KR20240012220A - Method for recommending user-personalized plants based on artificial intelligence and plant management system performing the same - Google Patents

Abstract

Embodiments include receiving the user's account information and generating the user's account registration data; Generating user space registration data by analyzing space-related information of the user; Analyzing the user's plant-related information to generate the user's plant registration data; generating profile data for the user based on the user's account registration data, the user's space registration data, and the user's plant registration data; and calculating a first type of customized plant for the user based on the user's profile data and the profile data of the plant species. It relates to an artificial intelligence-based user customized plant recommendation method and a plant management system that performs the same.

Description

Method for recommending user-personalized plants based on artificial intelligence and plant management system performing the same}

Embodiments of the present application are a method of recommending a first type or a second type of customized plant for the user by analyzing profile data or history data including the user's individual characteristics based on artificial intelligence technology, and performing the same. It's about doing.

As the quality of life improves, users' needs for living environments are increasing in multiple dimensions. For example, there is a significant need to improve the user's living environment by placing flowers or other growing plants in the living room.

Growing plants in the living room not only beautifies the living environment, but can also improve the air quality in the living room.

However, each growing plant has a relatively sensitive growth habit. Moreover, the growth habits of each plant vary. Due to a lack of expert knowledge about these plants, there is a high probability that users will incorrectly select plants that are unsuitable for their living environment, which then makes it difficult to grow plants appropriately according to the growth habits of the plants in their living environment. If the user does not grow plants according to the plant's growth habits, the growing plants will die, ultimately causing a loss of economic resources due to disposal and repurchase of dead remains, and reducing the user's interest in growing plants.

Recently, there have been attempts to promote customized plants for users by taking into account the user's living environment, but the user's usage environment has not yet been analyzed in a high level, so there are limitations in recommending truly customized plants to the user.

Patent Registration Publication No. 10-2312180 (2021.10.06.)

In order to solve the above-described problem, embodiments of the present application provide a method of recommending customized plants for the user by analyzing profile data or history data including the user's individual characteristics based on artificial intelligence technology, and plant management that performs this. We would like to provide a system.

The artificial intelligence-based user-customized plant recommendation method according to one aspect of the present application may be performed by one or more computing devices. The artificial intelligence-based user-customized plant recommendation method includes receiving user account information and generating user account registration data; Generating user space registration data by analyzing space-related information of the user; Analyzing the user's plant-related information to generate the user's plant registration data; generating profile data for the user based on the user's account registration data, the user's space registration data, and the user's plant registration data; and calculating a first type of customized plant for the user based on the user's profile data and the plant species' profile data.

In one embodiment, the step of analyzing the user's space-related information to generate the user's space registration data includes receiving space-related information including a user's space-captured image from the user's electronic device; Recognizing objects in the spatial captured image and calculating object characteristics for each recognized object; Obtaining space type characteristics from the space-related information; calculating style characteristics of space based on object characteristics of the recognized object; Calculating ventilation characteristics of the space based on the number of windows; Receiving the illuminance value of the space measured from the user's electronic device; Obtaining the received illuminance measurement value as illuminance characteristics; And it may include generating space registration data of the user including one or more space characteristics of the user's space identifier, object characteristics placed in the space of the identifier, space type characteristics, style characteristics of the space, and illuminance characteristics. The user's spatial identifier is assigned in the step of creating the user's account registration data.

In one embodiment, the step of calculating the style characteristics may be recognizing the style of the space in which the object is recognized using a pre-learned style recognition model. The style recognition model is a machine learning model configured to predict the style characteristics of the space in which the input object is recognized by inferring the correlation between the input object characteristics of the space where the style is to be recognized and the style of the input space. The style recognition model is learned using a training data set consisting of a plurality of training samples, and each training sample has training data and label data, and the training data is a sample object calculated from a sample space image taken of the sample space. Includes characteristics. The input object characteristics and sample object characteristics include one or more object information among the location of the object, the arrangement relationship of the objects, and the appearance of the object.

In one embodiment, the ventilation characteristic indicates a ventilation level with a higher value as ventilation becomes easier. The step of calculating the ventilation characteristics of the space based on the number of windows may be calculating the ventilation level of the space based on at least two space characteristics of the ventilation space type, number of windows, and ventilation area of the windows. there is.

In one embodiment, the step of calculating a first type of customized plant for the user based on the user's profile data and the profile data of the plant species includes selecting one or more plants matching a plant search filtering condition from the user's profile data. searching for a first type of customized plant for the user; calculating a user's suitability score for the searched first type of customized plants, and generating a list of first types of customized plants based on the suitability scores; and transmitting the first type of customized plant list to the user's electronic device. The filtering condition includes at least one characteristic item among characteristic items included in the user's profile data. The user's suitability score for the searched plant is calculated based on the number of characteristics matching between the profile data of the plant species and the user's profile data.

In one embodiment, the step of generating a first type of customized plant list based on the suitability score includes, when the user's profile data includes a plurality of spatial registration data, searched based on spatial characteristics in each spatial registration data. This may be to classify plants by each registered space and create a customized plant list for each registered space.

In one embodiment, the artificial intelligence-based user customized plant recommendation method receives the user's evaluation score for the first type of customized plant calculated and recommended for the user, and the recommended first type of customized plant and evaluation score Generating evaluation history data including; And based on the evaluation history of a plurality of users for whom one or more first types of customized plants are recommended for each user, a plurality of factors constituting the correlation between the first type of customized plants and the evaluation of the recommendation are analyzed to determine each first type. The method may further include calculating each user's recommendation score for each type of customized plant, and calculating a second type of customized plant based on the user's recommendation score for each first type of customized plant.

In one embodiment, the artificial intelligence-based user-customized plant recommendation method includes at least one of plant content production history data and plant content consumption history data when one or more of the user's plant content production event and plant content consumption event occurs. Generating activity history data; And it may further include calculating a second type of customized plant for the user based on the user's activity history data. The step of calculating a second type of customized plant for the user based on the user's activity history data includes extracting a keyword of interest of the user from plant content production history data and consumption history data; calculating a user's interest score based on the extracted interest keywords; generating user interest record data including the interest score by associating corresponding activity history data and interest keywords; forming a plurality of clusters consisting of some users with similar interest tendencies among the plurality of users based on interest record data of each of the plurality of users; and calculating a second type of customized plant for the user based on keywords of interest related to the community to which the user belongs.

In one embodiment, the step of calculating a second type of customized plant for the user based on the interest keyword related to the cluster to which the user belongs includes selecting the user's interest among a plurality of interest keywords related to the cluster to which the user belongs. A second type of customized plant for the user may be calculated based only on the remaining interest keywords not included in the recorded data.

In one embodiment, calculating the user's interest score based on the extracted interest keywords includes assigning an interest score to the interest keyword extracted from consumption history data using a pre-stored interest score table; And based on the local frequency of the extracted keyword of interest appearing in the content produced by the user from which the keyword of interest was extracted and the global frequency of the appearance of the extracted keyword of interest in the entire produced content of a plurality of users, It may also include the step of assigning the user's interest score and assigning the calculated interest score to the corresponding produced content. The global frequency may be the frequency of appearance in the entire range of individually produced content of all users over the entire period on the system or over a certain period of time.

In one embodiment, the artificial intelligence-based user-customized plant recommendation method includes: generating purchase history data for the user when a plant purchase event occurs; And it may further include calculating a second type of customized plant for the user from the user's purchase history data using a pre-trained purchase prediction model. The purchase prediction model is a machine learning model configured to infer the correlation between a plurality of characteristics included in user-related data and the plants the user purchases, and to custom-predict plants to be purchased by the user of an input data set containing a plurality of input characteristics. . The purchase prediction model is learned using a training data set consisting of a plurality of training samples, and each training sample in the set includes training data and label data for each sample user. The label data has a value indicating the purchased plant for a user whose purchase history includes purchased plants. In the case of a user without purchase history, the label data has a value representing the user's experienced plants included in the user's registered plant data in the user's profile data. The training data may be formed based on at least one user's data among profile data of each user, activity history data of the user, and purchase history data of the user.

A computer-readable recording medium according to another aspect of the present application may record a program consisting of instructions for performing the artificial intelligence-based user-customized plant recommendation method according to the above-described embodiments.

The plant management system according to another aspect of the present application is to input the user's account information to subscribe to the plant management system, and the plant management system provides plant-related services in consideration of the user's space and the plants grown or being grown. A user's electronic device configured to input the user's space-related information and plant-related information, and to receive customized plant information for the user calculated using the user's profile data and the plant's profile data; And receive the user's account information to generate the user's account registration data, analyze the user's space-related information to generate the user's space registration data, and analyze the user's plant-related information to generate the user's plant registration data. , generating profile data for the user based on the user's account registration data, the user's space registration data, and the user's plant registration data, and generating a first profile data for the user based on the user's profile data and the plant species' profile data. It may also include a service server configured to produce a type of customized plant.

In one embodiment, the electronic device may be further configured to input the user's evaluation score for the received first type of customized plant and transmit it to the service server. The service server receives the user's evaluation score for the first type of customized plant calculated and recommended for the user and generates evaluation history data including the recommended first type of customized plant and the evaluation score, and Based on the evaluation history of a plurality of users for whom at least one type of first type of customized plant is recommended for each user, a plurality of factors constituting the correlation between the evaluation of the first type of customized plant and the recommendation are analyzed for each first type. It may be further configured to calculate a recommendation score of each user for each of the customized plants, and calculate a second type of customized plant based on the user's recommendation score for each first type of customized plant.

In one embodiment, the electronic device receives a user input for consuming plant content or producing plant content in the system, transmits a request for executing a service operation according to the user input to a service server, and responds to the request. It may be further configured to receive processing results for service operations. When receiving the request, the service server generates activity history data including one or more of plant content production history data and plant content consumption history data, and creates a second type for the user based on the user's activity history data. It may be further configured to yield customized plants.

In one embodiment, the electronic device receives a user input for purchasing a plant through a commerce system included in or connected to the system, transmits a purchase request according to the user input to the commerce system, and sends a request to the commerce system. It may be further configured to receive a purchase processing result according to. The service server, upon receiving the purchase request, generates purchase history data of the corresponding user; and may be further configured to calculate a second type of customized plant for the user from the user's purchase history data using a pre-learned purchase prediction model.

In the plant management system according to the above-described embodiments, the service server includes a second type of customized plant calculated based on evaluation history data, a second type of customized plant calculated based on activity history data, and purchase history data. When the second type of customized plant calculated based on is calculated, plants including some or all of each second type of customized plant may be calculated as a final customized plant and transmitted to the electronic device. The final customized plant calculation result is a recommendation result screen composed of a multi-layer structure, and the recommendation result screen includes multiple sub-screens including recommendation results for each recommendation method. The electronic device may be further configured to display the recommendation result screen and move a sub-screen by scrolling on the recommendation result screen.

The plant management system according to one aspect of the present invention may predict and recommend a first type of customized plant for the user by analyzing profile data including user characteristics and plant species characteristics based on artificial intelligence technology.

In addition, the plant management system analyzes the evaluation history of the recommendation of the first type of customized plant, activity details such as content consumption/production within the management system platform, and purchase history on the commerce system based on artificial intelligence technology for the corresponding user. It is also possible to predict and recommend a second type of customized plant for .

The plant management system calculates plants tailored to the user in various aspects, enabling more accurate recommendations. As a result, even if users lack expert knowledge about plants, the probability of incorrectly selecting plants that are unsuitable for their living environment is reduced. As a result, leakage of plant consumption amounts can be prevented.

