WO2019139460A2

WO2019139460A2 - Artificial intelligence apparatus

Info

Publication number: WO2019139460A2
Application number: PCT/KR2019/001449
Authority: WO
Inventors: 이준원; 정의경; 정해융
Original assignee: 엘지전자 주식회사
Priority date: 2019-02-01
Filing date: 2019-02-01
Publication date: 2019-07-18
Also published as: WO2019139460A3; KR20210089783A

Abstract

An Artificial intelligence apparatus is disclosed. An artificial intelligence apparatus according to an embodiment of the present invention comprises: a communication unit for receiving use data from at least one home appliance placed in a home; an analysis module for acquiring a use pattern of the home on the basis of the use data; a learning model for outputting a resulting value obtained by classifying homes in a particular area according to specific elements of a type, by using the use patterns acquired from the homes in the particular area; and a processor for providing distribution data of the specific elements of the types of the homes in the particular area to a business operator by using the resulting value, wherein the learning model is a neural network model having been trained using the use patterns of the multiple homes.

Description

Artificial Intelligence Device

The present invention relates to an artificial intelligence apparatus that divides the assumptions in a specific region by use of usage data of a home appliance and provides distribution data on detailed types of assumptions to a business entity.

Artificial intelligence is a field of computer engineering and information technology that researches how computers can do thinking, learning, and self-development that can be done by human intelligence. To be imitable.

In addition, artificial intelligence does not exist in its own right, but has many direct and indirect links with other areas of computer science. Especially, in the field of information technology, the artificial intelligence has been introduced in many fields and it has been actively attempted to utilize it to solve problems in the field.

On the other hand, techniques for recognizing and learning the situation using artificial intelligence, providing information desired by the user in a desired form, or performing a desired action or function are being actively studied. Recognition models based on artificial intelligence technology grow as recognition models are updated by recognizing and learning user characteristics of new environments or artificial intelligence devices.

Meanwhile, as the spread of home appliances based on artificial intelligence is expanded, various data can be obtained from various sensors provided in the home appliances. The manufacturer of the home appliance can utilize the data obtained from the home appliance to improve the product or develop new products. On the other hand, the manufacturer can also worry about the business utilization of the data obtained from the home appliance.

The problem to be solved by the present invention is to provide an artificial intelligence device capable of generating business revenues by classifying the local households according to the detailed types.

Another object to be solved by the present invention is to provide an artificial intelligence device capable of generating additional profit by calculating sales strategy information using results obtained by dividing households in detail by type.

An artificial intelligence device according to an embodiment of the present invention receives usage data from at least one home appliance disposed in a home to obtain a usage pattern, inputs usage patterns to a learning model, , And distributes the distribution data that classifies the assumptions within a specific region according to the type of detail elements.

The artificial intelligence device according to another embodiment of the present invention acquires sales strategy information based on distribution data on the details of types of assumptions in a specific area and transmits the acquired sales strategy information to the business entity.

According to the present invention, it is possible to generate business revenues by selling furniture distribution type data to a business nearby, for example, a convenience store, a bakery, a coffee shop, and a delivery service.

In addition, it is possible to elaborate the sale of goods and the provision of services, such as selling and displaying customized foods, goods packages and services corresponding to furniture types more efficiently.

1 is a conceptual diagram of a system including an artificial intelligence device according to an embodiment of the present invention.

2 is a schematic block diagram showing a control configuration of an artificial intelligence apparatus according to an embodiment of the present invention.

3 is a flowchart illustrating an operation of the artificial intelligence apparatus according to an embodiment of the present invention.

4 is a diagram for explaining a method of generating a learning model according to an embodiment of the present invention.

5 is a diagram for explaining a method of acquiring distribution data using usage data received from a home appliance according to an embodiment of the present invention.

FIG. 6 is a diagram showing a distribution diagram generated by using information obtained by dividing assumptions within a specific region according to detailed elements of a type.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, wherein like reference numerals are used to designate identical or similar elements, and redundant description thereof will be omitted. The suffix "module" and " part "for the components used in the following description are given or mixed in consideration of ease of specification, and do not have their own meaning or role. In the following description of the embodiments of the present invention, a detailed description of related arts will be omitted when it is determined that the gist of the embodiments disclosed herein may be blurred. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are intended to provide further explanation of the invention as claimed. , &Lt; / RTI > equivalents, and alternatives.

