fields or elements.

[0048] According to an embodiment, the programs can include an operating system, middleware, or an application for controlling one or more resources of an electronic device, executable on the operating system. The operating system can include, for example, Android TM, iOS TM, Windows TM, Symbian TM, Tizen TM, or Bada TM. At least some of the programs can be preloaded into the electronic device, for example, during manufacture, or can be downloaded or updated from an external electronic device (e.g., an electronic device or a server) while in use by a user. All or part of the programs can include a neural network.

[0049] The operating system can control the management (e.g., allocation or recovery) of one or more system resources (e.g., process, memory, or power) of the electronic device. Additionally or alternatively, the operating system can include other hardware devices of the electronic device, such as an input module, an audio output module, a display module, an audio module, a sensor module, an interface, a haptic module, a camera module, a power management module, a battery, and it can include one or more drivers for driving a communication module, a subscriber identification module, or an antenna module.

[0050] The middleware can provide various functions to the application program so that functions or information provided from one or more resources of the electronic device can be used by the application program. According to one embodiment, the middleware can dynamically delete some existing components or add new components. According to one embodiment, at least a portion of the middleware can be included as a part of the operating system or can be implemented as a separate software from the operating system.

[0051] The application program can include, for example, various application programs. According to one embodiment, the application program can also include an information exchange application program capable of supporting information exchange between the electronic device and an external electronic device. The information exchange application program can include, for example, a notification relay application program configured to transmit designated information (e.g., a call, a message, or an alarm) to an external electronic device, or a device management application program configured to manage an external electronic device. For example, the notification relay application can deliver notification information corresponding to a designated event (e.g., mail receipt) generated in another application (e.g., an e-mail application 269) of the electronic device to an external electronic device. Additionally or alternatively, the notification relay application can receive notification information from an external electronic device and provide it to a user of the electronic device.

[0052] The device management application program can, for example, control the power (e.g., turn on or off) or function of an external electronic device or some components thereof (e.g., a display module or a camera module of the external electronic device) in communication with the external electronic device. The electronic device (e.g., brightness, resolution, or focus) can be controlled. The device management application program can additionally or alternatively support the installation, deletion, or update of an application program running on the external electronic device.

[0053] Throughout this specification, neural networks, neural networks, and network functions can be used with the same meaning. A neural network can consist of a set of interconnected computing units, often referred to as "nodes". These "nodes" can also be referred to as "neurons". A neural network consists of at least two or more nodes. The nodes (or neurons) that make up the neural network can be interconnected by one or more "links".

[0054] In a neural network, two or more nodes connected by a link can form a relative input node and output node relationship. The concepts of input node and output node are relative, any node that is in an output node relationship with one node can be in an input node relationship with another node, and vice versa. As described above, an input node to output node relationship can be created around a link. One or more output nodes can be connected to an input node through a link, and vice versa.

[0055] In the relationship between the input node and the output node connected by a link, the value of the output node can be determined based on the data input to the input node. Here, the nodes connecting the input node and the output node can have weights. The weights can be variable and can be changed by a user or an algorithm in order for the neural network to perform the desired function. Here, the edges or links interconnecting the input nodes and the output nodes have weights that can be variably applied by a user or an algorithm to perform the functions required by the neural network. For example, when one or more input nodes are connected to an output node through respective links, the output node is set to the output of the input nodes connected to the output node and the link corresponding to each input node. The node value can be determined according to the weight.

[0056] As described above, in a neural network, two or more nodes are interconnected by one or more links to form an input node and output node relationship within the neural network. The characteristics of the neural network can be determined according to the number of nodes and links in the neural network, the correlation between the nodes and links, and the weight values assigned to each link. For example, if there are two neural networks with the same number of nodes and links and different weight values between the links, the two neural networks can be identified as different from each other.

[0057] Figure 1 A method according to an embodiment is shown, in which a server provides information about at least one matching product to a user terminal using a neural network. Figure 1 Embodiments of the present disclosure can be combined with various embodiments of the present disclosure.

[0058] Referring to Figure 1 In step S110, the server acquires basic information of the user terminal and online activity history information of the user terminal according to the behavior of the user terminal (e.g., an electronic device) accessing the first web page.

[0059] The server can be a server providing a product sale service for the user terminal. For example, the server can be a server of a website providing information on products such as fashion accessories, perfumes, and cosmetics and managing sale of the products. For example, the server can collect various information on the user terminal and determine preference information of the user terminal. The server can determine a recommended product according to the preference information of the user terminal, and generate an incentive message inducing the user terminal to purchase the recommended product and transmit the same to the user terminal. At this time, the server can provide the user terminal with a web page composed of the product recommended for the user terminal according to interaction (e.g., a click) with the attraction message of the user terminal. Based on the user terminal searching for a specific recommended product on the web page composed of the recommended product of the user terminal, the server can determine information of at least one matching product matching the specific recommended product and provide the same to the user terminal.

[0060] The user terminal can be a terminal using a product sale service provided by the server. For example, the user terminal can be a terminal connected to a website providing information on products such as fashion accessories, perfumes, and cosmetics and selling the products.

[0061] For example, the user terminal can log in to a website providing a product sale service of the server. For example, the server can acquire basic information of the user terminal and online activity history information of the user terminal according to login of the user terminal. For example, the user terminal can log in to the website by inputting an ID and a password registered in the website.

[0062] For example, the user terminal can access a first web page through login. Here, the first web page can be a main web page of the website providing the product sale service.

[0063] The basic information of the user terminal is basic information of the user, and for example, can include an ID, an age, a gender, an address, body information, etc. Here, the ID can include at least one of an ID of the user terminal subscribing to the website or an SNS (Social Network Service) ID. The body information can be information on a body input by the user terminal. For example, the body information can include a height, a weight, a shoulder width, a bust, and a waist.

[0064] The information on the online activity history of the user terminal can be information indicating an activity history of the user terminal on the website. For example, the information on the online activity history can include purchase history information, query history information, and a web log.

[0065] The purchase history information is history information of a product purchased by the user terminal, for example, a product type of the purchased product, a brand of the purchased product, a product price of the purchased product, a plurality of features related to the purchased product, and can include a time of purchasing the product. The product type can be classified by a major category, a middle category, and a minor category. For example, the first major category can include a first middle category related to a top, a middle category related to a bottom, a middle category related to a shoe, and a middle category related to an outerwear. For example, the second major category can include a middle category related to a handbag, a middle category related to a shoulder bag, a middle category related to a backpack, and a middle category related to a wallet. The third major category can include a middle category related to skin care, a middle category related to makeup, a middle category related to hair care, and a middle category related to perfume. For example, each middle category can be classified into a plurality of minor categories. However, the product type is not limited to the embodiment of the present disclosure, and can include more types of products and be classified into various forms. For example, a value representing each of a plurality of product types can be preset on the server. For example, the value representing each of the plurality of product types can include a product management code. The product management code can include a first code corresponding to the major category, a second code corresponding to the middle category, and a third code corresponding to the minor category. The plurality of features related to the purchased product can represent various features of the purchased product, for example, a pattern of the product, a material of the product, a color of the product, a composition of the product, and a smell of the product. For example, a value representing each of the plurality of features can be preset in the server. The product purchase time can include a date and time at which the user terminal purchases the product.

[0066] The inquiry history information is history information of a product searched by the user terminal, for example, a product type of the inquired product, a brand of the inquired product, a product price of the inquired product, a plurality of features related to the inquiry, etc. The product, and the product inquiry time of the inquired product can include an inquiry time. The product inquiry time can include a date and time at which the user terminal views the product and a time of staying on the product page.

[0067] The web log can be information in which the user terminal records a website activity. For example, the web log includes a web page on which the user terminal stays within the website, a time from when the user terminal first accesses the first web page to when the product is purchased, and a time from when the user terminal first accesses the first web page to when the product is purchased. It can include a second time taken by the user terminal to access the first web page, a total stay time of each of a plurality of web pages, and a stay time of each scroll position. Each web page.