Additionally, the plant management system can form a list of various plant products and suggest them to users who cannot enter keywords for new plants due to lack of plant background knowledge. As a result, the user's plant knowledge base expands.

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

In order to more clearly explain the technical solutions of the embodiments of the present invention or the prior art, drawings necessary in the description of the embodiments are briefly introduced below. It should be understood that the drawings below are for illustrative purposes only and not for limiting purposes of the present specification. Additionally, for clarity of explanation, some elements may be shown in the drawings below with various modifications, such as exaggeration or omission.
1 is a block diagram of a plant management system according to an embodiment of the present application.
Figure 2 is a schematic diagram of service operations performed by the plant management system of Figure 1;
Figure 3 shows profile data of plant species stored in a plant species profile DB, according to an embodiment of the present application.
Figure 4 shows user profile data stored in a user profile DB according to an embodiment of the present application.
Figure 5 is a schematic diagram of a process for calculating a recommendation score according to an embodiment of the present application.
Figure 6 is a schematic diagram of a keyword-of-interest extraction operation according to an embodiment of the present application.
Figure 7 is a schematic diagram of a learning process of the purchase prediction model, according to an embodiment of the present application.
8A to 8C illustrate a recommendation result screen provided in a multi-layer structure according to an embodiment of the present application.
9A to 9D are flowcharts of an artificial intelligence-based user-customized plant recommendation method according to another aspect of the present application.
FIG. 10 is a schematic diagram of the process of generating spatial registration data of FIG. 9A.
Figure 11 is a schematic diagram of the process of generating plant registration data for the user of Figure 9A.
Figure 12 is a schematic diagram of a process for generating a first type of customized plant for the user of Figure 9A.
Figure 13 shows a screen for entering an evaluation score according to an embodiment of the present application.
Figure 14 is a schematic diagram of a process for generating user interest record data based on content consumption history data, according to an embodiment of the present application.
Figure 15 illustrates a user request corresponding to a consumption event, according to an embodiment of the present application.
Figure 16 is a schematic diagram of a process for generating user interest record data based on content production history data, according to an embodiment of the present application.

Hereinafter, embodiments of the present application will be examined in detail with reference to the drawings.

However, this disclosure is not intended to limit the disclosure to specific embodiments, and should be understood to include various modifications, equivalents, and/or alternatives to the embodiments of the disclosure. . In connection with the description of the drawings, similar reference numbers may be used for similar components.

In this specification, expressions such as “have,” “may have,” “includes,” or “may include” refer to the corresponding features (e.g., numerical values, functions, operations, steps, parts, elements, and/or components). It refers to the presence of components such as etc.) and does not exclude the presence or addition of additional features.

When a component is said to be "connected" or "connected" to another component, it is understood that it may be directly connected to or connected to the other component, but that other components may exist in between. It should be. On the other hand, when it is mentioned that a component is “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

Expressions such as “first,” “second,” “first,” or “second” used in various embodiments may modify various elements regardless of order and/or importance, and limit the elements. I never do that. The above expressions can be used to distinguish one component from another. For example, the first component and the second component may represent different components, regardless of order or importance.

Embodiments of singular expressions used in this specification also include embodiments of plural expressions, unless phrases related to the singular expression clearly indicate the contrary.

As used herein, the expression “configured to” may mean, for example, “suitable for,” “having the capacity to,” or “having the capacity to.” ," can be used interchangeably with "designed to," "adapted to," "made to," or "capable of." The term “configured (or set) to” may not necessarily mean “specifically designed to” in hardware. Instead, in some contexts, the expression “a device configured to” may mean that the device is “capable of” working with other devices or components. For example, the phrase "processor configured (or set) to perform A, B, and C" refers to a processor dedicated to performing the operations (e.g., an embedded processor), or executing one or more software programs stored on a memory device. By doing so, it may mean a general-purpose processor (eg, CPU or application processor) capable of performing the corresponding operations.

Figure 1 is a block diagram of a plant management system, according to an embodiment of the present application, and Figure 2 is a schematic diagram of service operations performed by the plant management system of Figure 1.

The plant management system 1 according to embodiments may be entirely hardware, entirely software, or have aspects that are partly hardware and partly software. For example, a system may collectively refer to hardware equipped with data processing capabilities and operating software for running it. In this specification, terms such as “unit,” “system,” and “device” are intended to refer to a combination of hardware and software driven by the hardware. For example, the hardware may be a data processing device that includes a Central Processing Unit (CPU), Graphics Processing Unit (GPU), or other processor. Additionally, software may refer to a running process, object, executable, thread of execution, program, etc.

1 and 2, the plant management system 1 is configured to recommend customized plants to each user by a service server 200 connected to one or more electronic devices 100 of the user. .

The electronic device 100 is a client terminal device that communicates with the service server 200 and includes at least one processor capable of processing data, a memory for storing data, and a communication unit for transmitting/receiving data. Such electronic devices 100 may be implemented in the form of, for example, laptop computers, other computing devices, tablets, cellular phones, smart phones, smart watches, smart glasses, head-mounted displays (HMDs), other mobile devices, and other wearable devices. It could be.

In certain embodiments, the electronic device 100 includes a memory 110, a communication module 120, a processor 130, an input device 140, an output device 150, and a camera ( 160) and an illumination sensor 170.

The memory 110 is connected to the processor 130 and may store data such as basic programs for operation of the processor 130, application programs, setting information, and information generated by operations of the processor 130. Additionally, the memory 110 may provide stored data according to a request from the processor 130. Additionally, the memory 110 may permanently or temporarily store data received from the communication unit 120, data captured by the camera 160, and data generated or processed by the processor 130.

The memory 110 may be comprised of volatile memory, non-volatile memory, or a combination of volatile memory and non-volatile memory. The memory 110 includes HDD (Hard Disk Drive), SSD (Solid State Drive), CD (Compact Disc), RAM (Random Access Memory), Rom (Read Only Memory), and various types of data that store data permanently, semi-permanently, or temporarily. It may also include other storage devices.

In certain embodiments, the memory 110 may store a program (or application) for implementing the operation of a plant management system, such as an artificial intelligence-based user-customized plant recommendation method.

The communication module 120 may be referred to as a transmitter, receiver, transceiver, communication unit, communication modem, or communication circuit. The communication module 120 is a wired access system and a wireless access system, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.xx system, IEEE Wi-Fi system, 3rd generation partnership project (3GPP) system, and 3GPP LTE (long term evolution). It can support at least one of various wireless communication standards such as system, 3GPP 5G NR (new radio) system, 3GPP2 system, and Bluetooth.

The processor 130 may be configured to implement the procedures and/or methods proposed in the present invention. The processor 130 may determine at least one executable operation of the electronic device 100 based on information determined or generated using a data analysis algorithm. Additionally, the processor 130 may control the components of the electronic device 100 to perform the determined operation. This processor 130 may be implemented with at least one processor. The processor 130 may include a CPU (Central processing unit) or a neuromorphic processor designed to be advantageous for artificial neural network calculations by incorporating neurons and synapses of the human brain.

In certain embodiments, the processor 130 provides raw data to be used by the service server 200 to generate the user's profile data, history data, and executes the data received from the service server 200. ) controls the overall operations.

The input device 140 is a component configured to receive commands related to user input. The input device 140 may include a touch unit or other input unit.

The touch unit is a component that uses a part of the user's body or another object as a pointing object to input user commands. The touch unit may include, but is not limited to, a pressure-sensitive or electrostatic sensor. The other input units include, for example, buttons, keyboards, dials, switches, sticks, etc.

The output device 150 is a component that displays information stored and/or processed in the electronic device 100 and may include, for example, an LCD, OLED, or flexible screen, but is not limited thereto.

Although the input device 140 and the output device 150 are separated in FIG. 1, in many embodiments, the input device 140 and the output device 150 are implemented as a single component to receive input and output information. It can be. For example, the input device 140 and the output device 150 may be a touch panel implemented as a touch screen forming a layer structure with a screen. Touch input is input by a pointing object (eg, including the user's body, a tool, etc.).

The camera 160 captures a subject and obtains an image frame such as a still image or video. The acquired image frame may be displayed on the display of the output device 150, transmitted to an external device through the communication unit 120, processed by the processor 130, or stored in the memory 110.

The camera 160 may be composed of one or more photographing units. The imaging unit may generate an image of the subject in response to wavelengths. The camera 160 may acquire RGB images by including an RGB sensor. Additionally, the camera 160 may further include an IR sensor or a depth sensor to further acquire an IR image or depth image.

In certain embodiments, the camera 160 may capture part or all of the user's space where the user can place growing plants. Additionally, the camera 160 may capture part or all of the user's growing plants.

The sensor 170 is configured to convert physical changes around the electronic device 100 into electrical signals. In certain embodiments, the sensor 170 includes an illuminance sensor that detects changes in illuminance of a space. The illuminance sensor 170 is a sensor that measures surrounding brightness. The illuminance sensor 170 includes a photoelectric element that has a photoelectric effect in which conductivity changes by generating electrons excited to the internal structure when optical energy (eg, light) is acquired. The photoelectric device may be, for example, a cadmium sulfide (Cds)-based device, but is not limited thereto.

Additionally, in some embodiments, the sensor 170 may further include one or more sensors that detect changes in the spatial environment in addition to illuminance.

The electronic device 100 having these components is configured to interact with the service server 200 to perform various operations to receive the plant management service of the plant management system 1.

In certain embodiments, the electronic device 100 installs a plant management application to provide the service of the plant management system 1, executes the installed plant management application to access the service server 200, and accesses the service server 200. ) can also interact with.

The electronic device 100 is configured to input the user's account information to join the plant management system 1. The entered account information is used to create account registration data.

In addition, the electronic device 100 provides the user's space-related information, plants, etc. so that the plant management system 1 can provide plant-related services (e.g., customized plant recommendation service) in consideration of the user's space and the plants grown or being grown. It is configured to input related information and receive customized plant information for the user calculated based on the user's profile data and plant profile data.

Additionally, the electronic device 100 is configured to allow the user to experience various platform activities on the system.

The electronic device 100 is configured to input the user's evaluation score for the received customized plant and transmit it to the service server 200. The received customized plant may be a customized plant of the first type below.

The electronic device 100 receives user input for consuming content in the system or producing content, and executes a service operation (e.g., a content call request for consumption or a content creation request) according to the user input. It is configured to transmit a request to the service server 200 and receive a processing result (eg, desired consumption content or desired production content) for a service operation according to the request.

The electronic device 100 receives a user input for purchasing a plant through a commerce system included in the system or connected to the system, transmits a purchase request according to the user input to the commerce system, and makes a purchase according to the request. configured to receive processing results.

As the operation of the electronic device 100 is closely related to the service server 200, it will be described in more detail below with reference to the service server 200.

The service server 200 is two or more computer systems or computer software implemented as network servers. Here, a network server is a computer system and computer that is connected to a sub-device that can communicate with other network servers through a computer network such as a private intranet or the Internet, receives a request to perform a task, performs the task, and provides a performance result. Refers to software (network server program). However, in addition to these network server programs, it should be understood as a broad concept that includes a series of application programs operating on a network server and, in some cases, various databases built within it. The service server 200 may be implemented as any type or combination of types of computing devices, such as a network server, web server, file server, supercomputer, desktop computer, etc.