Terms including ordinals, such as first, second, etc., may be used to describe various elements, but the elements are not limited to these terms. The terms are used only for the purpose of distinguishing one component from another.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the terms "comprises", "having", and the like are used to specify that a feature, a number, a step, an operation, an element, a component, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

Prior to a specific description of the present invention, a brief description of artificial intelligence (AI) will be given.

Artificial intelligence (AI) is a field of computer engineering and information technology that researches how computers can do thinking, learning, and self-development that can be done by human intelligence. It means to be able to imitate behavior.

Machine learning is a field of artificial intelligence, a field of study that gives computers the ability to learn without explicit programming.

Specifically, machine learning is a technology for studying and constructing a system and algorithms for learning, predicting, and improving its performance based on empirical data. Machine learning algorithms are not designed to perform strictly defined static program instructions but rather take the form of building a specific model to derive predictions or decisions based on input data.

The term 'machine learning' can be used interchangeably with the term 'machine learning'.

Many machine learning algorithms have been developed on how to classify data in machine learning. Decision Tree, Bayesian network, support vector machine (SVM), and Artificial Neural Network (ANN) are typical examples.

Decision trees are an analytical method that classifies and predicts decision rules by plotting them in a tree structure.

The Bayesian network is a graphical representation of the stochastic relations (conditional independence) between multiple variables. Bayesian networks are suitable for data mining through unsupervised learning.

The support vector machine is a model of supervised learning for pattern recognition and data analysis, mainly used for classification and regression analysis.

An artificial neural network is an information processing system in which a number of neurons, called nodes or processing elements, are connected in the form of a layer structure, modeling the connection between neurons and the operating principle of biological neurons.

An artificial neural network is a model used in machine learning. It is a statistical learning algorithm that is inspired by neural networks of biology (especially the brain of the animal's central nervous system) in machine learning and cognitive science.

Specifically, an artificial neural network can refer to an entire model having a problem-solving ability by changing the synaptic bond strength through artificial neurons (nodes) forming a network by the synapse combination.

The term artificial neural network can be used interchangeably with the term neural network.

The artificial neural network may comprise a plurality of layers, each of which may comprise a plurality of neurons. Artificial neural networks can also include synapses that connect neurons to neurons.

The artificial neural network generally takes the following three factors: (1) a connection pattern between neurons in different layers, (2) a learning process to update the weights of connections, (3) &Lt; / RTI >

The artificial neural network may include network models such as DNN (Deep Neural Network), RNN (Recurrent Neural Network), BRDNN (Bidirectional Recurrent Deep Neural Network), MLP (Multilayer Perceptron), CNN (Convolutional Neural Network) , But is not limited thereto.

The term " layer " may be used interchangeably with the term " layer " herein.

Artificial neural networks are classified into single-layer neural networks and multi-layer neural networks according to the number of layers.

A general neural network consists of an input layer and an output layer.

In addition, a general multilayer neural network consists of an input layer, a hidden layer, and an output layer.

The input layer is a layer that accepts external data. The number of neurons in the input layer is equal to the number of input variables. The hidden layer is located between the input layer and the output layer, receives signals from the input layer, do. The output layer receives a signal from the hidden layer and outputs it to the outside. The input signal between neurons is multiplied by the respective connection strengths between 0 and 1, and then summed. If the sum is larger than the threshold value of the neuron, the neuron is activated and implemented as an output value through the activation function.

On the other hand, the in-depth neural network including a plurality of hidden layers between the input layer and the output layer may be a representative artificial neural network that implements deep learning, which is one type of machine learning technology.

Meanwhile, the term 'deep learning' can be used in combination with the term 'in-depth learning'.

Artificial neural networks can be trained using training data. Here, learning means a process of determining a parameter of an artificial neural network using learning data in order to achieve classification, classification, regression, clustering, and the like of input data . Representative examples of parameters of an artificial neural network include a weight applied to a synapse or a bias applied to a neuron.

The artificial neural network learned by the training data can sort or group the input data according to the pattern of the input data.

Meanwhile, the artificial neural network learned using the training data can be referred to as a trained model in this specification.

The following explains the learning method of artificial neural network.

The learning methods of artificial neural networks can be roughly divided into instructional learning, non-instructional learning, semi-supervised learning, and reinforcement learning.

Map learning is a method of machine learning to derive a function from training data.

Among such inferred functions, outputting consecutive values is called regression, and classifying is performed by predicting and outputting a class of an input vector.