[0068] In step S120, the server can determine preference information of the user terminal based on information about the online activity history.

[0069] The preference information of the user terminal can be preference information of products purchased by the user terminal. For example, the preference information of the user terminal can include a preference price value for each type of product, a preference brand value for each type of product, a preference color value for each type of product, and a preference value for each type of product. Related to the preferred scent. For example, the value of the preferred price of each type of product, the value of the preferred brand of each type of product, and the value of the preferred color of each type of product are the values of the preferred price of each type of product. Middle class category, the preferred brand value of each type of product, and the preferred color value of each type of product. It can include values for brands and colors.

[0070] For example, the value of each of a plurality of brands, the value of each of a plurality of colors, and the value of each of a plurality of perfumes can be pre-stored in the server.

[0071] For example, the value related to the preferred scent can include a value of a scent related to a top note, a value of a scent related to a middle note, a value of a scent related to a base note, and a value of a scent related to a main note. There are. The top note refers to the faintest and shortest scent that is felt when the perfume is first sprayed. The middle note refers to the intermediate scent that appears after the top note disappears. The base note refers to the scent with the longest duration. The main scent refers to the scent that is mainly felt in the incense.

[0072] For example, based on the capacity of the purchase history information included in the online activity history information being greater than a preset capacity, the server can provide the preference information of the user terminal by employing a preference analysis model of a neural network. It can be decided from the purchase history information.

[0073] For example, the preset capacity can be determined by Equation 1 below.

[0074] [Equation 1]

[0075]

[0076] In Equation 1, C th is the preset capacity, N learn is the number of training data sets used to learn the neural network, N layer is the number of layers included in the neural network, N need is N def is the minimum number of training data sets required for training of the neural network, N def is the minimum number of layers required for training of the neural network, C ref may be a reference capacity.

[0077] For example, the reference capacity can be a value pre-stored on the server.

[0078] By doing so, if the capacity of the purchase history information is less than the preset capacity, the query history information that generally guarantees the amount of data can be used. In addition, instead of setting the preset capacity as a fixed value, the preset capacity is adjusted according to the data used for learning of the neural network and the number of layers constituting the neural network, so that the neural network can be changed in the following manner: if there is additional learning or update, you can flexibly change the preset capacity.

[0079] For example, the server data pre-processes the purchase history information, acquires a product type value for each purchased product, a brand value for each purchased product, a product price value for each purchased product, a feature value related to each purchased product, and a purchase time value for each purchased product, you can create a purchase history vector containing.

[0080] The value of each purchased product type can be a value representing the product type of each of the plurality of products purchased after the user terminal joins the website. The brand value of each purchased product can be a value representing the brand of each of the plurality of products purchased after the user terminal joins the website. The value of each purchased product price can be a value representing the price of each of the plurality of products purchased after the user terminal joins the website. The feature value related to each purchased product can be a value representing the feature of each of the plurality of products purchased after the user terminal joins the website. The value of the purchase time of each purchased product can be a value representing the purchase time of each of the plurality of products purchased after the user terminal joins the website.

[0081] For example, the value of the product type can be a product management code.

[0082] For example, when the type of the purchased product is the first major category or the second major category, the feature value related to the product can include a color value. For example, when the type of the purchased product is the third major category, the feature value related to the product can include a value related to the scent. The value related to the scent can include a value of the scent related to the top note, a value of the scent related to the middle note, a value of the scent related to the base note, and a value of the scent related to the main note. For example, if the type of the purchased product is the third major category and is a cosmetic other than a perfume, the value of the perfume related to the top note, the value of the perfume related to the middle note, and the value of the scent related to the base note are determined to be 0 and can include a hit value.

[0083] For example, the value of the product purchase time can include a value of the date of purchase of the product and a value of the time of purchase of the product.

[0084] For example, the server can acquire the preference information of the user terminal output from the preference analysis model by inputting the purchase history vector into the preference analysis model.

[0085] For example, a preference analysis model can be learned based on a plurality of purchase history vectors and a plurality of correct answer preference information. The preference analysis model can use a CNN-based neural network. For example, the neural network can include an input layer, one or more hidden layers, and an output layer.

[0086] For example, learning data consisting of a plurality of purchase history vectors and a plurality of correct answer preference information is input to the input layer, passes through one or more hidden layers and the output layer, and is output as an output vector, the output vector is the input of the layer connected to the output layer, the loss function layer outputs a loss value using a loss function that compares the output vector of each training data with the correct answer vector, and the parameters of the neural network. Learning can be performed in the direction in which the loss value decreases.

[0087] One or more hidden layers can include one or more convolution layers and one or more pooling layers. For example, a plurality of purchase history vectors can be filtered in the convolution layer, and a feature map can be formed through the convolution layer.

[0088] For example, by reducing the dimension by selecting a fixed vector related to a feature based on the feature map formed in the pooling layer and subsampling the formed feature map, a feature related to preference information can be extracted from the vectorized time series data. For example, the pooling layer can be a max pooling layer that extracts the maximum value. For example, the pooling layer can be an average pooling layer that extracts the average value. For example, at this time, the parameters of the neural network can include parameters related to the convolution layer and the pooling layer (size of the feature map, size of the filter, depth, step, zero padding).

[0089] For example, one purchase history vector can constitute one preference information and one learning data set. At this time, a plurality of learning data sets can be pre-stored.

[0090] The correct answer preference information can include a value of a preferred price for each type of product, a value of a preferred brand for each type of product, a value of a preferred color for each type of product, and a value related to a preferred smell.

[0091] For example, the value of the preferred price included in the correct answer preference information can be determined for each intermediate classification category. At this time, by dividing the total number of purchases by the number of purchases of each product in the intermediate category and multiplying the price of that intermediate category, the preferred price of each product included in the correct answer preference information can be determined. It can be a value.

[0092] In addition, for example, the value of the preferred price included in the correct answer preference information can be determined by Equation 2 below.

[0093] [Equation 2]

[0094]

[0095] In Equation 2, Price p is the value of the preferred price of the middle category, k is the number of purchase products included in the middle category, p i is the price of the i-th purchase product, n i is the number of purchase products of the i-th purchase product, n total can be the total number of times the purchase product is purchased in the corresponding middle category.

[0096] For example, the value of the preferred brand of each type of product included in the correct answer preference information can be determined for each middle category. At this time, the value of the preferred brand included in the correct answer preference information is the value representing the first brand having the most number of purchases in the middle category and the value of the first brand consisting of the number of purchases of the products of the first brand. The value representing the second brand having the most number of purchases after the first brand, the value of the second brand consisting of the number of purchases of the products of the second brand, and the value representing the third brand can include the value of the third brand consisting of the value represented thereby and the number of purchases of the products of the third brand. For example, the value of the preferred brand included in the correct answer preference information can be determined for each middle category.

[0097] For example, the value of the preferred color for each type of product included in the correct answer preference information can be determined for each middle category. At this time, the value of the preferred color included in the correct answer preference information is the RGB value of the first color corresponding to the color having the most number of purchases in the middle category and the value of the first color consisting of the number of purchases. The product having the first color, the RGB value of the second color corresponding to the color having the most number of purchases after the first color, and the value of the second color consisting of the number of purchases of the product. The product having the second color, followed by the second color, can include the RGB value representing the third color corresponding to the color having the most number of purchases and the value of the third color consisting of the number of purchases. The product having the third color has been purchased. For example, the value of the preferred color included in the correct answer preference information can be determined for each middle category.