The service server 200 is configured to provide various plant-related services to the user through the user's electronic device 100. In certain embodiments, the service server 200 includes a communication module 220, an account management module 231, a profile analysis module 233, and a history analysis module 235. Additionally, in some embodiments, the service server 200 may further include a plant species profile DB 211, a user profile DB 213, and a user interest record DB 215.

The service server 200 may receive various data from the user's electronic device 100 through the communication module 220. Since the communication module 220 is similar to the communication module 110 of the electronic device 100, detailed description is omitted.

Each of the modules 231 to 239 may be implemented with at least one processor and/or memory.

The account management module 231 may generate the user's account registration data and grant access rights based on the user's account information received from the user's electronic device 100. Additionally, the account management module 231 may perform management operations to modify or update registered account registration information. Additionally, when the account management module 231 receives a system access request from the user's electronic device 100, the account management module 231 may check whether the user has access rights based on the requesting user's account registration data and approve the access.

The account registration data will be described in more detail with reference to FIG. 4 below.

The profile analysis module 233 is configured to receive space-related information about the user from the user's electronic device 100 and calculate the user's space characteristics by analyzing the space-related information.

In some embodiments, the user's space-related information may further include space type information. The space type information may include, for example, room (including studio), living room, bathroom, veranda, kitchen, office, outdoor, and other types. The service server 200 may obtain one or more space types in which the user can place plants from the user's electronic device 100.

The user's space-related information includes an image captured of the user's space (hereinafter, “space captured image”). Then, the profile analysis module 233 may calculate the user's spatial characteristics based on space-related information including the spatial captured image.

The user's spatial characteristics are the result of analyzing the space in terms of the architectural structural aspect of the space and the current utilization of the space by the user. The user's space characteristics may include one or more characteristics of object characteristics arranged in the space, space type characteristics, and space style characteristics.

Object characteristics include information on objects recognized in spatial images among objects placed in space. The object properties may be a list of recognized objects.

The profile analysis module 233 is configured to recognize the object expressed in the input image using a pre-learned object recognition model.

The object recognition model is a machine learning model with a neural network structure configured to extract features required to recognize an object from an input image, classify the class to which the object belongs based on the extracted features, and recognize the object. It may be possible. The object recognition model may be composed of, for example, a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or another vision recognition Neural Network (NN) structure, but is not limited thereto.

In some embodiments, the object recognition model may be configured to detect a foreground area distinct from the background in a single input image, extract features from a sub-image of the detected foreground area, and recognize which object the individual foreground is.

In the object recognition model, a class is a classification item that recognizes an object. Various product items that can be placed in space may be designated as classes. The plurality of classes may be, for example, windows, sofas, picture frames, pillows, lights, curtains, and/or other furniture.

The object recognition model is learned using a training data set consisting of a plurality of training samples. Each training sample includes a training image of the sample object and label data indicating the actual recognition value of the sample object. In some embodiments, the training image of the sample object may be a sample space image of the space where the sample object is placed.

When the training image of each training sample is input, the object recognition model extracts features from the input image and calculates an operation result value (i.e., prediction value) for recognizing the object of the input image. The parameters of the object recognition model are learned so that the error between the operation result value and the value (actual value) of the label data is further reduced. The learned object recognition model has parameters that minimize the error.

When the profile analysis module 233 receives space-related information including a spatially captured image from the user's electronic device 100, it inputs the spatially captured image into a previously learned object recognition model. The object recognition model extracts features from the input spatial capture image, calculates a probability value that the object included in the spatial capture image belongs to each of a plurality of preset classes based on the extracted features, and the calculated probability value is the highest. The object may be recognized as a high-class product item.

Some or all of the recognition results output from the object recognition model are used as object characteristics. The recognition result includes class information (eg, recognition item) to which the object is determined to belong. In some embodiments, the recognition result may further include a probability value that the object belongs to the determined class.

When a plurality of spatial images taken from different views of a single space and/or a plurality of objects in a space captured from a single space captured image including a plurality of objects are recognized, the profile analysis module 233 is You can also create a list of recognized object properties.

The profile analysis module 233 is configured to calculate the appearance of the recognized object. The appearance of the object may include color, shape, and size.

Additionally, the profile analysis module 233 is configured to calculate the location of the recognized object. The position of the object may be expressed as an absolute position on the frame of the spatially captured image. Additionally, the profile analysis module 233 may be further configured to calculate placement relationships between objects. The placement relationship of objects expresses the relative positional relationship between an object and other objects under the internal structure of space.

The profile analysis module 233 may calculate object characteristics including one or more object information among some or all of the recognition results (eg, object name), the location of the object, the arrangement relationship of the objects, and the appearance of the object.

Additionally, the profile analysis module 233 may be further configured to calculate the style of the corresponding space based on the object characteristics. The style of the space expresses the internal atmosphere and tendency and may be referred to as an interior style.

The profile analysis module 233 may calculate the space style from the space structure map through various methods.

In some embodiments, the profile analysis module 233 may pre-store a reference style data set including object characteristics belonging to each of a plurality of styles. Each reference style data set is subset into a respective style. Each subset consists of object features in the sample space classified by that style.

The profile analysis module 233 compares the object characteristics of the space for which a style is to be determined with the object characteristics in the reference style data set, searches for matching object characteristics, and matches the reference style to which the searched object characteristic belongs in the input structure map. You can also decide based on the style of the space. In this case, the profile analysis module 233 may select object characteristics of the sample space that match object characteristics of the desired space using cosine similarity, Euclidean similarity, or other vector similarity calculation methods.

In some other embodiments, the profile analysis module 233 may be configured to recognize a style for a space in which an object is recognized using a pre-learned style recognition model.

Since the style recognition model is similar to the object recognition model, the differences will be mainly described.

The style recognition model is a machine learning model configured to predict the style characteristics of the space in which the input object is recognized by inferring the correlation between the input object characteristics of the space whose style is to be recognized and the style of the input space. The machine learning model has a neural network structure.

The neural network structure is configured to recognize the style of the class in which the desired space is most likely to be classified as the style of the input space, based on the correlation between the input object characteristics and the style of the input space. For example, the neural network structure may be composed of a multi-layer perceptron (MLP) or other NN structure including a fully connected layer.

The input object characteristics include object information for each placement location. The location of an object based on space may be converted to correspond to object characteristics. Additionally, the input object characteristics may further include other object characteristics (color, appearance, etc.).

The style recognition model is learned using a training data set consisting of a plurality of training samples. Each training sample has training data and label data, respectively. The training data includes sample object characteristics calculated from a sample space image taken of the sample space. The sample object characteristics correspond to the input object characteristics.

If the first sample has the value v1 of the first object feature and the second sample has the value v2 of the second feature, the training data of the first sample is {v1, 0}, and the training data of the second sample is {0, v2. } may also be included.

The label data has a value indicating the actual style of the corresponding sample space.

When training data (eg, sample object characteristics) of each training sample is input, the style recognition model calculates a result value (eg, predicted value) by processing the training data with internal parameters. The parameters of the style recognition model are learned so that the error between the operation result value and the value (actual value) of the label data is further reduced. The learned object recognition model has parameters that minimize the error.

Additionally, the profile analysis module 233 may be further configured to calculate the ventilation level of the space. The ventilation level indicates the ease of ventilation of the space. In one example, the ventilation level may be expressed as a higher value as ventilation improves. For example, the ventilation level may indicate better ventilation at level 5 than at level 1. However, the ventilation level is not limited thereto, and the ventilation level may be expressed inversely.

The profile analysis module 233 may calculate the ventilation level of the space based on the number of windows.

Additionally, in some embodiments, the profile analysis module 233 may calculate the ventilation level of the space based on at least one of the number of windows, the space type, and the ventilation area of the windows.

Ventilation levels may also depend on the type of space. The type of space is closely related to the internal volume of the space, and the larger the internal volume, the better the internal circulation. The profile analysis module 233 may also store preset ventilation levels for each space type. For example, the degree of ease of ventilation for each space type may be specified in advance in the following order: outdoor (or garden), veranda, living room, room (studio), bathroom, and kitchen. The value of ventilation level for each space type may be based on various statistical data or may be arbitrarily predetermined by the system operator.

The profile analysis module 233 may calculate the ventilation level as shown in the table below. In the table below, higher level values indicate easier ventilation.

division Window O (1 piece) window x Room (studio) 3 One living room 5 2 restroom 2 One kitchen 2 One Veranda 5 3 Outdoor (garden) 5 5

In some embodiments, the profile analysis module 233 may calculate the ventilation level of the space through the following equation based on the space type and number of windows.

Here, Aspace_window represents the ventilation area of the wth window. m represents the total number of recognized windows. Fspace_type represents the ventilation level for each space type depending on the presence/absence/area of windows.

The ventilation area may be less than half of the total area of the recognized window. This is because windows are generally configured so that only an area less than half the size of the entire transparent area is open.

In some embodiments, the profile analysis module 233 may pre-store a space type-ventilation level table that maps ventilation levels according to window conditions for each space type. The Fspace_type may be a function corresponding to the space type-ventilation level table. F is a function that converts an input value into a range of pre-designated ventilation levels. For example, F may be configured to convert an input value (i.e., space_type(n*space_window)) to any one of levels 1 to 5. Through Equation 1, the profile analysis module 233 may calculate the ventilation level in Table 1.

In addition, the system 1 may obtain information about the user's surrounding environment and calculate the user's space characteristics, including the illuminance characteristics of the user space.

The user's space-related information obtained from the electronic device 100 may further include information on the user's surrounding environment. The surrounding environment information includes illuminance information acquired by the illuminance sensor 170. The illuminance information may include a measured value of illuminance detected by the illuminance sensor 170. The measured value may be a digitally converted illuminance measurement value based on the detected light energy. The digital measurement value may be, for example, Lux, but is not limited thereto. The converted digital measurement value is used as the illuminance characteristic value.

Then, the user's spatial characteristics calculated by the profile analysis module 233 may further include illuminance characteristics.

The profile analysis module 233 may generate space registration data of the user including one or more space characteristics among object characteristics arranged in space, space type characteristics, space style characteristics, and illuminance characteristics. The profile analysis module 233 may assign a user space identifier (eg, user space ID) and generate user space registration data including the assigned user space ID and the one or more space characteristics. This user space registration data will be described in more detail with reference to FIG. 4 below.

Additionally, the profile analysis module 233 is configured to receive plant-related information about the user from the user's electronic device 100 and generate plant registration data for the user.

The plant-related information includes information describing plants grown or grown by the user (hereinafter, “experienced plant information”). The experienced plant refers to a plant that is currently being experienced or has been experienced in the past. This experiential plant information may include one or more of the name, nickname, photograph, and other plant information of the plant that the user is currently growing.

Additionally, the experienced plant information includes the number of experienced plants. For example, the experience plant information includes the number of registered plants currently being grown and the number of plants that have been grown.

In some embodiments, the plant-related information may further include information that describes the surrounding environment of the registered space where the plant is to be placed at a level that can be easily recognized by the user (hereinafter, “user-perceived environment information”). The registered space is designated through space-related information.

The user recognized environment information may include one or more of a light recognized environment, a ventilation recognized environment, a flower pot recognized environment, and other recognized environment information.

The light recognition environment and ventilation recognition environment describe the environment around the plant more non-quantitatively than the above-mentioned illuminance characteristics and ventilation characteristics, taking into account the level of awareness of users, including non-experts.