In map learning, an artificial neural network is learned in a state where a label for training data is given.

Here, the label may mean a correct answer (or a result value) to be inferred by the artificial neural network when the training data is input to the artificial neural network.

In this specification, the correct answer (or result value) to be inferred by the artificial neural network when training data is input is called labeling or labeling data.

In this specification, labeling of training data for training of artificial neural networks is labeled as labeling data for training data.

In this case, the training data and the label corresponding to the training data) constitute one training set, and the artificial neural network may be input in the form of a training set.

On the other hand, training data represents a plurality of features, and labeling labels on training data may mean that the features represented by the training data are labeled. In this case, the training data can represent the characteristics of the input object in a vector form.

Artificial neural networks can derive a function of the association between training data and labeling data, using training data and labeling data. In addition, artificial neural networks can determine (optimize) parameters of artificial neural networks by evaluating the inferred functions.

Non-instructional learning is a kind of machine learning, and there is no label for training data.

Specifically, the non-guidance learning may be a learning method in which the artificial neural network is learned so as to find and classify patterns in the training data itself, rather than association of labels corresponding to training data and training data.

Examples of non-instructional learning include clustering or Independent Component Analysis.

The term " clustering " as used herein can be used interchangeably with the term " clustering ".

Examples of artificial neural networks that use non-bipartite learning include Generative Adversarial Network (GAN) and Autoencoder (AE).

Generative hostile neural network is a machine learning method that improves performance by competing two different artificial intelligences, a generator and a discriminator.

In this case, the generator is a model for creating new data, and can generate new data based on the original data.

In addition, the discriminator is a model for recognizing patterns of data, and can discriminate authenticity of new data generated by the generator based on original data.

And the generator learns the data which is not deceived by the discriminator, and the discriminator learns the data from the generator by receiving it. Thus, the generator can evolve to make the discriminator as foolish as possible, and can evolve to distinguish between the original data of the discriminator and the data generated by the generator.

The auto encoder is a neural network that aims to reproduce the input itself as an output.

The auto encoder includes an input layer, a hidden layer, and an output layer, and the input data passes through the input layer and enters the concealment layer.

In this case, since the number of nodes in the hidden layer is smaller than the number of nodes in the input layer, the dimension of the data is reduced, and compression or encoding is performed.

Also, the data output from the hidden layer enters the output layer. In this case, since the number of nodes in the output layer is larger than the number of nodes in the hidden layer, the dimension of the data is increased, so that decompression or decoding is performed.

On the other hand, the auto encoder adjusts the connection strength of neurons through learning, so that the input data is expressed as hidden layer data. In the hidden layer, the information can be represented by a smaller number of neurons than the input layer. The fact that the input data can be reproduced as output can mean that the hidden layer has discovered and expressed a hidden pattern from the input data.

Semi-instructional learning is a kind of machine learning, which can be a learning method that uses labeled training data and training data that is not labeled.

One of the techniques of semi-instruction learning is to infer labels of unlabeled training data and then perform learning using inferred labels. This technique is useful when labeling cost is high. .

Reinforcement learning is the theory that if an agent is able to determine what action each agent should do at any given moment, then the best path to experience can be found without data.

Reinforcement learning can be performed mainly by the Markov Decision Process (MDP).

The Markov decision process is described as follows: First, the agent is given an environment in which information necessary for the next action is configured. Second, the agent defines how the agent behaves in the environment. Third, and the penalty will be given to what if you fail to do what you can do, and the fourth time you will experience the best policy until the future compensation reaches its peak.

The structure of the artificial neural network is specified by the structure of a model, an activation function, a loss function or a cost function, a learning algorithm, an optimization algorithm, etc., and a hyperparameter is pre- And then the model parameter is set through learning to specify the content.

For example, factors for determining the structure of the artificial neural network may include the number of hidden layers, the number of hidden nodes included in each hidden layer, an input feature vector, and a target feature vector.

The hyper parameters include various parameters that must be initially set for learning, such as initial values of the model parameters. The model parameters include various parameters to be determined through learning.

For example, the hyperparameter may include a node initial weight value, an initial node deflection value, a mini-batch size, a learning iteration count, a learning rate, and the like. The model parameters may include weight between nodes, deviation between nodes, and the like.

The loss function can be used as an index to determine the optimal model parameters in the learning process of artificial neural networks. Learning in artificial neural networks means manipulating the model parameters to reduce the loss function, and the goal of learning is to determine the model parameters that minimize the loss function.