[0098] For example, the value related to the preferred scent included in the correct answer preference information can be determined based on the third major category related to cosmetics. For example, the value related to the preferred scent included in the answer preference information is the value of the scent related to the most overlapping top note and the most overlapping middle note among a plurality of purchase products corresponding to the third major category related to cosmetics. It can be determined as the value of the scent associated with the note, the value of the scent associated with the most overlapping base tone, and the value of the most overlapping tonic note.

[0099] In addition, for example, if there is no history of purchasing a product included in the third category, a value related to a preferred flavor included in the correct answer preference information can be set to a value 0.

[0100] Through this, the preference analysis model can be learned to determine the preference information based on the input purchase history vector.

[0101] In addition, for example, the preference analysis model can include a price preference analysis model, a brand preference analysis model, a color preference analysis model, and an odor preference analysis model. At this time, each analysis model can use the neural network described above. For example, the purchase history vector can be input into each analysis model, and the preference information includes a value of a preferred price of each product, a value of a preferred brand of each product, and a value of a preferred color of each product. For each analysis model, a value of each type of product and a value related to a preferred flavor can be obtained in parallel.

[0102] For example, based on the size of the purchase history information included in the online activity history information being less than a preset capacity, the server predicts a preference using a neural network from query history information included in the information. Through this model, the online activity history preference information of the user terminal can be determined.

[0103] For example, the server provides a value of whether to purchase for each viewed product, a value of a product type provided for each viewed product, a value of a brand provided for each viewed product, a value of a price provided for each viewed product by data preprocessing of the query history information. You can create a search history vector including a related feature value and a product search time value for each searched product.

[0104] The value of whether to purchase each viewed product can be a value indicating whether to purchase each of a plurality of products viewed after the user terminal joins the website. The value of each searched product type can be a value indicating the product type of each of a plurality of products viewed after the user terminal joins the website. The value of the brand of each viewed product can be a value representing the brand of each of a plurality of products viewed after the user terminal joins the website. The value of the price of each viewed product can be a value indicating the price of each of a plurality of products viewed after the user terminal joins the website. The value of each searched product related feature can be a value indicating the feature of each of a plurality of products viewed after the user terminal joins the website. The value of each viewed product search time can be a value representing the search time of each of a plurality of products viewed after the user terminal joins the website.

[0105] For example, for a product that has not been purchased, the value of whether to purchase can be determined as 0, and for a product that has been purchased, it can be determined as 1.

[0106] For example, the value of the product viewing time can include a value of a date on which the product is viewed, a value of a time at which the product is viewed, and a value of a time spent on a product page.

[0107] For example, the server can acquire the preference information of the user terminal output by the preference prediction model by inputting the query history vector into the preference prediction model.

[0108] The preference prediction model can be learned based on a plurality of query history vectors and a plurality of correct answer preference information. For example, the preference prediction model can use a neural network based on GRU (Gated Recurrent Unit). The preference prediction model is explained in detail in Figure 2 , which will be described later.

[0109] In step S130, the server can transmit an attraction message containing attraction information about the user terminal to the user terminal after a preset time elapses from when the first web page is accessed by the user terminal.

[0110] The attraction information can be information used to generate an attraction message transmitted to induce interest of the user terminal. For example, the lure information can include information about a final lure phrase, information about a preset time, and information about a preset position. Among them, the preset time can indicate a time at which the final lure phrase is displayed within a web page in which the user terminal is currently located. The preset position can indicate a position at which the final attraction phrase is displayed in the web page in which the user terminal is currently located.

[0111] The lure information can be displayed at the preset position of the web page in which the user terminal is located.

[0112] The lure message can be a message for linking to a second web page, which is a customized web page created to induce a purchase of the user terminal. For example, the lure message can link to a URL (Uniform Resource Locator) for accessing the second web page. The URL refers to a web address, which is a protocol for indicating a location of a resource on a network.

[0113] For example, the preset time can be a time from when the user terminal first accesses the first web page to when the product is purchased, and a time from when the first web page is accessed to when the access to the website is terminated. Based on the time required to arrive at this point for the second time.

[0114] For example, the preset position can be determined according to a total dwell time in which the user terminal dwells in each web page within the website and a dwell time for each scroll position of each web page.

[0115] According to an embodiment, the server can determine a product recommended for the user terminal based on the preference information of the user terminal.

[0116] For example, the server can determine a recommendation level for each of a plurality of preset products based on the preference information of the user terminal. Here, the plurality of preset products can be all products sold on a product sales website. The server determines a first similarity between a price of a product and a preferred price, a second similarity between a brand of the product and a preferred brand, and a third similarity between a color of the product for each of the plurality of preset products. And can determine a recommendation level according to a fourth similarity between a smell of the product and a preferred smell.

[0117] In addition, for example, the recommendation level can be determined by Equation 3 below.

[0118] [Equation 3]

[0119]

[0120] In Equation 3, R level is a recommendation level, Pr r is a preferred price corresponding to a product category, Pr0 is a price corresponding to a product, SIM1 is a preferred brand corresponding to a product category and a brand of the product category. SIM2 is a similarity between a preferred color corresponding to a product category and a color of the product, and SIM3 is a similarity between a value related to a preferred smell corresponding to a product category and a value related to a product. The smell of the product, K can be a correction coefficient.

[0121] For example, the recommendation level can be determined as a value greater than 0 and less than or equal to 100.

[0122] For example, SIM1 can be inversely proportional to a distance value between a coordinate value set for each of the first to third brands and a coordinate value set for a brand of the corresponding product. For example, a first distance between a coordinate value set for the first brand and a coordinate value set for the brand of the product, a second distance between a coordinate value set for the second brand and a coordinate value set for the brand of the product can be determined as a value divided by 2 for a third distance between a coordinate value set for the third brand and a coordinate value set for the brand of the product divided by a correction coefficient related thereto.

[0123] For example, SIM2 can be inversely proportional to a distance value between an RGB value set for each of the first to third colors and an RGB value set for a color of the corresponding product. For example, a first distance between an RGB value set for the first color and an RGB value set for the color of the product, and a second distance between an RGB value set for the second color and an RGB value set for the color. It can be determined as a value divided by 2 and a third distance between an RGB value set for the third color and an RGB value set for the color of the product divided by 4 divided by a correction coefficient related to the color.

[0124] For example, the SIM 3 can be determined in inverse proportion to a distance value between a coordinate value set for each of a scent related to a pre-scent, a scent related to a middle scent, a scent related to a base scent, and a scent related to a main scent, and a coordinate value set for a scent of a corresponding product. For example, the SIM 3 can be determined as a first distance between a coordinate value set for a scent related to a pre-scent and a coordinate value set for a scent of a product divided by 10, and a second distance between a coordinate value set for a scent related to a middle scent and a coordinate value set for a scent of a product can be determined as a sum of distance values between coordinate values set for a scent of a product divided by a correction coefficient related to taste.

[0125] For example, a plurality of brands, a plurality of colors, and a plurality of scents can be collected in advance.

[0126] For example, the server can set a coordinate of each of a plurality of brands on an xy coordinate axis, set as an x-axis indicating luxury and a y-axis indicating popularity according to a keyword of each of a plurality of brands. For example, as the number of keywords indicating a high-class feature in the total number of collected keywords increases, an x value can be determined as a large value. For example, the more the number of keywords indicating popularity in the total number of collected keywords, the greater y value can be determined.

[0127] For example, the server can preset an RGB value for each of a plurality of colors.

[0128] For example, the server can match each of a plurality of scents with a circular graph (perfume wheel) showing a relationship between olfactory groups inferred based on similarity and difference of scents. At this time, the circular graph can be classified according to four main accords: Fresh, Floral, Oriental, Woody; Floral, Soft Floral, Floral Oriental, Soft Oriental, Oriental, woody can include Oriental, Woody, Mossy Woody, Dry Woody, Fougere, Citrus, Aquatic, Green, and Fruity. At this time, the coordinate values can be set in order from Floral to Fruity. For example, the coordinate values can be set as values of 1 to 14 in order from Floral to Fruity.