For example, the light perception environment is a value representing 'receiving sunlight from a window', 'receiving sunlight from inside a window', 'receiving plant lighting', or 'receiving no sunlight or light'. You can also have

The ventilation recognition environment may have a value indicating 'there is a window' or 'there is no window but the wind flows through'. In some embodiments, 'having a window' may be distinguished as having a window within or outside a certain distance (eg, 5 m) from the currently experienced plant.

The profile analysis module 233 may generate user plant registration data including the plant-related information.

The profile analysis module 233 acquires some or all of the user's experienced plant information and user-perceived environment information included in the plant-related information as the user's plant characteristics.

The profile analysis module 233 assigns an identifier (e.g., user experience plant ID) to the user's plant, and uses a pre-assigned and stored plant identifier indicating each plant (or plant variety) based on the plant name. Search. The profile analysis module 233 may generate the user's plant registration data including the user's plant characteristics obtained from the assigned user experience plant ID, the searched plant ID, and the plant-related information.

The user's plant may correspond to a nickname. For example, if a user is growing two identical olive trees by nicknames Tree 1 and Tree 2, the profile data for the plant of Tree 1 includes a first ID pointing to Tree 1 and an olive tree ID pointing to the olive tree. The profile data of the plant of Tree 2 may include a second ID indicating Tree 2, and the Olimu Tree ID. This user's plant registration data will be described in more detail with reference to FIG. 4 below.

In some embodiments, the plant ID may be searched in the plant species profile DB 211. The plant species profile DB 211 is described in more detail with reference to FIG. 3 below.

Figure 3 shows profile data of plant species stored in a plant species profile DB, according to an embodiment of the present application.

Referring to FIG. 3, the plant species profile DB 211 stores plant species profile data for each of a plurality of plant species.

Plant species profile data includes plant species characteristics that describe the variety and cultivation information for each plant. The variety information and cultivation information are obtained from plant illustrations, various plant organ data, etc.

The cultivation information describes the cultivation method and cultivation environment. As shown in FIG. 3, the cultivation information may include, for example, one or more conditions among light conditions, ventilation conditions, pot conditions, and companion animal conditions required for the plant to be well grown without dying. The ventilation condition may be expressed as the ventilation level described above.

In some embodiments, the plant species profile data may further include the popularity of the plant. The popularity may also be expressed as the number of views for the plant on an Internet portal site or commerce system. To this end, the profile analysis module 233 may obtain the number of searches for each plant during a certain period of time, including the Internet search volume inquiry API.

The profile analysis module 233 generates profile data of the plant species by assigning a unique system identifier (eg, plant ID) to each plant.

The service server 200 may generate the user's profile data based on the user's account registration data, the user's space registration data, and the user's plant registration data. The plant registration data may include some data retrieved from the plant species profile DB 211. The generated user profile data may be stored in the user profile DB 213.

Figure 4 shows user profile data stored in a user profile DB according to an embodiment of the present application.

Referring to FIG. 4, the user profile DB 213 stores profile data for each of a plurality of users.

The user's profile data describes the user's account registration information, the user's space registration information, and the user's plant registration information.

The user's account registration data includes the user's system identifier (eg, user ID), account name, and the user's geographic location. Additionally, in some embodiments, the user's account registration data includes an identifier for the user's space (e.g., user space ID), an identifier for the user's experience plants (e.g., user experience plant ID), and an identifier for the user's growing experience. It may also include one or more items of identifier (e.g., plant ID) for the plant and its value.

The user's space registration data includes the user space ID. When a user registers an experience plant, the user's space registration data further includes the user experience plant ID. Additionally, in some embodiments, the user's space registration data may include items and values of one or more of space type characteristics, window characteristics, ventilation level characteristics, interior style characteristics, and illuminance characteristics. The window characteristics may include one or more of the presence or absence of a window, the number of windows, and the ventilation area of the window.

The user's plant registration data includes a user experience plant ID, a plant ID, a captured image of the user experience plant, and the user's plant characteristics. For example, as shown in FIG. 4, the user's plant registration data may include items and values of one or more characteristics of a light recognition environment, a ventilation recognition environment, and a pollen recognition environment included in the user recognition environment information. It may be possible.

Referring again to FIGS. 1 and 2, the service server 200 is configured to predict customized plants for the user based on the user's profile data and the plant species' profile data. The customized plant for the user is a plant predicted to be well grown or of high interest by the individual user, depending on the user's registered space, user experience, etc. Some or all of the custom plants may be recommended to that user.

In certain embodiments, the profile analysis module 233 may search for a first type of customized plant based on the user's profile data and the plant species' profile data. Information on some or all of the searched first type of customized plants may be transmitted to the electronic device 100 and some or all of the searched first type of customized plants may be recommended to the user.

The searched plants are plants in which at least one of the plant species characteristics included in the plant species profile data of each plant matches a characteristic in the user's profile data. The profile analysis module 233 searches for one or more plants that match the plant search filtering conditions among the user's profile data.

The filtering condition may include at least one characteristic item among characteristic items included in the user's profile data. For example, the filtering condition may use at least one of plant characteristics and space characteristics, such as one or more of light quantity, ventilation, and companion animals, as a search filtering condition.

In some embodiments, the profile analysis module 233 calculates the user's suitability score for each searched plant based on the number of characteristics matching between the profile data of the plant species and the user's profile data, and adds the suitability score to the plant species. It is also possible to create a custom plant list of the first type based on the first type. The provided search result of the first type of customized plant may include a list of the first type of customized plant.

The fitness score is given a higher score as the number of matching characteristics increases.

In the customized plant list, the searched plants are arranged so that plants with relatively high suitability scores are provided to the user first. For example, a screen of a customized plant list arranged in order of high suitability score may be provided to the user's electronic device 100.

Additionally, in some embodiments, when the user's profile data includes a plurality of spatial registration data, the profile analysis module 233 selects plants searched based on spatial characteristics within each spatial registration data for each registered space. You can also categorize and create customized plant lists for each registered space.

Specifically, the profile analysis module 233 calculates a suitability score for each registered space based on the number of characteristics matching between the profile data of the plant species of the searched plant and the spatial characteristics in each spatial registration data, and calculates a suitability score for each registered space. You can also create a customized plant list for each registered space based on star suitability scores.

The user's profile data may include spatial characteristics for the first space and spatial characteristics for the second space.

For example, for a user who has registered two rooms (first space and second space) as user spaces, a customized plant list can be created for each room. A specific plant (A) may match the spatial characteristics of the first space but may not match the spatial characteristics of the second space. Then, the specific plant (A) is classified as a customized plant for the first space.

Since the customized plant list for each registered space is similar to the customized plant list described above, detailed description is omitted.

Referring again to Figures 1 and 2, the service server 200 may utilize the user's history data to additionally recommend plants customized to the user. Recommendation of customized plants based on history data may be performed after, before, or in parallel with the recommendation of customized plants based on the above-described profile data, depending on the type of history data. The customized plants based on history data and the customized plants based on profile data may be the same or at least partially different from each other.

The service server 200 predicts customized plants based on the history data by the history analysis module 235. Hereinafter, for clarity of explanation, the customized plants predicted based on the historical data will be referred to as a second type of customized plant to be more clearly distinguished from the first type of customized plant, and the embodiments of the present application will be described in more detail.

The history analysis module 235 may generate user history data for each of a plurality of users who have signed up for the system.

In certain embodiments, the historical data includes one or more of the user's rating history data, activity history data, and purchase history data. The history analysis module 235 may predict a second type of customized plant for the user based on at least one of the user's evaluation history data, activity history data, and purchase history data.

The historical analysis module 235 may generate historical evaluation data including some or all of the user's registered account data (e.g., user ID) and evaluations of recommendations of the first type of personalized plants.

Specifically, when the history analysis module 235 receives the user's evaluation of the first type of customized plant recommended by the profile analysis module 233, the history analysis module 235 includes an evaluation of the recommendation of the first type of customized plant. User evaluation history data can also be generated.

The user's evaluation may be obtained as a value (hereinafter referred to as “evaluation score”) that evaluates the user's satisfaction with receiving a customized plant recommendation. The evaluation score may be the user's review value for the customized plant. In some embodiments, the evaluation score may be a subjective value about the user's feeling of actually growing the recommended first type of customized plant.

Then, the evaluation history data includes an already recommended first type of customized plant and an evaluation score for the recommendation of the first type of customized plant.

In some embodiments, the evaluation history data may further include some or all of the user's profile data used to recommend a first type of customized plant. For example, the evaluation history data may further include spatial characteristics within registered spatial data.

In addition, in some embodiments, the history analysis module 235 evaluates the first type of customized plant and the recommendation based on the evaluation history of a plurality of users for which one or more first type of customized plants are recommended for each user. It is also possible to calculate each user's recommendation score for each first type of customized plant by analyzing a plurality of factors that constitute the correlation between them. For some users, the recommendation score may be calculated for a first type of customized plant that was not recommended to the user but was recommended by another user.

The recommendation score is a score obtained by predicting how satisfied the user will be when a plant is recommended to the user in terms of the plurality of factors. The user's recommendation score for a different first type of customized plant that is recommended only to other users and not to the user is a quantification of the user's predicted satisfaction level when the other first type of customized plant is recommended to the user. . On the other hand, the user's recommendation score for the first type of customized plant that has already been recommended to the user may be a value that objectively or quantitatively quantifies the user's level of satisfaction with the first type of customized plant in terms of the plurality of factors. This is because the evaluation score in the evaluation details represents the user's subjective evaluation value of having been recommended the first type of customized plant.

The history analysis module 235 forms an evaluation history table consisting of evaluation scores in evaluation history data of a plurality of users for the plurality of first customized plants, and uses the evaluation history data to create a plurality of first customized plants for each user. It is also possible to calculate a recommendation score for each type of customized plant.

The evaluation history data may be expressed in matrix form. In the rating history matrix, one axis (eg, columns) represents a plurality of users who have signed up for the system, and another axis (eg, rows) represents a first type of customized plant recommended to the plurality of users. In the evaluation history matrix, each matrix element represents the value of the user's evaluation score for the recommendation of the corresponding first type of customized plant.

Figure 5 is a schematic diagram of a process for calculating a recommendation score according to an embodiment of the present application.

Referring to FIG. 5, the evaluation history table may be expressed in matrix form. Then, the history analysis module 235 decomposes this evaluation history matrix into a first sub-matrix and a second sub-matrix based on a plurality of factors constituting the correlation between the recommended customized plant and the evaluation score, and Based on the first sub-matrix and the second sub-matrix, a recommendation score for each of the plurality of first types of customized plants may be calculated for each user.

The details analysis module 235 may decompose the evaluation details matrix into a first sub-matrix and a second sub-matrix by applying a preset matrix decomposition method. The first sub-matrix and the first sub-matrix are matrices of lower dimension than the evaluation history matrix.

The matrix decomposition method may be, for example, Singular Value Decomposition (SVD), Non-Negative Matrix Factorization (NMF), stochastic gradient descent (SGD), Alternating Least Squres (ALS), or other methods.

In the first sub-matrix, one axis (eg, column) represents a plurality of users who have subscribed to the system, and another axis (eg, row) indicates the plurality of factors. In a first sub-matrix, one axis (eg, column) represents the plurality of factors and another axis (eg, column) represents the plurality of recommended customized plants of the first type.

The history analysis module 235 calculates a recommendation score for the user and the customized plant based on the first and second sub-matrices.