The loss function may mainly use a mean squared error (MSE) or a cross entropy error (CEE), and the present invention is not limited thereto.

The cross entropy error can be used when the correct label is one-hot encoded. The raw hot encoding is an encoding method in which the correct label value is set to 1 only for the neuron corresponding to the correct answer and the correct label value is set to 0 for the neuron not the correct answer.

In the machine learning or the deep learning, a learning optimization algorithm can be used to minimize the loss function. The learning optimization algorithm includes GD (Gradient Descent), Stochastic Gradient Descent (SGD), Momentum ), NAG (Nesterov Accelerate Gradient), Adagrad, AdaDelta, RMSProp, Adam, and Nadam.

The slope descent method is a technique of adjusting the model parameters in the direction of decreasing the loss function value in consideration of the slope of the loss function in the current state.

The direction in which the model parameter is adjusted is referred to as a step direction, and the size to be adjusted is referred to as a step size (size).

At this time, the step size may mean the learning rate.

The slope descent method can be updated by partially differentiating the loss function with each model parameter to obtain the slope and changing the learning parameters in the gradient direction in which the model parameters are obtained.

The stochastic descent method divides training data into mini - batches, and performs a slope descent method for each mini - batch to increase the frequency of slope descent.

Adagrad, AdaDelta, and RMSProp are techniques that improve the optimization accuracy by adjusting the step size in SGD. In SGD, momentum and NAG are techniques that increase the optimization accuracy by adjusting the step direction. Adam combines momentum and RMSProp to adjust the step size and step direction to improve optimization accuracy. Nadam is a technique that combines NAG and RMSProp to improve the optimization accuracy by adjusting the step size and step direction.

The learning speed and accuracy of the artificial neural network are characterized not only by the structure of the artificial neural network and the kinds of learning optimization algorithms, but also by the hyper parameters. Therefore, in order to obtain a good learning model, it is important not only to determine a proper artificial neural network structure and a learning algorithm, but also to set an appropriate hyperparameter.

Typically, the hyperparameter is experimentally set to various values, and the artificial neural network is learned, and the learning result is set to an optimal value that provides stable learning speed and accuracy.

In the following drawings, the business entity 200 is a terminal (or a company) operated by a person (or a company) who provides commerce-related information to a person (or a company) who provides a commerce service to a home or a person Or server.

Meanwhile, the artificial intelligence device 10 may be a server operated by the manufacturer of the home appliance 20, but it is not necessarily the case.

The artificial intelligence apparatus 10 may be connected via a network with at least one home appliance disposed in a plurality of

assumptions

2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200.

The plurality of

assumptions

2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200 can be located in various regions.

For example, the first group of

assumptions

2200, 2300, 2400, and 2500 may be stored in a first region, the second group of

assumptions

2600, 2700, 2800, and 2900 may be stored in a second region, , 3100, 3200) may be located in the third area.

Here, the area may refer to administrative districts such as Eup, Myeon, Dong, Ward, City, and Province, but it is not limited to this and can be classified by various criteria.

The artificial intelligence device 10 may receive usage data from at least one home appliance disposed within a plurality of

assumptions

2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200.

The artificial intelligence device 10 may receive usage data for each of the at least one home appliances 2210 from at least one home appliance 2210 disposed in the first home 2200. [

On the other hand, the artificial intelligence apparatus 10 can acquire a usage pattern in the home based on usage data.

Also, the artificial intelligence device 10 can output the results obtained by dividing the assumptions in the specific region according to the detailed elements of the type, using the usage pattern obtained from the assumptions within the specific region.

The artificial intelligence apparatus 10 may then use the resultant values to obtain distribution data on the types of assumptions in the specific area and to provide the obtained distribution data to the business entity 200. [

The home appliance 2210 corresponds to a device that is disposed in a specific space (e.g., in a home) and provides a unique function. As shown in FIG. 1, the home appliance 2210 may include an air conditioner, a robot cleaner, a refrigerator, a washing machine, and the like.

Referring to FIG. 2, the artificial intelligence apparatus 100 may include a processor 110, a communication unit 120, a memory 130, an analysis module 140, and a learning model 150.

The processor 110 may control the overall operation of the configurations contained in the artificial intelligence device 10. For example, the processor 11 may include at least one integrated circuit, an application specific integrated circuit (ASIC), a CPU, an application processor (AP), a microcomputer, or the like.