[0129] For example, in the case where a product corresponds to the first or second major category, a scent set for the product can be determined according to a scent value of a perfume most frequently purchased by a plurality of terminals that purchase the product.

[0130] For example, the server can determine a product having a recommendation level higher than a preset recommendation level among a plurality of preset products as a product recommended to the user terminal.

[0131] According to one embodiment, the server can generate information of the final attraction phrase for the user terminal based on the basic information of the user terminal, the preference information of the user terminal, and a product recommended for the user terminal.

[0132] For example, the server can determine that the user terminal is in one of a plurality of groups based on a profile vector composed of a basic vector and a preference vector.

[0133] The basic vector can be generated by data preprocessing the basic information of the user terminal. For example, the basic vector can include values of age, gender, address, and body-related values. The value of the address can include a value of an area in which the user terminal is located. For example, a plurality of values of areas can be pre-stored on the server. The plurality of areas can include areas based on domestic administrative units. For example, the plurality of areas can include municipal areas. The body-related values can include values of height, weight, shoulder width, bust, and waist.

[0134] The preference vector can be generated by data preprocessing the preference information of the user terminal. For example, the preference vector can include a value related to a preferred price, a value related to a preferred brand, a value related to a preferred color, and a value related to a preferred scent.

[0135] The value related to the preferred price can be a value of a price preference type of the user terminal determined based on a preferred price value for each product type. For example, a plurality of price preference types can be pre-set on the server.

[0136] The value related to the preferred brand can be a value of a brand preference type of the user terminal determined based on a preferred brand value for each product type. For example, a plurality of brand preference types can be pre-set on the server.

[0137] The value related to the preferred color can be an RGB value indicating a most preferred color determined based on a preferred color for each type of product. For example, the value related to the preferred color can be an RGB value indicating a color having the most number of purchases or an RGB value indicating a color having the most number of views.

[0138] The value related to the preferred scent can include a value of a top note, a value of a middle note, a value of a base note, and a value of a main note.

[0139] For example, the plurality of groups can be classified by clustering based on the plurality of base vectors and the plurality of preference vectors. That is, the plurality of groups can be determined by a clustering technique based on the profile vectors including the base vectors and the preference vectors. Here, clustering can refer to unsupervised learning that groups data having similar attributes into a certain number of clusters. Specifically, the plurality of preference vectors can be reduced in dimension to three-dimensional vectors or less by various dimension reduction techniques. For example, the server can reduce the dimension of the plurality of profile vectors to three dimensions or less by various dimension reduction techniques. For example, the server can reduce the profile vectors to three dimensions or less by a principal component analysis (PCA) technique. For example, the server can determine a data axis having the greatest variance when projecting the plurality of profile vectors onto principal component axes, and reduce the dimension to the determined axis. For example, the server can generate a first axis based on the maximum variance in the plurality of profile vectors, and a second axis can generate a vector perpendicular to the first vector axis. Thereafter, the server can create a third axis as a vector perpendicular to the second axis again. When the server projects the original data onto the three vector axes created, the server can reduce the original data to a dimension equal to the number of vector axes. Hereinafter, the plurality of preference vectors are generated, and the vectors in which the generated user vectors are reduced in dimension by various dimension reduction techniques can be referred to as reduced dimension vectors. For example, the plurality of groups can be determined by a DBSCAN (density-based spatial clustering of applications with noise) technique based on the plurality of reduced dimension vectors. For example, DBSCAN assumes that if a certain element (point) belongs to a cluster, it must be close to many other elements in the cluster, and for this calculation, a radius and a minimum point can be used. For example, the diameter can be a radius based on a certain data element, which can be referred to as a dense region. For example, the minimum element can indicate how many elements are needed around a core point to designate it. Also, each element in the data set can be divided into a core point, a boundary point, and an outlier point. For example, the server can check the diameter size of each element and find how many elements are around it. Thereafter, if there are k or more elements within the diameter range, the server can determine that the element is a key element. In addition, the server can determine elements included within the diameter range from the core element as boundary elements. In addition, the server can determine elements not included within the diameter range from the core element as outlier elements, and can exclude the outlier elements from the corresponding cluster. Also, if the distance between the key elements is less than the diameter, the server can classify them into the same cluster.

[0140] For example, a price vector containing the value of the preferred price of each type of product can be generated for the plurality of user terminals. For example, the plurality of price preference types can be determined based on the plurality of price vectors by the clustering technique described above.

[0141] For example, a brand vector including a value of a preferred brand of each type of product can be created for a plurality of user terminals. For example, a plurality of brand preference types can be determined based on a plurality of brand vectors by the clustering technique described above. A plurality of brand preference groups can be determined by the clustering technique based on a plurality of brand vectors, and each brand preference group can represent one brand preference type. For example, a brand preference group most similar to a preferred price value of each product of a user terminal among a plurality of brand preference groups can be determined, and a value group for a corresponding brand preference can be determined as a brand preference type of the user terminal, which can be determined by the value.

[0142] For example, a value of a type of an attractive phrase can be set for each of a plurality of groups. The attractive phrase can refer to a phrase that induces a user terminal to click on an attractive message. For example, values of an inducing statement type are: type 1 stimulates curiosity, type 2 conveys a quantity limit, type 3 includes another person's experience, type 4 emphasizes a benefit, and type 4 can include a value. At least one of a fifth type and a sixth type indicates a time limit. In other words, the type of the incentive phrase can be a combination of a plurality of types.

[0143] For example, sample inducing phrases for each combination of the inducing phrase type can be transmitted as an inducing message to a plurality of user terminals included in each of a plurality of groups. At this time, a value of a phrase type corresponding to an attractive message on which a plurality of user terminals click the most can be set in a corresponding group.

[0144] For example, a final inducing phrase can be generated as a phrase that combines a sample inducing phrase matching a value of a type of an inducing phrase and a name of a product with the highest recommendation degree among products recommended to a user terminal. Here, the sample attractive phrase can be a sample attractive phrase on which a plurality of user terminals of a corresponding group click the most among a preset number of sample attractive phrases. For example, a plurality of sample inducing phrases matching a value of a phrase type can be transmitted as an inducing message to a plurality of user terminals included in each of a plurality of groups. At this time, a preset number of sample attractive phrases can be set in a corresponding group in order of a number of times a plurality of user terminals click on an attractive message. For example, the preset number can be three or more.

[0145] In step S140, the server can transmit information about a second web page to the user terminal based on a request of the user terminal to access a link to the attractive message.

[0146] For example, when the user terminal clicks on the decoy message, a connection request message requesting access to a URL for accessing the second web page can be transmitted to the server.

[0147] For example, the server can transmit information about a second web page including product information recommended for the user terminal to the user terminal based on the connection request message.

[0148] For example, the recommended product information can include a value of a product type of a recommended product.

[0149] For example, the server can determine a plurality of candidate templates of the web page by using a web page creation model of a neural network based on the basic information of the user terminal and the preference information of the user terminal. Here, the template can refer to information that defines a structure, design elements, and a text position of the web page in advance. For example, the template can include setting information of a header area of the web page, setting information of a navigation area of the web page, setting information of a content area of the web page, setting information of a content area of the web page, and the like. The setting information of a sidebar area of the web page and the navigation area of the web page can include setting information of a footer area. For example, the header area can include a logo area, a core menu area, a profile area, and a notification area. For example, the navigation area can include a home area, a menu item area, and a search area. For example, the content area can include a text area, an image area, and a video area. For example, the sidebar area can include a menu area and an additional search area. For example, the footer area can include an additional information area and a contact area. For example, the setting information includes information related to a text including arrangement of the text, size of the text, font of the text, thickness of the text, and the like, and arrangement of an image, position of the image, size of the image, and the like. It can include color-related information such as image-related information, color by area, brightness by area, and saturation by area.