In some embodiments, the history analysis module 235 performs matrix decomposition (e.g., through SVD) on all matrix component values, including undetermined values in the evaluation history matrix, into a 1-1 sub-matrix and a 1-2 sub-matrix. Obtain, process the transpose matrix of any one of the 1-1 sub-matrix and the 1-2 sub-matrix and the remaining sub-matrix to perform an inner product, and re-calculate the recommendation score for each user and each customized plant. You may. For example, the history analysis module 235 may process the inner product of the transpose matrix of the 1-1 sub-matrix and the 1-2 sub-matrix.

In the above example, when the evaluation value at the matrix position of the user in row u and the first type of customized plant in column i in the evaluation history matrix is referred to as r _ui , the history analysis module 235 uses the following equation: Through this, a recommendation score may be calculated for the user in row u and the first type of customized plant in column i.

Here, x represents the 1-1st sub-matrix and y represents the 1-2nd sub-matrix.

In some other embodiments, the history analysis module 235 may calculate the entire matrix component values including undetermined values in the evaluation history matrix having the u column and the i row to any 2-1 sub-matrix having the u column and the i row. is decomposed into an arbitrary 2-2 sub-matrix, and the transpose matrix of any one of the 2-1 sub-matrix and the 2-2 sub-matrix and the inner product of the remaining sub-matrix are processed for each user and each Calculate a prediction matrix having a recommendation score for a first type of customized plant as a matrix component value, and the 2-1 sub such that the difference between the matrix component value of the prediction matrix and the matrix component value of the evaluation history matrix is minimized. The matrix component values of the matrix and/or the 2-2 sub-matrix are updated, and the transpose matrix and the remaining sub-matrix of any one of the updated 2-1 sub-matrix and 2-2 sub-matrix are inner product processed. It may also be configured to recalculate a recommendation score for each user and each customized plant.

In some embodiments, the difference between matrix component values of the prediction matrix and matrix component values of the evaluation history matrix may be referred to as a loss function. The loss function may be expressed as the following equation.

Then, the history analysis module 235 may newly recommend a second type of customized plant based on the recommendation scores for the user and the first type of customized plant calculated through Equation 2 or Equation 3. The history analysis module 235 may calculate a second type of customized plant when the recommendation score is greater than or equal to a preset recommendation threshold score. The recommendation threshold score may be based on statistical data or may be arbitrarily pre-designated by the system operator.

In some embodiments, the history analysis module 235 may generate a second type of customized plant list based on the user's recommendation score for each calculated second type of customized plant.

The second type of customized plant list is a search result of a new customized plant provided after recommending the existing search results provided as the first type of customized plant list.

The second type of customized plant list includes a first type of customized plant that has already been recommended to a specific user and another first type of customized plant that has been recommended to another user, as shown in Figure 5. It comes true.

The second type of customized plants are arranged so that plants with relatively high recommendation scores in the second type of customized plant list are provided to the user first. For example, a screen of a second type of customized plant list arranged in order of high recommendation score may be provided to the user's electronic device 100.

In some embodiments, when two or more first types of customized plants among the plurality of second types of customized plants have the same recommendation score, the two or more second types of customized plants in the second type of customized plants list Plants can also be arranged in order of popularity.

As mentioned above, popularity can also be expressed as the number of views for the plant on Internet portal sites and commerce systems. The history analysis module 235 may search the popularity of each of the two or more plants in the plant species profile DB 211 and arrange them so that plants with a high recognition value (i.e., number of views) are provided to the user first.

Additionally, in some embodiments, the history analysis module 235 filters the plurality of second types of custom plants in the list for at least some of the second type of custom plants that match the user's registered plant data in the user's profile data. Thus, the second type of customized plant list can be created.

The user's registered plant data includes plant information that the user is currently experiencing or has experienced in the past. The history analysis module 235 may generate a second type of customized plant list by excluding experienced plants that match experienced plant information in the user's registered plant data from the second type of customized plants. As a result, the history analysis module 235 may intensively recommend new plants that the user has not experienced as the second type of customized plants.

Referring again to FIGS. 1 and 2 , the history analysis module 235 may generate user activity history data based on a request received from the user's electronic device 100 or a service provided by the system. In the system 1, the user's activities may be treated as events.

User activities include consuming plant content and producing plant content.

The plant content refers to data that describes plant-related information in text, image, video, or multimedia format. The plant content may include, for example, one or more of a plant guide, plant guide details, a gardening blog, a planter feed, a list of my plants, a list of plants I have grown, and other content.

The activity history data includes plant content consumption history data and/or production history data. The history analysis module 235 provides plant content production history data describing the user's plant content consumption activity on the system platform when one or more of the user's plant content production event and plant content consumption event occurs, and the user Activity history data may be generated including one or more of plant content consumption history data that describes the activity that produced the plant content.

Each activity history data includes at least one of user account information, plant content information for consumption or production, activity time such as consumption or production, and activity details performed at that time.

The activity time may be expressed in units of years/months/days and/or hours/minutes/seconds.

For example, plant content consumption history data may include at least some of the user account information and consumption activity details.

In some embodiments, the activity content included in the plant content consumption history data includes keyword search on the plant guide or gardening blog search page, plant guide content view, gardening blog view, bookmark registration of the viewed content, feed content interaction, and my plant list. You can also include one or more activities such as adding a plant to or adding a plant to the list of plants you have grown.

The keywords include keywords of interest to the user. The keywords of interest may include keywords of plant species in which the user is interested or is likely to have interest. Consumption history data including the activity content may include keywords of interest (eg, plant species keywords) used in the activity.

Additionally, the plant content production history data may include at least some of the user account information, content content produced by the user, and production time.

The activity content included in the plant content production history data may include one or more of the following: chat creation, post creation, plant guide or gardening blog creation, and feed content creation.

Activities through which plant content consumption history data and production history data are obtained will be described in more detail with reference to FIGS. 14 to 16 below.

The history analysis module 235 may generate user interest record data based on activity history data.

The history analysis module 235 may extract a user's interest keyword from consumption history data and production history data, and generate user interest record data by associating the consumption/production history data and the interest keyword.

The history analysis module 235 may obtain the user's keywords of interest from user input obtained during the activity process.

For example, when plant species text is input to search plants, the plant species text is obtained as a keyword of interest. Alternatively, when text is input to create content, some of the input text may be obtained as a keyword of interest.

In some embodiments, the history analysis module 235 may be configured to extract a user's keywords of interest from the user's produced content. The history analysis module 235 is configured to recognize text data in user-produced content and extract keywords of interest by processing the recognized text data into natural language.

If the content produced by the user includes an image, the history analysis module 235 may recognize the user's text data using the OCR method.

Figure 6 is a schematic diagram of a keyword-of-interest extraction operation according to an embodiment of the present application.

Referring to FIG. 6, the history analysis module 235 may extract keywords of interest to the user from the user's text data using text mining or other natural language processing methods.

The history analysis module 235 generates interest record data including extracted interest keywords and activity history data from which the interest keywords were extracted.

Additionally, in some embodiments, the history analysis module 235 assigns an interest score that quantifies the user interest in the interest keyword extracted from content related to activity history data to the extracted interest keyword and further includes the interest score. Interest history data may also be generated.

The history analysis module 235 may assign interest scores to interest keywords extracted from consumption history data using a pre-stored interest score table.

The interest score table represents the relationship between activity content and interest scores. The relationship between activity content and interest scores includes a relationship in which consumption activities mainly performed by users with a high interest in a specific plant are given a high interest score, or a relationship in which a lower interest score is given when there is no interest in a specific plant. . The interest score may have a positive integer, 0, or negative integer value.

In some embodiments, the interest score table may include pre-specified interest scores for each activity content. For example, the interest score table may be structured as follows.

Activity details
(event_name) Activities interest score Click_plantwiki_searchterm Simple query One Click_plantwiki_suggestion Simple query One Click_blog Simple query One Click_plantwiki_goodmatch Simple query One Click_plantwiki_bookmark bookmark 2 Click_plantwiki_interaction Like, bookmark, comment 2 Myplant_past_complete Register plants you have grown -One Myplnat_present_complete My plant registration -2

The history analysis module 235 determines the local frequency in which the extracted keyword of interest appears in the content produced by the user from which the keyword of interest was extracted and the global frequency in which the extracted keyword of interest appears in the entire content produced by a plurality of users. Based on this, the user's interest score may be assigned to the interest keyword, and the calculated interest score may be assigned to the corresponding produced content. The global frequency is the frequency of appearance in the entire range of individually produced content of all users over the entire period or a certain period of time on the system.

The history analysis module 235 may assign a higher user interest score to an interest keyword whose local frequency is greater in the ratio of the local frequency to the global frequency. If the proportion of the local frequency is greater, this user used the keyword of interest more than other users, which ultimately means that the user had more interest in the keyword of interest than other users.

In one example, the history analysis module 235 may assign an interest score to each user for each interest keyword using a TF-IDF (Term Frequency-Invserse Document Frequency) method, but is not limited to this.

The history analysis module 235 forms a plurality of clusters consisting of some users with similar interest tendencies among the plurality of users based on the interest record data of each of the plurality of users.

The history analysis module 235 may classify a plurality of users into a plurality of clusters using a preset clustering method.

In one example, the preset clustering method coordinates each user's interest record data into vector points to form a random initial cluster, and continuously updates the center point of the cluster using the average coordinates of the vector points belonging to each initial cluster. This may be a K-means clustering method that determines multiple clusters, but is not limited to this. The preset clustering method is

In some embodiments, the history analysis module 235 may calculate the number of clusters appropriate for the range of interest record data of the plurality of users. Then, the initial cluster is formed based on the calculated number of clusters.

In one example, the history analysis module 235 may calculate the optimal number of clusters using the elbow method. The history analysis module 235 may calculate the ratio of the variance between clusters and the total variance, and calculate the point where the increase in the variance ratio decreases as the number of clusters.

The history analysis module 235 may recommend a second type of customized plant for the user based on keywords of interest related to the group to which the user belongs. The history analysis module 235 may calculate plants matching a plurality of keywords of interest related to the group to which the user belongs as a second type of customized plant. For example, if a user (A) with a keyword of interest (WA) and a user (B) with a keyword of interest (WB) belong to the same cluster, the plants matching their keyword of interest (WA) and the interests of other users Plants matching the keyword (WB) may be calculated as a second type of customized plant for each user (A) and user (B).

Additionally, the history analysis module 235 may provide a second type of customized plant list consisting of the second type of customized plants.

In some embodiments, the history analysis module 235 provides a list of second types of customized plants in which the second type of customized plants retrieved through matching are arranged based on the user's interest score to the user's electronic device 100. You can also provide it.

Then, individual second type customized plants in the second type customized plant list may be arranged based on the interest score of the cluster.

The customized plants are arranged so that plants of the second type with relatively high cluster interest scores are provided to the user first in the list of customized plants of the second type. For example, a screen of a second type of customized plant list in which the clusters are arranged in order of high interest scores may be provided to the user's electronic device 100.

In some embodiments, the history analysis module 235 may perform a second type of customization for the user based on only the remaining interest keywords that are not included in the user's interest record data among the plurality of interest keywords related to the cluster to which the user belongs. You can also provide plants. Then, the user may receive new recommendations for plants that he or she was not previously aware of but are suitable for him or her. For example, if a user (A) with a keyword of interest (WA) and a user (B) with a keyword of interest (WB) belong to the same cluster, the user (A) has plants that match only the keyword of interest (WB). , a second type of customized plant list may be provided.