The communication unit 120 may include at least one communication module for connecting the artificial intelligence device 100 with at least one home appliance 20 through a network. Each of the at least one communication module may support different wired / wireless communication schemes. The artificial intelligence apparatus 100 may receive event information from at least one home appliance 2210 through the communication unit 120. [

In addition, the communication unit 120 may connect the artificial intelligence apparatus 100 to a plurality of home appliances and businesses 200 using at least one communication module.

The memory 130 may store control data for controlling the operation of the components included in the artificial intelligent device 100 or information or data received through the communication unit 120 or the like.

In particular, the memory 130 may store program data for execution of the analysis module 140 and the learning model 150, and data and algorithms related to the operation of each

module

140, 150.

For example, the usage pattern generated based on the usage data may be stored in the memory 130. [

The analysis module 140 may obtain a usage pattern in the home based on the usage data received from the home appliance.

The learning model 150 may output a result obtained by dividing the assumptions within a specific region by the detail elements of the type using the usage pattern obtained from the assumptions within the specific region.

Referring to FIG. 3, an artificial intelligence apparatus according to an embodiment of the present invention includes a step of generating a learning model by training a neural network using a plurality of assumption patterns (S310) The method comprising: receiving usage data from an appliance (S320); acquiring a usage pattern in the assumption (S330) based on the usage data; using the usage pattern obtained from assumptions in a specific area, (S340) of outputting a result obtained by dividing the assumptions in the region according to the detailed elements of the type (S340), and providing the distribution data on the type of the assumptions in the specific region to the business entity using the resultant value can do.

The step of generating a learning model (S310) by training a neural network using usage patterns of a plurality of assumptions will be described with reference to FIG.

Hereinafter, it is assumed that the learning model 530 is generated in the learning apparatus 1000 of the neural network. However, the following process can also be performed in the artificial intelligence apparatus 100. [

The learning apparatus 1000 of the neural network may include a data input unit 1010, a processor 1020, and a neural network 1030. [

The data input unit 1010 can receive input data. In this case, the data input unit 1010 may receive training data and may receive unprocessed data.

When the data input unit 1010 receives the raw data, the processor 1020 can preprocess the received data to generate training data that can be input to the neural network 1030. [

The neural network 1030 may be implemented in hardware, software, or a combination of hardware and software, and in the case where some or all of the neural network 1030 is implemented in software, (Not shown) included in the learning apparatus 1000 of FIG.

The processor 1020 may enter training data or a training set into the neural network 1030 to train the neural network 1030. [

The processor 1020 can determine (optimize) the parameters of the Artificial Neural Network (ANN) by repeatedly learning the Artificial Neural Network (ANN).

The artificial neural network whose parameters have been determined by learning using the training data can be referred to as a trained model 530 in this specification.

On the other hand, a trained model 530 can be used to infer results for new input data rather than training data

First, a method of training a neural network based on map learning will be described.

In the present invention, the usage pattern 410 may be used as training data, and the details of the type may be input to the neural network together with the usage pattern 410 as a label.

Specifically, the processor 1020 may train the neural network using a type of detail element corresponding to the usage pattern 410 and the usage pattern 410. [

The usage pattern 410 may include a hold state 411 of the home appliance and an operation state 412 of the home appliance.

For example, the holding state 411 of the home appliance indicates whether or not the refrigerator is held, whether or not the dishwasher is held, whether the air purifier is held, whether the microwave oven is held, whether the oven has a gas range, Or home appliances deployed in the home.

The operation state 412 of the home appliance may include the number of times the home appliance is used, the time when the home appliance is used, the time interval using the home appliance, the setting of the home appliance (for example, oven function or thawing function for a microwave oven, A set temperature, an operation mode, etc.).

The type can also be the number of household members, household members or income type.

If the type is the number of household members, the details of the type can include one person, two people, three people, four or more people, and multi-family households.

If the type is a household member, the detailed elements of the type include household composed of a couple, family composed of a couple and a child, gender of a member, household composed of members under 30 members, household composed of members of 60 members or more .

If the type is an income type, the details of the type may include household income, family income, etc.

On the other hand, usage pattern 410 may be a usage pattern of one or more home appliances in a particular home, and a sub-type of type corresponding to usage pattern 410 may be a household composition of the particular home.