[0150] Each of the plurality of candidate templates can be matched with each of a plurality of preset time intervals. For example, the plurality of preset time intervals can be configured in units of 10 seconds, such as 0 to 10 seconds, 10 to 20 seconds, 20 to 30 seconds, and the like. For example, each of the plurality of candidate templates can be matched every 10 seconds. At this time, a candidate template having a larger value of text thickness and color brightness can be matched to the last time interval unit.

[0151] The final template can be selected from among the candidate templates including a time interval from a first time point at which the bait message is transmitted to a second time point at which the connection request message is transmitted, among the plurality of preset time intervals. The second web page can have the final template applied to the product information recommended for the user terminal. That is, the second web page can be a web page in which the final template includes one or more products recommended for the user terminal. That is, the second web page can be a web page in which the final template includes one or more products recommended for the user terminal.

[0152] For example, a basic vector including an age value, a gender value, an address value, and a body-related value can be generated by data preprocessing of basic information of the user terminal. A preference vector including a value related to a preferred price, a value related to a preferred brand, a value related to a preferred color, and a value related to a preferred smell can be generated by preprocessing preference information data of the user terminal.

[0153] The server can obtain a value of the output candidate template set by inputting the default vector and the preference vector into the web creation model. At this time, the candidate template set can include a plurality of candidate templates. That is, the value of the candidate template set can indicate a plurality of candidate templates. For example, a plurality of candidate template sets can be pre-stored on the server.

[0154] The web generation model can be learned based on a set of a plurality of basic vectors, a plurality of preference vectors, and a plurality of correct answer candidate template sets.

[0155] The web creation model can use a neural network based on CNN. For example, the neural network can include an input layer, one or more hidden layers, and an output layer.

[0156] For example, training data consisting of a set of a plurality of basic vectors, a plurality of preference vectors, and a plurality of correct answer candidate templates is input to the input layer, passes through one or more hidden layers and the output layer, and is output as an output vector, which is input to a loss function layer connected to the layer as an output vector, and the loss function layer outputs a loss value using a loss function that compares the output vector of each learning data with a correct answer vector. The parameters of the neural network are set in the direction in which the loss value becomes smaller.

[0157] The one or more hidden layers can include one or more convolution layers and one or more pooling layers. For example, in the convolution layer, the plurality of basic vectors and the plurality of preference vectors can be filtered, and a feature map can be formed through the convolution layer.

[0158] For example, by reducing the dimension according to the fixed vector related to the feature selected from the feature map formed in the pooling layer, and performing subsampling on the formed feature map, a feature related to the candidate template set in the vectorized time series data can be extracted. For example, the pooling layer can be a max pooling layer that extracts a maximum value. For example, the pooling layer can be an average pooling layer that extracts an average value. For example, at this time, the parameters of the neural network can include parameters related to the convolution layer and the pooling layer (size of the feature map, size of the filter, depth, step, zero padding).

[0159] For example, one basic vector and one preference vector can constitute a set of correct candidate templates and a set of training data. At this time, a plurality of learning data sets can be pre-stored.

[0160] The correct answer candidate template set can be determined as a candidate template set that is most frequently selected by the plurality of user terminals in the group that matches the basic vector and the preference vector constituted by the learning data set. For example, the server can transmit sample images of each of the plurality of candidate template sets of each of the plurality of groups to the plurality of user terminals included in the group. The sample images can include images of each of the plurality of candidate templates of the web page. At this time, each of the plurality of user terminals in the group can select a sample image that the user terminal likes and transmit the same to the server. The correct answer candidate template set can be determined as a candidate template set to which the sample images most frequently selected by the plurality of user terminals in the group correspond.

[0161] For example, the plurality of candidate template sets can be updated at a predetermined time interval.

[0162] By doing so, the web page generation model can be trained to determine a set of candidate templates based on the input pool and the preference vector.

[0163] In step S150, the server can transmit information on at least one matching product to the user terminal based on the user terminal viewing a product page of the recommended product included in the second web page.

[0164] The information on at least one matching product can be a product that matches the recommended product included in the second web page. For example, at least one matching product can be determined by using a product matching model of a neural network based on the basic information of the user terminal, the preference information of the user terminal, and the recommended product information for the user terminal.

[0165] For example, a display position of an image of at least one matching product on a product page of a recommended product can be determined according to a matching relationship between the recommended product and the at least one matching product. Here, the product page of the recommended product can be a web page displayed on the user terminal for purchasing the recommended product when the user terminal searches for the recommended product included in the second web page.

[0166] Specifically, for example, the server can determine a group to which the user terminal belongs among a plurality of groups according to a profile vector constituted by the basic vector and the preference vector of the user terminal. The server can acquire purchase history information of the group to which the user terminal belongs. The purchase history information of the group to which the user terminal belongs can include purchase history information of each of all user terminals included in the group.

[0167] For example, the server pre-processes data on purchase history information to determine a product type value of each purchased product of a user terminal that purchased a recommended product within the recommended product, a purchase date of each purchased product, and a number of purchases of each purchased product. You can create a purchased product vector containing a frequency count. At this time, the purchased product can be a purchased product that is different from the recommended product in the intermediate category.

[0168] For example, the server can generate a recommended product vector including a value of at least one product type by data pre-processing on recommended product information of a user terminal.

[0169] For example, when the server inputs the purchased product vector and the recommended product vector to a product matching model using a neural network, a matching product vector including a value of at least one product type can be output. The server can determine at least one matching product as a product corresponding to the matching product vector.

[0170] For example, the product matching model can be learned based on a plurality of purchased product vectors, a plurality of recommended product vectors, and a plurality of correct answer matching product vectors.

[0171] The product matching model can use a neural network based on CNN. For example, the neural network can include an input layer, one or more hidden layers, and an output layer.

[0172] For example, learning data consisting of a plurality of purchased product vectors, a plurality of recommended product vectors, and a plurality of correct answer matching product vectors is input to the input layer, passes through one or more hidden layers and the output layer, and is output as an output vector. The loss function layer connected to the output layer is input, the loss function layer outputs a loss value using a loss function that compares the output vector of each learning data with the correct answer vector, and the neural network parameters used in the product matching model can be learned in a direction to reduce the loss value.

[0173] One or more hidden layers can include one or more convolution layers and one or more pooling layers. For example, a plurality of purchased product vectors and a plurality of recommended product vectors can be filtered in the convolution layer, and a feature map can be formed through the convolution layer.

[0174] For example, by reducing the dimension by selecting a fixed vector related to a feature from the feature map formed in the pooling layer, and performing subsampling on the formed feature map, a feature related to the matching product vector in the vectorized time series data can be extracted. For example, the pooling layer can be a max pooling layer that extracts a maximum value. For example, the pooling layer can be an average pooling layer that extracts an average value. For example, at this time, the parameters of the neural network can include parameters related to the convolution layer and the pooling layer (size of the feature map, size of the filter, depth, step, zero padding).

[0175] For example, one purchase product vector and one recommended product vector can constitute a matching product vector and a learning data set. At this time, a plurality of learning data sets can be stored in advance.

[0176] The correct answer matching product vector can include a value of at least one product type. For example, the value of at least one product type included in the correct answer matching product vector can include a value of a product type included in one purchase product vector consisting of a learning data set. In other words, a product corresponding to the correct answer matching product vector can be determined from a plurality of purchased products. For example, the product corresponding to the correct answer matching product vector can be determined as a purchase product in which the purchase frequency is greater than or equal to a predetermined frequency within a predetermined time among a plurality of purchase products consisting of a learning data set. At this time, the purchase frequency can be determined by applying a weight related to the time of purchase of the product to the number of user terminals that have purchased the product in the group. Here, the weight related to the purchase time of the corresponding purchase product can be set to a value greater than 0 and less than or equal to 1, and the closer to the most recent date, the closer to 1.