In some embodiments, if two or more customized plants in the second type of customized plant list have the same interest score, the two or more customized plants in the second type of customized plant list will be arranged in order according to the popularity of the plants. It may be possible.

Additionally, the history analysis module 235 may predict a second type of customized plant for the user based on the user's purchase history data. In the system 1, purchases may also be treated as events.

The purchase history data is generated when a user purchases a plant through a commerce system connected inside or outside the system. The purchase history data includes information describing this purchase event. The purchase history data includes purchased plant information. The purchased plant information may include at least part of the registration account data of the purchaser, that is, the user (eg, user ID), and at least part of the profile data of the plant species of the purchased plant (eg, plant species ID).

Additionally, in some embodiments, the purchase history data may further include purchase price, purchase time, seller information, etc.

In certain embodiments, the history analysis module 235 may calculate a second type of customized plant for the user from the user's purchase history data using a pre-trained purchase prediction model.

The purchase prediction model is a machine learning model configured to infer the correlation between a plurality of characteristics included in user-related data and the plants the user purchases, and to custom-predict plants to be purchased by the user of an input data set containing a plurality of input characteristics. . The machine learning model has a neural network structure.

The neural network structure is configured to calculate the class of plants that the user is most likely to purchase as a second type of customized plant for the user based on the correlation between the plurality of input characteristics and the user. For example, the neural network structure may be composed of a multi-layer perceptron (MLP) or other NN structure including a fully connected layer.

Figure 7 is a schematic diagram of a learning process of the purchase prediction model, according to an embodiment of the present application.

Referring to Figure 7, the purchase prediction model is learned using a training data set consisting of a plurality of training samples. Training samples are user-specific. Each training sample in the set includes training data and label data for each sample user.

In the case of a user who has purchase history including purchased plants, the label data may have a value indicating the purchased plant (eg, a plant identifier value). In some embodiments, the label data may have a value representing the user's experienced plants included in the user's registered plant data in the user's profile data in the case of a user without purchase history (e.g., a first-time registered user). This is because the purchase history has not been secured through a commerce system connected inside or outside the above system, but it has been raised or can currently be raised because the purchase was made somewhere. Even if the plant is transferred from another person, such as a gift, this transfer can also be treated as an event equivalent to a purchase by the user. This is because they did not refuse the transfer, and in general, others are more likely to transfer the plant considering the tastes of the transferee, that is, the user.

Training data is formed based on at least one user's data among profile data of each user, activity history data of the user, and purchase history data of the user. For example, a user's training sample may include some or all of the characteristic items included in the user's profile data, activity history data, and purchase history data. In the system 1, raw data of various historical data for each user is input to the purchase prediction model as training data.

As shown in FIG. 7, the purchase prediction model may be an MLP model. The MLP model includes an input layer, a plurality of hidden layers, and an output layer. Each layer contains one or more nodes. In some embodiments, the input layer may have as many nodes as the number of feature items in the training data. The value of a specific item is entered into the corresponding specific node.

The hidden layer is configured to receive the output of the previous layer, process the operation, and pass the operation result as the input of the next layer. The computational processing of the hidden layer is performed based on parameters learned to infer correlations between a plurality of input characteristics and users.

The output layer is configured to calculate the probability that a user will purchase a specific plant as a predicted value based on the correlation between the user and a plurality of input characteristics calculated while passing through a plurality of hidden layers.

As shown in Figure 7, when each training data (eg, a plurality of characteristic items) is input, the purchase prediction model computes the training data with internal parameters to calculate a result value (ie, a predicted value). The parameters of the purchase prediction model are learned so that the error between the calculation result value and the value of the label data (i.e., the actual value) is further reduced.

The error may be expressed as a logistic loss function or cross entropy loss function.

The learned purchase prediction model has parameters that minimize the error. Through this learning process, the purchase prediction model learns the factors that most influenced the occurrence of a purchase event, and eventually, the learned purchase prediction model can predict plants that the user is likely to purchase during the next purchase. The history analysis module 235 calculates the plants predicted to be purchased as a second type of customized plant for the user.

In some embodiments, the purchase prediction model may determine one or more plants whose calculated purchase probability for each plant is higher than a preset threshold probability as the second type of customized plant. The threshold probability may be, for example, 0.5, but is not limited thereto.

Output data including plant information of the determined second type of customized plant is produced. Additionally, the output data may further include the purchase probability value.

Additionally, the history analysis module 235 may generate a second type of customized plant list consisting of the selected second type of customized plants.

In some embodiments, when two or more plants among the plurality of selected customized plants of the second type have the same purchase probability, the two or more plants in the list of customized plants of the second type are arranged in order according to the popularity of each plant. It could be.

As mentioned above, popularity can also be expressed as the number of views for the plant on Internet portal sites and commerce systems. The history analysis module 235 may search the popularity of each of the two or more plants in the plant species profile DB 211 and arrange them so that plants with a high recognition value (i.e., number of views) are provided to the user first.

Additionally, in some other embodiments, the history analysis module 235 may calculate a second type of customized plant from the user's activity history data using a pre-learned purchase prediction model.

Since the purchase prediction model into which the user's activity history data is input is similar to the purchase prediction model of FIG. 7, the differences will be mainly described.

The training data of the purchase prediction model may be activity history data of a sample user. For example, each training sample may be {sample user's consumption history data, label data} or {sample user's production history data, label data}.

When the user's activity history data is input into the purchase prediction model learned in this way, the correlation between the user's input activity history data and the purchase probability is inferred, and the purchase probability of the user with the input activity history data is calculated. Here, plants whose purchase probability is greater than or equal to a preset threshold probability are calculated as the second type of customized plant.

Since the learning process and recommendation process using the training data set consisting of the training data are similar to the learning process and recommendation process of FIG. 7, detailed descriptions are omitted.

The history analysis module 235 selects one of the recommended results based on each of the evaluation history data, activity history data, and purchase history data to provide a single customized plant (or list) for the user, or three recommendations. You can also provide multiple custom plants (or lists) using each method.

8A to 8C illustrate a recommendation result screen provided in a multi-layer structure according to an embodiment of the present application.

Referring to FIGS. 8A to 8C, the recommendation results of the multiple customized plants may be provided in one view.

The history analysis module 235 may transmit a recommendation result screen composed of a multi-layer structure to the user's electronic device 100. The recommendation result screen may have an area larger than the size of the output device 150 of the electronic device 100.

Each of the multiple layers includes recommendation results for each recommendation method. The recommendation result screen may include multiple sub-screens corresponding to each of multiple layers. Each sub-screen can also display a customized plant list calculated through the corresponding recommendation method.

For example, when the electronic device 100 is a smart phone, the multi-layer structure-based recommendation result screen is configured to move a customized plant displayed according to a scroll movement input on the screen. The area of the sub-screen displayed may be moved by scrolling on this screen.

In some other embodiments, the history analysis module 235 may integrate the three recommendation results to generate multiple customized plant lists. In this case, plants included repeatedly in each list are included as a single item in the single customized plant list.

Additionally, in some other embodiments, the profile analysis module 233 may calculate a second type of customized plant from the user's profile data using a pre-learned purchase prediction model.

Since the purchase prediction model into which the user's profile data is input is similar to the purchase prediction model of FIG. 7, the differences will be mainly described.

The training data of the purchase prediction model may be profile data of a sample user.

When the user's profile data is input into the purchase prediction model learned in this way, the correlation between the user's input profile data and the purchase probability is inferred, and the purchase probability of the user with the input profile data is calculated. Plants whose purchase probability is greater than or equal to a preset threshold probability are calculated as a second type of customized plant.

Since the learning process and recommendation process using the training data set consisting of the training data are similar to the learning process and recommendation process of FIG. 7, detailed descriptions are omitted.

It will be clear to those skilled in the art that the electronic device 100, service server 200, or system 1 may include other components. For example, the service server 200 may include other hardware elements necessary for the operations described herein, including input devices for data entry and output devices for printing or other data display. In addition, the system 1 may further include a network, network interface, and protocol connecting the service server 200 and an external device (eg, a user terminal, an external database, etc.).

9A to 9D are flowcharts of an artificial intelligence-based user-customized plant recommendation method according to another aspect of the present application.

The method of FIG. 9A may be performed by one or more computing devices, such as the plant management system of FIG. 1. 9B to 9D are detailed flowcharts of step S960.

Referring to Figure 9a, the artificial intelligence-based user-customized plant recommendation method includes receiving the user's account information and generating the user's account registration data (S901); Analyzing the user's space-related information to generate the user's space registration data (S910); Analyzing the user's plant-related information to generate the user's plant registration data (S920); Generating user profile data based on the user's account registration data, the user's space registration data, and the user's plant registration data (S930); and calculating a first type of customized plant for the user based on the user's profile data and the plant species profile data (S940).

The step (S940) includes transmitting a first type of customized plant for the user to the user's electronic device 100.

FIG. 10 is a schematic diagram of the process of generating spatial registration data of FIG. 9.

Referring to FIG. 10, the step of generating the user's spatial registration data (S910) includes receiving space-related information including the user's spatial captured image from the user's electronic device 100 (S911); Recognizing objects in the space captured image and calculating object characteristics for each recognized object (S912); Obtaining space type characteristics from the space-related information (S913); Calculating style characteristics of space based on object characteristics of the recognized object (S914); Calculating ventilation characteristics of the space based on the number of windows (S915); Receiving the illuminance value of the space measured from the electronic device 100 (S916); Obtaining the received illuminance measurement value as illuminance characteristics (S917); and generating space registration data of the user including one or more space characteristics of the user's space identifier, object characteristics placed in the space of the identifier, space type characteristics, style characteristics of the space, and illuminance characteristics (S918).

The user's spatial identifier is assigned in the process of creating the user's account registration data in step S901.

In step S911, the user's electronic device 100 may photograph a space to be registered and transmit the space-related information to the service server 200.

In some embodiments, step S912 may recognize an object expressed in the input image using a previously learned object recognition model. The object recognition operation using the object recognition model was described above with reference to the profile analysis module 233. Detailed explanation is omitted.

In step S912, the object characteristics may include one or more object information among the location of the object, the arrangement relationship between the objects, and the appearance of the object.

In step S913, the electronic device 100 may display a first input screen that prompts the user to input space type information. The electronic device 100 may transmit space-related information including input space type information to the service server 200.

In step S914, the style of the space in which the object is recognized may be recognized using a style recognition model learned in advance. It may be a machine learning model configured to predict the style characteristics of the space in which the input object is recognized by inferring the correlation between the input object characteristics of the space where the style is to be recognized and the style of the input space.

The style recognition model is learned using a training data set consisting of a plurality of training samples, and each training sample has training data and label data, and the training data is a sample object calculated from a sample space image taken of the sample space. It may also contain characteristics.

In step S915, the ventilation characteristic represents the ventilation level. In some embodiments, step S915 may calculate the ventilation level of the space based on at least two spatial characteristics of the ventilation space type, the number of windows, and the ventilation area of the windows.

Figure 11 is a schematic diagram of the process of generating plant registration data for the user of Figure 9A.

Referring to FIG. 11, the step of generating the user's plant registration data (S920) includes receiving the user's plant-related information, including user experience plant information and user-perceived environment information, from the user's electronic device 100. (S921); Extracting at least some information from the user's plant-related information as the user's plant characteristics (S922); Obtaining a spatial identifier of the user's account registration data (S923); and generating plant registration data including a space identifier of the user's registered space and the plant characteristics (S924).