[0177] By doing so, a product matching model can be learned to determine a matching product based on an input purchase product vector and a recommended product vector.

[0178] For example, the matching relationship between the recommended product and the at least one matching product can represent a relationship between the use location information of the recommended product and the use location information of the at least one matching product. Here, the use location information can be information indicating a location where a user uses or wears a product. For example, the use location information can include a value of at least one area of a plurality of predetermined areas indicating a location where a user uses or wears a product. For example, display location information based on a combination of product-specific use location information and use location information can be stored in advance in the server. The display location information according to the combination of use location information can indicate display location information of a matching product according to a combination of each area.

[0179] For example, the predetermined plurality of areas can include a first area for hair, a second area for a face, a third area for a chest, a fourth area for an abdomen, a fifth area for an arm, and a sixth area for an arm. A seventh area for a leg and an eighth area for a foot. For example, if the product in question is a top, the use location information can include values of the third area, the fourth area, and the fifth area. For example, if the product in question is a lotion, the use location information can include a value of the second area.

[0180] For example, the server creates an image for at least one matched product according to a display position of an image of a recommended product on a product page according to a relationship between the use location information of the recommended product and the use location information of the user. You can decide the display position of at least one matched product. For example, if the recommended product is a lotion and the matched products are a shampoo and a coat, the position where the image of the shampoo is displayed includes use location information including the value of the second area and the value of the first area. According to the matching relationship between the use location information (for example, expressed as [2, 1]), the display position of the first area can be determined according to the second area. The position where the image of the court is displayed is the matching relationship (for example, [2, (3, 4, 5, 7)]) between the use location information including the value of the second area and the use location information including the values of the third to fifth areas and the seventh area, and the display position of the area consisting of the third to fifth areas and the seventh area can be determined based on the second area. For example, the image of the shampoo can be displayed at the top of the lotion product page according to the position where the image of the lotion is displayed, and the image of the pants can be displayed at the bottom according to the display position of the image of the lotion. It can be displayed in the area.

[0181] According to one embodiment, the server can collect a plurality of images of at least one matched product based on the determined at least one matched product. For example, the server can collect a plurality of images of at least one matched product by web crawling.

[0182] For example, the server can generate a candidate image vector including a pixel value of an image of at least one product by data preprocessing each of a plurality of images.

[0183] For example, the server can generate a reference image vector including a pixel value of a recommended product by data preprocessing an image of a recommended product.

[0184] For example, the server can input a plurality of candidate image vectors and a reference image vector into an image decision model using a second neural network, and select a value having the greatest similarity related to the composition of the reference image vector among the compositions of the plurality of candidate image vectors and the reference image vector. The candidate image vector can be obtained as an output image vector.

[0185] For example, the server can determine that the output image vector is an image of at least one matched product.

[0186] For example, the image decision model can be learned based on a plurality of candidate image vectors, a plurality of reference image vectors, and a plurality of correct output image vectors.

[0187] For example, the neural network used in the image decision model can use a CNN-based neural network. For example, the neural network can include an input layer, one or more hidden layers, and an output layer.

[0188] For example, training data composed of a plurality of candidate image vectors, a plurality of reference image vectors, and a plurality of correct answer output image vectors is input to an input layer, passes through one or more hidden layers and an output layer, and an output is obtained as an output vector input to a loss function layer connected to the output layer, and the loss function layer outputs a loss value using a loss function that compares the output vector with a correct answer vector of each training data. The network parameters used in the image decision model can be learned in the direction of reducing the loss value.

[0189] One or more hidden layers can include one or more convolutional layers and one or more pooling layers. For example, the plurality of candidate image vectors and the plurality of reference image vectors can be filtered in the convolutional layer, and a feature map can be formed through the convolutional layer.

[0190] For example, by reducing the dimension by selecting a fixed vector related to the feature from the feature map formed in the pooling layer, and by performing subsampling on the formed feature map, a feature related to the output image vector in the vectorized time series data can be extracted. For example, the pooling layer can be a max pooling layer that extracts the maximum value. For example, the pooling layer can be an average pooling layer that extracts the average value. For example, at this time, the parameters of the neural network can include parameters related to the convolutional layer and the pooling layer (size of the feature map, size of the filter, depth, step, zero padding).

[0191] For example, one candidate image vector and one reference image vector can constitute one output image vector and one training data set. At this time, a plurality of learning data sets can be pre-stored.

[0192] The correct answer output image vector can be a candidate image vector having the highest similarity value related to the composition of the reference image vector among a plurality of candidate image vectors composed of one learning data set. For example, the similarity related to the composition between the candidate image vector and the reference image vector can be determined based on the size of the object on the 3D coordinates and the angle at which the object is placed on the 3D coordinates. For example, the 3D coordinates of the product can be determined by implementing the product in 3D based on a plurality of candidate image vectors and a reference image vector.

[0193] In addition, for example, the similarity related to the composition between the candidate image vector and the reference image vector can be determined by Equation 4 below.

[0194]

Equation 4

[0195]

[0196] In Equation 4, Similarity is a similarity related to a composition between a candidate image vector and a reference image vector, a is a first coefficient for adjusting a size weight, Size1 is a similarity of the reference image vector. Size2 is a size of the object in the candidate image vector, β is a second coefficient for adjusting an angle weight, Angle1 is an angle of the object in the reference image vector, Angle2 is a feature of the object in the angle candidate image vector, and γ can be a third coefficient for adjusting a weight between the size and the angle.

[0197] For example, the values of the first to third coefficients can be greater than 0 and less than 1.

[0198] Through this, it is possible to learn an image decision model to determine an output image vector based on the input of a plurality of candidate image vectors and a reference image vector.

[0199] Figure 2 FIG. is a diagram illustrating an example of a preference prediction model according to an embodiment. Figure 2 Embodiments of the can be combined with various embodiments of the present disclosure.

[0200] Referring to Figure 2 The neural network used in the preference prediction model can be a gated recurrent unit (GRU)-based neural network 200. Here, the GRU can be a modified model of an RNN (recurrent neural network).

[0201] For example, the neural network 200 used in the preference prediction model can include an input layer 210, one or more hidden layers 220, and an output layer 230. Learning data composed of a plurality of query history vectors and a plurality of correct answer preference information is input to the input layer 210, passes through the one or more hidden layers 220 and the output layer 230, and is output as an output vector. The output vector is output to a first loss function layer connected to the layer 230, which outputs a loss value using a loss function that compares the output vector of each learning data with a correct answer vector. The preference prediction model can learn parameters of the neural network in a direction to reduce the loss value.

[0202] One search history vector used as learning data can be composed of one correct answer preference information and a set. For example, a plurality of sets can be pre-stored on a server.

[0203] The correct answer preference information can include a value of a preferred price for each type of product, a value of a preferred brand for each type of product, a value of a preferred color for each type of product, and a value related to a preferred smell.

[0204] For example, the value of the preferred price included in the correct answer preference information can be determined for each intermediate classification category. At this time, the value of the preferred price of each product included in the correct answer preference information is the number of times each product in the intermediate category is viewed divided by the total number of views multiplied by the sum of the prices. The weight of the product and the value applied with the purchase weight. This can be a value determined for each reclassification category.

[0205] For example, the value of the preferred price included in the correct answer preference information can be determined by changing the product purchased to the equation of the product viewed in Equation 2 above. At this time, if it is determined that the product viewed has been purchased according to the purchase value, a purchase weight can be applied to the product viewed determined as purchased in the changed equation. The purchase weight can be a preset value. For example, the purchase weight can be set to 1.2 or 1.5.

[0206] For example, the value of the preferred brand of each type of product included in the correct answer preference information can be determined for each intermediate category. At this time, the value of the preferred brand included in the correct answer preference information is the value representing the first brand having the most number of purchases in the intermediate category and the value of the first brand consisting of the number of times the product is queried. The value of the second brand having the most number of views after the representative first brand, the value of the second brand representing the number of times the product of the second brand is viewed, and the value of the third brand having the most number of views after the second brand can include the value of the third brand consisting of the value represented thereby and the number of times the product of the third brand is viewed. For example, the value of the preferred brand included in the correct answer preference information can be determined for each intermediate classification category. At this time, when it is determined that the product viewed has been purchased according to the value of whether or not it is purchased, a purchase weight can be applied to the number of times the brand corresponding to the product viewed determined as purchased is purchased. For example, the purchase weight can be set to 1.2 or 1.5.

[0207] For example, the value of the preferred color for each type of product included in the correct answer preference information can be determined for each intermediate classification category. At this time, the value of the preferred color included in the correct answer preference information is the RGB value representing the first color corresponding to the color with the most number of views in the intermediate classification category, and the first color consisting of the color with the most number of views. The number of times the first color of the viewed product has been viewed; the value of the color; the RGB value representing the second color corresponding to the color with the most number of views after the second color; and the value of the second color consisting of the number of colors. Thereafter, the RGB value representing the third color corresponding to the color with a high number of views, and the value of the third color consisting of the number of views can be included. The number of times the third color of the viewed product has been viewed. For example, the value of the preferred color included in the correct answer preference information can be determined for each intermediate classification category. At this time, if it is determined that the viewed product has been purchased according to the purchase status value, a purchase weight can be applied to determine the number of times the viewed product has been viewed. For example, the purchase weight can be set to 1.2 or 1.5.

[0208] For example, the value of the preferred smell related to the preferred smell included in the correct answer preference information can be determined based on the third major category related to cosmetics. For example, the value of the preferred smell related to the preferred smell included in the answer preference information is the value of the smell related to the most overlapping top note and the most overlapping middle note among the plurality of search products corresponding to the third major category related to cosmetics. It can be determined as the value of the smell associated with the note, the value of the smell associated with the most overlapping base note, and the value of the most overlapping tonic note. At this time, when it is determined that the viewed product has been purchased according to the purchase value, a purchase weight can be applied to the number of overlapping scents corresponding to the viewed product determined to be purchased. For example, the purchase weight can be set to 1.2 or 1.5.

[0209] One or more hidden layers 220 include one or more GRU blocks, and one GRU block can include a reset gate and an update gate. Here, the reset gate and the update gate can include a sigmoid layer. For example, the sigmoid layer can be a layer whose activation function is a sigmoid function For example, the hidden state can be controlled by the reset gate and the update gate, and weights can exist for each gate and input.

[0210] The reset gate resets past information, and the weight r(t) derived through the previous hidden layer can be determined.

[0211] For example, a plurality of query history vectors and a plurality of correct answer preference information are input to an input layer, and a reset gate generates a current input value (xt) from the plurality of query history vectors and the plurality of correct answers. At the time of input, a dot product of a weight Wr of the current time and a hidden state (h(t-1)) of the previous time is generated from the plurality of search history vectors and the plurality of correct answer preference information, which is a dot product of a weight Ur of the last time, and finally the two values are added and input to a sigmoid function, and the result can be output as a value between 0 and 1. Through these values between 0 and 1, it can be determined how much the hidden state value of the previous point will be utilized.

[0212] The update gate decides the update rate of the past and present information, and z(t) can be determined as the amount of information of the current time.

[0213] For example, when the input value (xt) of the current time is input, a dot product with the weight Wz of the current time point, a dot product with the hidden state (h(t-1)) of the previous time point, and a dot product with the weight Uz of the previous time point are added and input to a sigmoid function, and the result can be output as a value between 0 and 1. And 1-z(t) can be multiplied by the information (h(t-1)) of the previous hidden layer.

[0214] In this way, z(t) can reflect how much current information will be used, and 1-z(t) can reflect how much past information will be used.

[0215] By multiplying the result of the reset gate, the information candidate group of the current time t can be determined. For example, when the input value (xt) of the current time is input, the first value is a dot product with the weight Wh of the current time, and the hidden state (h(t-1)) of the previous time can be multiplied by the value of the previous weight Uh and input to a tanh function. For example, tanh represents a nonlinear activation function (hyperbolic tangent function).

[0216] By combining the results of the update gate and the candidate group, the weight of the current time hidden layer can be determined. For example, the output value z(t) of the update gate is multiplied by the current hidden state (h(t)), the value 1-z(t) discarded by the update gate, and the hidden state (h(t-1)) of the previous time. The weight of the current time hidden layer can be determined by the sum of the values multiplied by t-1.

[0217] By this, the server can learn a preference prediction model to predict preference information by considering a query history vector.

[0218] Figure 3 A flowchart of a method according to one embodiment is shown, in which a server provides information about at least one matching product to a user terminal using a neural network. Figure 3 Embodiments of the present disclosure can be combined with various embodiments of the present disclosure.

[0219] Referring to Figure 3 In step S310, the server can acquire basic information of the user terminal and online activity history information of the user terminal according to the user terminal accessing the first web page.

[0220] The first web page can be a main web page of a website providing a product sales service. The basic information can include an identification card, age, gender, address, and body information. The information about the online activity history can include purchase history information, inquiry history information, and a web log.

[0221] For example, the server can determine whether a capacity of the purchase history information included in the information about the online activity history is greater than a preset capacity.

[0222] For example, if the capacity of the purchase history information included in the online activity history information is greater than the preset capacity, the server determines preference information of the user terminal by using a preference analysis model of a neural network based on the purchase history. The purchase history information can include a product type of a purchased product, a brand of the purchased product, a product price of the purchased product, a plurality of features related to the purchased product, and a product purchase time of the purchased product. The preference information can include a value of a preferred price of each type of product, a value of a preferred brand of each type of product, a value of a preferred color of each type of product, and a value related to a preferred smell.

[0223] For example, if the capacity of the purchase history information included in the online activity history information is less than the preset capacity, the server can determine the preference information of the user terminal by using a preference prediction model based on a neural network. The inquiry history information can include a product type of an inquired product, a brand of the inquired product, a product price of the inquired product, a plurality of features related to the inquired product, and a product inquiry time of the inquired product.

[0224] For example, the server can determine a recommendation level of each of a plurality of preset products based on the preference information of the user terminal.

[0225] For example, the server can determine a product, of which a recommendation level is higher than a preset recommendation level, among the plurality of preset products as a product to be recommended to the user terminal.

[0226] In addition, for example, the server can change the preset recommendation based on a communication state of the user terminal and a number of user terminals currently connected to the server. For example, the server can decrease a preset recommendation level based on the communication state of the user terminal being good and the number of user terminals currently connected to the server being less than a preset threshold number. Finally, the server can increase the number of products recommended to the user terminal and increase a load of the server allocated to the user terminal by increasing the number of recommended products included in a customized web page provided to the user terminal.

[0227] In step S320, the server can determine whether the user terminal maintains access to the website for a preset amount of time.

[0228] For example, if a preset time has elapsed since the user terminal accesses the first web page, the server can determine that the user terminal maintains access to the website for the preset time.

[0229] In step S330, if the user terminal maintains access to the website for a preset period of time, the server can transmit an attraction message containing information about the attraction of the user terminal to the user terminal.

[0230] For example, the induction information can include information about a final induction phrase, information about a preset time, and information about a preset position. Among them, the preset time can indicate a time at which the final attraction phrase is displayed within the web page in which the user terminal is currently located. The preset position can indicate a position at which the final attraction phrase is displayed in the web page in which the user terminal is currently located.

[0231] For example, whenever the user terminal accesses the first web page, the server determines a first time from when the user terminal accesses the first web page to when a product is purchased, and a time at which the user terminal accesses the website. Thereafter, a second time required to terminate the connection can be stored for each user terminal.