In step S921, the electronic device 100 may display a second input screen that prompts the user to input plant-related information. The electronic device 100 may transmit the input plant-related information to the service server 200.

In step S921, the plant-related information includes the number of experienced plants.

The plant registration data in step S924 is used to generate user profile data to recommend customized plants for the user.

Referring again to FIG. 10, in step S930, the service server 200 integrates the information of items shared in the user's account registration data, the user's space registration data, and the user's plant registration data into a single item of information. You can also create user profile data.

Figure 12 is a schematic diagram of a process for generating a first type of customized plant for the user of Figure 9A.

Referring to FIG. 12, the step of calculating a first type of customized plant for the user based on the user's profile data and the profile data of the plant species (S940) includes matching plant search filtering conditions among the user's profile data. Retrieving one or more plants as a first type of customized plant for the user (S941); and transmitting the first type of customized plant list to the user's electronic device 100 (S943).

The profile analysis module 233 searches for one or more plants that match the plant search filtering conditions among the user's profile data. The filtering condition may include at least one characteristic item among characteristic items included in the user's profile data.

In some embodiments, step S940, before step S943, calculates a user's suitability score for the searched first type of customized plant, and generates a list of customized plants of the first type based on the suitability score. A step (S942) may be further included. In this case, the customized plant list of the first type is transmitted in step S943.

In step S942, the user's suitability score for each searched plant may be calculated based on the number of characteristics matching between the profile data of the plant species and the user's profile data. The fitness score is given a higher score as the number of matching characteristics increases.

In step S942, the searched plants are arranged so that plants with relatively high suitability scores in the customized plant list are provided to the user first.

Additionally, in some embodiments, in step S942, when the user's profile data includes a plurality of spatial registration data, the searched plants are classified into each registration space based on spatial characteristics in each spatial registration data, and , this could be creating a customized plant list for each registered space.

The steps (S910 to S940) have been described above with reference to the operation of the profile analysis module 233, and detailed description will be omitted.

Referring again to FIG. 9A, the artificial intelligence-based user-customized plant recommendation method includes generating user history data for each of a plurality of registered users (S950); and calculating a second type of customized plant for the user based on the history data (S960).

The history data includes one or more of user evaluation history data, activity history data, and purchase history data.

The step (S960) may be calculating a second type of customized plant for the user based on at least one of the user's evaluation history data, activity history data, and purchase history data.

The step (S960) includes transmitting a second type of customized plant for the user to the user's electronic device 100.

In certain embodiments, the step S950 may include receiving the user's evaluation score for the first type of customized plant calculated in step S940 and recommended for the user, and receiving the recommended first type of customized plant and It includes a step (S9510) of generating evaluation history data including evaluation scores.

Figure 13 shows a screen for entering an evaluation score according to an embodiment of the present application.

Referring to FIG. 13, an evaluation score indicating satisfaction with the recommendation result is input through the electronic device 100 of the user who has been recommended the first type of customized plant calculated in step S940, and the value of the input evaluation score is And a first type of customized plant to be evaluated may be received from the electronic device 100 and evaluation history data including the first type of customized plant and an evaluation score may be generated (S9510). In some embodiments, the evaluation score may be a subjective value about the user's feeling of actually growing the recommended first type of customized plant.

Referring to FIG. 9B, after the step (S9510), the step (S960) includes a first type of customized plant and A plurality of factors constituting the correlation between evaluations of recommendations are analyzed to calculate each user's recommendation score for each first type of customized plant, based on the user's recommendation score for each first type of customized plant. This includes calculating a second type of customized plant (S9610).

In the step (S9610), at least some of the first type of customized plants whose recommendation scores are equal to or higher than a preset recommendation threshold score may be calculated as new second type of customized plants.

Additionally, in some embodiments, the step (S960) further includes a step (S9620) of generating a list of customized plants of the second type based on the recommendation scores of each of the customized plants of the second type calculated in step (S9610). It may also be included. The second type of customized plants are arranged in the order in which the second type of customized plant with a high recommendation score is provided to the user first.

In some embodiments, when two or more customized plants in the second type of customized plant list have the same recommendation score, the two or more customized plants in the second type of customized plant list may be arranged in order according to the popularity of the plants. It may be possible.

The second type of customized plant list generated in step S9620 may be transmitted to the electronic device 100.

Referring again to FIG. 9A, the step (S950) may further include a step (S9530) of generating activity history data of the user when an activity event of the user occurs. The activity event includes a plant content production event and/or a plant content consumption event.

The step (S9530) includes generating corresponding plant content consumption history data when a plant content consumption event occurs (S9531); and/or generating plant content production history data when a plant content production event occurs (S9532).

In step S9530, the occurrence of a plant content consumption event may correspond to receiving a request related to activity content received from the user's electronic device 100. The occurrence of the plant content production event in the step (S9530) may correspond to producing the corresponding plant content according to the production request. Activity history data may then be generated when such a request is received.

FIG. 14 is a schematic diagram of a process for generating user interest record data based on content consumption history data according to an embodiment of the present application, and FIG. 15 is a schematic diagram of a user corresponding to a consumption event according to an embodiment of the present application. Show the request.

Referring to FIG. 14, when a consumption event occurs through the user's electronic device 100, consumption history data is generated. Plant content consumption history data includes information about the user who consumed it and activity details.

For example, as shown in Figure 15, the activity contents included in the plant content consumption history data include keyword search on the plant guide or gardening blog search page, plant guide content inquiry, gardening blog query, bookmark registration of the viewed content, and feed content. It may include one or more of the following activities: interaction, adding a plant to the list of plants I have grown, or adding a plant to the list of plants I have grown.

Figure 16 is a schematic diagram of a process for generating user interest record data based on content production history data, according to an embodiment of the present application.

Referring to FIG. 16, when a production event occurs through the user's electronic device 100, production history data is generated. Plant content production history data includes user information and activity content.

The activity content included in the plant content production history data may include one or more of the following: chat creation, post creation, plant guide or gardening blog creation, and feed content creation.

Then, the step (S960) may further include a step (S9630) of calculating a second type of customized plant for the user based on the user's activity history data.

Referring to FIG. 9C, in the step of calculating a second type of customized plant for the user based on the user's activity history data (S9630), keywords of interest of the user are extracted from plant content production history data and consumption history data. Step (S9631); Calculating the user's interest score based on the extracted interest keywords (S9632); Generating user interest record data including the interest score by associating the corresponding activity history data and interest keywords (S9633); forming a plurality of clusters consisting of some users with similar interest tendencies among the plurality of users based on interest record data of each of the plurality of users (S9634); And it may also include calculating a second type of customized plant for the user based on keywords of interest related to the group to which the user belongs (S9635).

In some embodiments, in the step of calculating the interest score (S9632), the history analysis module 235 may assign an interest score to an interest keyword extracted from consumption history data using a pre-stored interest score table. .

In addition, the step of calculating the interest score (S9632) determines the local frequency in which the extracted keyword of interest appears in the content produced by the user from which the keyword of interest was extracted and the keyword of interest extracted from the entire produced content of a plurality of users. The user's interest score may be assigned to the keyword of interest based on the global frequency of appearance, and the calculated interest score may be assigned to the corresponding produced content. The global frequency is the frequency of appearance in the entire range of individually produced content of all users over the entire period or a certain period of time on the system.

The interest record data generated in step S9633 may be stored in the user interest record DB 215.

The interest record data used in step S9634 may be retrieved from the user interest record DB 215.

In the above step (S9634), after the interest record data of each user is coordinated into vector points, a plurality of clusters may be formed from the interest record data sets of a plurality of users using a preset clustering method.

In the above step (S9635), plants matching a plurality of keywords of interest related to the community to which the user belongs may be calculated as a second type of customized plant.

Additionally, in some embodiments, the step (S960) further includes a step (S9640) of generating a list of customized plants of the second type based on the interest score of each customized plant of the second type calculated in step (S9630). It may also be included.

The second type of customized plant list generated in step S9640 may be transmitted to the electronic device 100.

Additionally, in some embodiments, the step of transmitting the second type of customized plant for the user to the user's electronic device 100 may include selecting the second type of customized plant found through matching based on the interest score of the user. The arranged second type of customized plant list may be provided to the user's electronic device 100. The second type of customized plants are arranged in the order in which the second type of customized plant with a high interest score is provided to the user first.

In some embodiments, if two or more customized plants in the second type of customized plant list have the same interest score, the two or more customized plants in the second type of customized plant list will be arranged in order according to the popularity of the plants. It may be possible.

In some embodiments, the step (S9635) is to create a second type of customized plant for the user based on only the remaining interest keywords that are not included in the user's interest record data among the plurality of interest keywords related to the cluster to which the user belongs. It may be calculated.

Then, the step (S9640) may be to generate a list of customized plants of the second type consisting of the customized plants of the second type based only on the remaining keywords of interest.

Referring again to FIG. 9A, the step (S950) may further include a step (S9550) of generating purchase history data for the user when a plant purchase event occurs.

In some embodiments, occurrence of the purchase event in step S9550 may correspond to receiving a purchase request from the device 100. Although it is important whether the payment has been completed and the purchase has actually been completed, it is more important to recommend a customized plant whether the user is interested enough in the plant to want to purchase it.

In addition, the step (S960) may further include a step (S9650) of calculating a second type of customized plant for the user based on the user's purchase history data.

Referring to FIG. 9D, the step (S9650) may be calculating a second type of customized plant for the user from the user's purchase history data using a pre-learned purchase prediction model.

The purchase prediction model is a machine learning model configured to infer the correlation between a plurality of characteristics included in user-related data and the plants the user purchases, and to custom-predict plants to be purchased by the user in an input data set containing a plurality of input characteristics. It may be possible.

The purchase prediction model is learned using a training data set consisting of a plurality of training samples. Training samples are user-specific. Each training sample in the set includes training data and label data for each sample user.

In the case of a user who has purchase history including purchased plants, the label data may have a value indicating the purchased plant (eg, a plant identifier value). In some embodiments, the label data may have a value representing the user's experienced plants included in the user's registered plant data in the user's profile data in the case of a user without purchase history (e.g., a first-time registered user).

The training data is formed based on at least one user's data among profile data of each user, activity history data of the user, and purchase history data of the user.

When each training data (eg, a plurality of characteristic items) is input, the purchase prediction model calculates a result value (ie, a predicted value) by processing the training data with internal parameters. The parameters of the purchase prediction model are learned so that the error between the calculation result value and the value of the label data (i.e., the actual value) is further reduced. The learned purchase prediction model has parameters that minimize the error.

In some embodiments, the purchase prediction model may determine one or more plants whose calculated purchase probability for each plant is higher than a preset threshold probability as the second type of customized plant. The output data of the purchase prediction model may include plant information of the determined second type of customized plant and a purchase probability value calculated for the decision.

Additionally, in some embodiments, step S960 further includes generating a list of customized plants of the second type based on the purchase probability of each customized plant of the second type calculated in step S9650. It may also be included. The second type of customized plants are arranged in the order in which the second type of customized plant with a high purchase probability is provided to the user first.

In some embodiments, when two or more plants among the plurality of selected customized plants of the second type have the same purchase probability, the two or more plants in the list of customized plants of the second type are arranged in order according to the popularity of each plant. It could be.

The second type of customized plant list generated in step S9660 may be transmitted to the electronic device 100.