[0232] For example, the server can determine a shorter time between an average of a plurality of first times of the user terminal and an average of a plurality of second times of the user terminal as a preset required time.

[0233] For example, the attractive message can be displayed at a preset position of the web page in which the user terminal is located.

[0234] For example, the server can store, for the user terminal, a position at which a stay time is the longest among stay times of each scroll position of each web page in which the user terminal stays. For example, the preset position can be determined as a position at which a stay time is the longest among stay times of each scroll position of each web page. For example, the incentive message can be displayed at a position at which the user terminal stays the longest among the web pages in which the user terminal stays.

[0235] For example, the server can generate information of a final phrase that is most attractive to the user terminal based on basic information of the user terminal, preference information of the user terminal, and a product recommended for the user terminal.

[0236] In step S340, if the user terminal does not maintain access to the website for a preset period of time, the server can adjust information related to the preset period of time.

[0237] For example, if the user terminal terminates the access to the website within a preset time, the server reflects an average of a plurality of second numbers of times between when the user terminal accesses the first web page and when the access to the website is terminated. The user terminal can be adjusted for the previously stored.

[0238] In step S350, the server can determine whether a connection request message requesting to access a URL linked to the decoy message is received from the user terminal within a preset time from when the decoy message is transmitted.

[0239] In step S360, if the server does not receive the connection request message from the user terminal within the preset time from when the decoy message is transmitted, the server can change the final decoy phrase of the user terminal.

[0240] For example, the server selects a sample decoy phrase having the highest number of clicks after the sample decoy phrase is transmitted to the user terminal among a preset number of sample decoy phrases and a product name having the highest recommendation degree in the recommended product. The user terminal can change the final decoy phrase and the combined phrase. Among a plurality of sample decoy phrases matching the value of the phrase type, the preset number of sample decoy phrases can be arranged in order of the number of clicks of the decoy message by a plurality of user terminals.

[0241] In addition, for example, after the preset time elapses and a delay weight has been applied to the preset time, the server can transmit a decoy message containing information about the changed final decoy phrase to the user terminal. For example, the delay weight can have a value of 0.8 or more and 1.2 or less.

[0242] At this time, the delay weight can be inversely proportional to the difference between the number of clicks of the sample decoy phrase that has been transmitted and the number of clicks of the sample decoy phrase corresponding to the changed final decoy phrase. That is, the greater the difference in the number of clicks, the smaller the weight is determined, and thus the decoy message can be transmitted faster than the existing preset time.

[0243] In step S370, when the connection request message is received from the user terminal within the preset time, the server can transmit information about a second web page including product information recommended for the user terminal.

[0244] In step S380, the server can determine whether the user terminal searches for a product page of the recommended product included in the second web page.

[0245] For example, the server can determine whether the user terminal views the product page of the recommended product included in the second web page through a web log of the user terminal.

[0246] In step S390, when the user terminal views a product page of the recommended product included in the second web page, the server can transmit information about at least one matching product to the user terminal.

[0247] The information about at least one matching product can be a product matching the recommended product included in the second web page. For example, at least one matching product can be determined by using a product matching model of a neural network based on the basic information of the user terminal, the preference information of the user terminal, and the recommended product information for the user terminal.

[0248] For example, a display position of an image of at least one matching product on a product page of a recommended product can be determined according to a matching relationship between the recommended product and at least one matching product.

[0249] Figure 4 is a block diagram illustrating a configuration of a server according to an embodiment. Figure 4 An embodiment of the server can be combined with various embodiments of the disclosure.

[0250] As Figure 4 illustrated, the server 400 can include a processor 410, a communication unit 420, and a memory 430. However, not all components illustrated in Figure 4 are essential components of the server 400. The server 400 can be implemented with more components than those illustrated in Figure 4 , or the server 400 can be implemented with fewer components than those illustrated in Figure 4 . For example, the server 400 according to some embodiments can include a user input interface (not shown), an output unit (not shown), etc., in addition to the processor 410, the communication unit 420, and the memory 430.

[0251] The processor 410 generally controls the overall operation of the server 400. The processor 410 can include one or more processors and control other components included in the server 400. For example, the processor 410 can generally control the communication unit 420 and the memory 430 by executing programs stored in the memory 430. In addition, the processor 410 can perform the functions of the server 400 illustrated in Figures 1 to 3 by executing programs stored in the memory 430.

[0252] The communication unit 420 can include one or more components that allow the server 400 to communicate with other devices (not shown) and servers (not shown). The other devices (not shown) can be a computing device such as the server 400 or a sensing device, but are not limited thereto. The communication unit 420 can receive a user input from another electronic device or receive data stored in an external device through a network from the external device.

[0253] For example, the communication unit 420 can transmit and receive a message to establish a connection with at least one device. The communication unit 420 can transmit information generated by the processor 410 to at least one device connected to a server. The communication unit 420 can receive information from at least one device connected to a server. The communication unit 420 can transmit information related to the received information in response to the information received from at least one device.

[0254] The memory 430 can store a program for processing and control of the processor 410. For example, the memory 430 can store information input to a server or information received from another device through a network. In addition, the memory 430 can store data generated by the processor 410. The memory 430 can store information input to or output from the server 400.

[0255] The memory 430 is a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., an SD or XD memory, etc.), and a RAM (RAM, Random Access Memory), SRAM (Static), Random Access Memory), ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), PROM (Programmable Read Only Memory), magnetic memory, magnetic disk, and can include at least one type of storage medium optical disk therein.

[0256] The above-described embodiments can be implemented using hardware components, software components, and / or combinations of hardware components and software components. For example, the apparatuses, methods, and components described in the embodiments can include processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, and field programmable gate arrays (FPGAs), for example. It can be implemented using one or more general-purpose or special-purpose computers, such as arrays, programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. The processing device can execute an operating system (OS) and one or more software applications running on the operating system. In addition, the processing device can access, store, manipulate, process, and generate data in response to the execution of software. For ease of understanding, it can be described as using a single processing device; however, those skilled in the art will understand that the processing device can include multiple processing elements and / or multiple types of processing elements. For example, the processing device can include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0257] The method according to the embodiments can be implemented in the form of program instructions capable of being executed through various computer means and recorded on a computer-readable medium. The computer-readable medium can include program instructions, data files, data structures, etc., individually or in combination. The program instructions recorded on the medium can be specifically designed and configured for the embodiments, or can be known and available to those skilled in the computer software field. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic-optical media such as floptical disks including optical media (magneto-optical media). Hardware devices specifically designed to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include machine language codes such as codes generated by compilers, and high-level language codes that can be executed by computers using interpreters, etc. The above hardware devices can be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Claims

1. A method of providing information about at least one matching product to a user terminal using a neural network at a server, wherein, comprising the steps of: acquiring basic information of the user terminal and online activity history information of the user terminal according to the user terminal accessing the first webpage; determining preference information of the user terminal according to the online activity history information of the user terminal; sending an attraction message containing attraction information of the user terminal to the user terminal after a preset time from the user terminal accessing the first webpage; displaying the temptation information at a preset position of the webpage where the user terminal is located, sending information about the second webpage to the user terminal based on the information that the user terminal requests to access the second webpage linked with the bait message; and the second webpage includes product information recommended for the user terminal, sending information of at least one matching product to the user terminal based on the user terminal viewing the product page of the recommended product contained in the second webpage, determining the product recommended for the user terminal according to the preference information of the user terminal, generating attraction information of the user terminal based on the basic information of the user terminal, the preference information of the user terminal, and the recommended product information of the user terminal, the at least one matching product is a product matched with the recommended product, the at least one matching product is determined based on the basic information of the user terminal, the preference information of the user terminal, and the recommended product information of the user terminal by using a product matching model of the first neural network, determining the display position of the picture of the at least one matching product on the product page according to the matching relationship between the recommended product and the at least one matching product.

Images