In this way, individual second types of customized plants based on respective history data may be recommended to the user in steps S9620, S9640, and S9660, respectively.

In some embodiments, the step (S960) includes a second type of customized plant calculated based on the evaluation history data of steps (S9610 and S9620), and a second type of customized plant calculated based on the activity history data of steps (S9630 and S9640). 2 types of customized plants, when a second type of customized plant is calculated based on the purchase history data of steps (S9650, S9660), plants including some or all of each second type of customized plant are selected as final customized plants. A step of calculating with plants (S9670) may be further included.

In some embodiments, step S9670 may involve generating a recommendation result screen with a multi-layer structure and transmitting the recommendation result screen to the user's electronic device 100.

Each layer of multiple layers includes recommendation results for each recommendation method. The recommendation result screen may include multiple sub-screens corresponding to each of the multiple layers.

In some other embodiments, step S9670 may generate a multi-customized plant list by integrating the three recommendation results. In this case, plants included repeatedly in each list are included as a single item in the single customized plant list.

The steps (S950 to S960) have been described above with reference to the operation of the history analysis module 235, and detailed descriptions will be omitted.

When implementing embodiments of the present invention using hardware, application specific integrated circuits (ASICs) or digital signal processors (DSPs), digital signal processing devices (DSPDs), or programmable PLDs (PLDs) configured to perform the embodiments of the present application. logic devices), field programmable gate arrays (FPGAs), etc. may be included in the components of the present application.

The artificial intelligence-based user-customized plant recommendation method according to the embodiments of the present application described above and the operation of the plant management system 1 that performs the same are at least partially implemented as a computer program and recorded on a computer-readable recording medium. It can be. For example, implemented with a program product comprised of a computer-readable medium containing program code, which can be executed by a processor to perform any or all steps, operations, or processes described.

The computer-readable recording medium includes all types of recording devices that store data that can be read by a computer. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. Additionally, computer-readable recording media may be distributed across computer systems connected to a network, and computer-readable codes may be stored and executed in a distributed manner. Additionally, functional programs, codes, and code segments for implementing this embodiment can be easily understood by those skilled in the art to which this embodiment belongs.

The present invention discussed above has been described with reference to the embodiments shown in the drawings, but these are merely illustrative examples, and those skilled in the art will understand that various modifications and modifications of the embodiments are possible therefrom. However, such modifications should be considered within the technical protection scope of the present invention. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the attached patent claims.

Claims (17)

In an artificial intelligence-based user-customized plant recommendation method performed by one or more computing devices,
Receiving the user's account information and generating the user's account registration data;
Generating user space registration data by analyzing space-related information of the user;
Analyzing the user's plant-related information to generate the user's plant registration data;
generating profile data for the user based on the user's account registration data, the user's space registration data, and the user's plant registration data; and
Comprising: calculating a first type of customized plant for the user based on the user's profile data and the plant species profile data,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 1, wherein the step of analyzing the user's space-related information to generate the user's space registration data comprises:
Receiving space-related information including a user's space capture image from the user's electronic device;
Recognizing objects in the spatial captured image and calculating object characteristics for each recognized object;
Obtaining space type characteristics from the space-related information;
calculating style characteristics of space based on object characteristics of the recognized object;
Calculating ventilation characteristics of the space based on the number of windows;
Receiving the illuminance value of the space measured from the user's electronic device;
Obtaining the received illuminance measurement value as illuminance characteristics; and
Generating space registration data for the user including one or more space characteristics of the user's space identifier, object characteristics placed in the space of the identifier, space type characteristics, style characteristics of the space, and illuminance characteristics;
Characterized in that the user's spatial identifier is assigned in the step of creating the user's account registration data,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 2, wherein calculating the style characteristics comprises:
Using a pre-learned style recognition model, the style of the space where the object is recognized is recognized,
The style recognition model is a machine learning model configured to predict the style characteristics of the space in which the input object is recognized by inferring the correlation between the input object characteristics of the space where the style is to be recognized and the style of the input space,
The style recognition model is learned using a training data set consisting of a plurality of training samples, and each training sample has training data and label data, and the training data is a sample object calculated from a sample space image taken of the sample space. Contains characteristics,
The input object characteristics and sample object characteristics include one or more object information among the location of the object, the arrangement relationship of the objects, and the appearance of the object,
Artificial intelligence-based user-tailored plant recommendation method. According to claim 2,
The ventilation characteristic indicates a ventilation level with a higher value as ventilation becomes easier,
The step of calculating the ventilation characteristics of the space based on the number of windows is,
Characterized in calculating the ventilation level of the space based on at least two space characteristics of the ventilation space type, number of windows, and ventilation area of the windows,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 1, wherein calculating a first type of customized plant for the user based on the user's profile data and the plant species profile data comprises:
Searching for one or more plants matching the plant search filtering conditions among the user's profile data as a first type of customized plant for the user;
calculating a user's suitability score for the searched first type of customized plants, and generating a list of first types of customized plants based on the suitability scores; and
transmitting a first type of customized plant list to the user's electronic device;
The filtering condition includes at least one characteristic item among characteristic items included in the user's profile data,
The user's suitability score for the searched plant is calculated based on the number of characteristics matching between the profile data of the plant species and the user's profile data,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 5, wherein generating a customized plant list of a first type based on the fitness score comprises:
When the user's profile data includes a plurality of spatial registration data, the searched plants are classified by each registered space based on the spatial characteristics in each spatial registration data, and a customized plant list is generated for each registered space. doing,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 6,
receiving the user's evaluation score for a first type of customized plant calculated and recommended for the user and generating evaluation history data including the recommended first type of customized plant and the evaluation score; and
Based on the evaluation history of a plurality of users for whom at least one type of first type of customized plant is recommended for each user, a plurality of factors constituting the correlation between the evaluation of the first type of customized plant and the recommendation are analyzed for each first type. Calculating each user's recommendation score for each first type of customized plant, and calculating a second type of customized plant based on the user's recommendation score for each first type of customized plant,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 1,
Generating activity history data including at least one of plant content production history data and plant content consumption history data when one or more of the user's plant content production event and plant content consumption event occurs; and
Further comprising calculating a second type of customized plant for the user based on the user's activity history data,
The step of calculating a second type of customized plant for the user based on the user's activity history data,
Extracting keywords of user interest from plant content production history data and consumption history data;
calculating a user's interest score based on the extracted interest keywords;
generating user interest record data including the interest score by associating corresponding activity history data and interest keywords;
forming a plurality of clusters consisting of some users with similar interest tendencies among the plurality of users based on interest record data of each of the plurality of users; and
Comprising a step of calculating a second type of customized plant for the user based on keywords of interest related to the community to which the user belongs,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 8, wherein the step of calculating a second type of customized plant for the user based on keywords of interest related to the group to which the user belongs comprises:
Characterized in calculating a second type of customized plant for the user based only on the remaining interest keywords that are not included in the user's interest record data among a plurality of interest keywords related to the cluster to which the user belongs,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 8, wherein calculating the user's interest score based on the extracted interest keywords comprises:
assigning interest scores to keywords of interest extracted from consumption history data using a pre-stored interest score table; and
A user for a keyword of interest based on the local frequency of the extracted keyword of interest appearing in the content produced by the user from which the keyword of interest was extracted and the global frequency of the keyword of interest appearing in the entire content produced by a plurality of users. A step of assigning an interest score and assigning the calculated interest score to the corresponding production content,
The global frequency is characterized in that the frequency of appearance in the entire range of individually produced content of all users during the entire period or a certain period of time on the system,
Artificial intelligence-based user-tailored plant recommendation method. The method of claim 1,
When a user's plant purchase event occurs, generating purchase history data for the user; and
Further comprising calculating a second type of customized plant for the user from the user's purchase history data using a pre-trained purchase prediction model,
The purchase prediction model is a machine learning model configured to infer the correlation between a plurality of characteristics included in user-related data and the plants the user purchases, and to custom-predict plants to be purchased by the user in an input data set containing a plurality of input characteristics. ,
The purchase prediction model is learned using a training data set consisting of a plurality of training samples, and each training sample in the set includes training data and label data for each sample user,
For a user who has a purchase history including purchased plants, the label data has a value indicating the purchased plant,
For a user without purchase history, the label data has a value representing the user's experienced plants included in the user's registered plant data in the user's profile data,
The training data is characterized in that it is formed based on at least one user data among the profile data of each user, the activity history data of the user, and the purchase history data of the user.
Artificial intelligence-based user-tailored plant recommendation method. A computer-readable recording medium recording a program consisting of instructions for performing an artificial intelligence-based user-customized plant recommendation method according to any one of claims 1 to 11. In the plant management system,
In order to sign up for the plant management system, the user's account information is entered, and the plant management system provides the user's space-related information and plant-related information so that it can provide plant-related services by considering the user's space and the plants that have been grown or are growing. a user's electronic device configured to input and receive customized plant information for the user calculated using the user's profile data and the plant's profile data; and
Receive the user's account information to generate the user's account registration data, analyze the user's space-related information to generate the user's space registration data, analyze the user's plant-related information to generate the user's plant registration data, generate profile data for the user based on the user's account registration data, the user's space registration data, and the user's plant registration data, and create a first type for the user based on the user's profile data and the plant species' profile data; Comprising a service server configured to produce customized plants of,
Plant management system. The method of claim 13,
The electronic device is,
further configured to input the user's evaluation score for the received first type of customized plant and transmit it to the service server;
The service server is,
receiving the user's evaluation score for a first type of customized plant calculated and recommended for the user to generate evaluation history data including the recommended first type of customized plant and the evaluation score; and
Based on the evaluation history of a plurality of users for whom at least one type of first type of customized plant is recommended for each user, a plurality of factors constituting the correlation between the evaluation of the first type of customized plant and the recommendation are analyzed for each first type. Characterized in that, further configured to calculate each user's recommendation score for each first type of customized plant, and calculate a second type of customized plant based on the user's recommendation score for each first type of customized plant,
Plant management system. The method of claim 13,
The electronic device is,
Receive user input for consuming plant content or producing plant content within the system, transmit a request to execute a service operation according to the user input to the service server, and receive processing results for the service operation according to the request. composed,
The service server is,
Upon receiving the request, generate activity history data including one or more of plant content production history data and plant content consumption history data, and
Characterized in that it is further configured to calculate a second type of customized plant for the user based on the user's activity history data,
Plant management system. The method of claim 13,
The electronic device is,
Further configured to receive a user input for purchasing a plant through a commerce system included in or connected to the system, transmit a purchase request according to the user input to the commerce system, and receive a purchase processing result according to the request. become,
The service server is,
When receiving the above purchase request, create purchase history data for the user, and
Characterized in that it is further configured to calculate a second type of customized plant for the user from the user's purchase history data using a pre-trained purchase prediction model,
Plant management system. According to any one of claims 14 to 16,
The service server is,
When a second type of customized plant calculated based on evaluation history data, a second type of customized plant calculated based on activity history data, and a second type of customized plant calculated based on purchase history data are calculated, respectively, further configured to produce plants including some or all of the second type of customized plants as final customized plants and transmit them to the electronic device;
The final customized plant calculation result is a recommendation result screen composed of a multi-layer structure, and the recommendation result screen includes multiple sub-screens including recommendation results for each recommendation method,
The electronic device is,
Display the above recommendation result screen, and
Characterized in that it is further configured to move the sub screen by scrolling on the recommendation result screen,
Plant management system.

Images