On the other hand, the details of the type can be obtained, for example, by conducting surveys when registering a home appliance with a server.

Meanwhile, the processor 1020 may train the neural network using the sub-elements of the type corresponding to usage patterns and usage patterns of a plurality of households.

For example, the processor 1020 may be configured to train a neural network using the usage patterns of one or more home appliances in the first household and the details of the type of first furniture, The neural network can be trained using the details of the type of the second household.

Accordingly, various usage patterns and sub-types of elements can be input to the neural network 1030 as training data.

In this case, the processor 1020 may repeatedly train the neural network using map learning.

In this case, the neural network 1030 can derive a function of the usage pattern and the relation of the sub-elements of the type using the sub-elements of the type corresponding to the usage pattern and the usage pattern. The neural network 1030 may also determine (optimize) the parameters of the neural network 1030 through an evaluation of the inferred function.

Thus, the trained neural network can classify usage patterns according to the type of detail elements.

The following describes how to train a neural network based on non-instructional learning.

Non - instructional learning can be a learning method that learns artificial neural networks to find and classify patterns in training data itself. And clustering can be a process of clustering feature vectors obtained from artificial neural networks into finite clusters.

On the other hand, as a method of non-mapping learning, a k-means clustering algorithm can be used.

The k-mean clustering algorithm is one of the separate clustering algorithms, and it is an algorithm that groups given data into k clusters. Each cluster has one centroid. Each object is assigned to the nearest center, and individuals assigned to the same center are gathered to form a cluster. The user must set the number of clusters (k) in advance to execute the algorithm. And operates in a manner that minimizes dispersion of the distance difference with each cluster.

Therefore, the detailed elements of the type or type are determined according to the classification goals of the households, and the processor 1020 can train the neural network so that the usage patterns can be classified according to the determined types or the detailed elements of the types.

For example, in the case of a household with a classification target of 1 or a multinational household, the number of clusters is determined to be 2, and an initial parameter for classifying a single household and a multinational household can be set. When the processor 1020 inputs the usage pattern as training data, the neural network can set the parameters to minimize the variance of the distance between the usage patterns and the respective clusters.

The trained neural network can thus clustering usage patterns according to the type of detail elements.

On the other hand, a trained model 530 can be used to infer results for new input data rather than training data.

While a trained model 530 may include one or more learning models each corresponding to one or more types.

Specifically, the a trained model 530 may include a first learning model corresponding to the number of household members, a second learning model corresponding to the household member, and a third learning model corresponding to the income type .

Where the first learning model may be a trained neural network that classifies or clusters the families into subdivisions according to the number of household members.

And the second learning model may be a trained neural network that classifies or clusters the families into subdivisions of household members.

The third learning model may also be a trained neural network that classifies or clusters the families into subdivisions of the income type.

On the other hand, the learning model 530 may be mounted on the artificial intelligence apparatus 100. [

Meanwhile, the learning model 530 may be implemented by hardware, software, or a combination of hardware and software. When a part or the whole of the learning model is implemented by software, one or more commands constituting the learning model may be generated by the artificial intelligence apparatus 100 And may be stored in the memory 130.

The following will be described with reference to FIG. 5 in connection with S320, S330, S340, and S350.

Referring to FIG. 5, the processor 110 may receive usage data from at least one home appliance disposed in a plurality of households through the communication unit 120.

In this case, the analysis module 120 may obtain a usage pattern in the home based on the usage data.

Specifically, the usage data may include at least one of identification data of the home appliance and operation data of the home appliance.

Here, the identification data of the home appliance is for identifying the type of the home appliance, and the analysis module 120 can acquire the holding state of the home appliance based on the identification data.

For example, the analysis module 120 may use at least one home appliance in the home to obtain information that the home appliance has a washing machine, a refrigerator, and an air conditioner in a specific home.

In this manner, the analysis module 120 can acquire the holding state of a plurality of home appliances of the home.

Meanwhile, the operation data of the home appliance can be used to obtain the operation state of the home appliance.

Specifically, the operation data may be information related to the operation of the home appliance. More specifically, the operation data is transmitted to the home appliance through the home appliance using information sensed by one or more sensors provided in the home appliance, information received through the input device of the home appliance, information sensed by the sensor, And the setting information of the home appliance.

The analysis module 120 can obtain the operation state of the home appliance using the operation data.

For example, the analysis module 120 can use the data acquired from the refrigerator opening / closing sensor to acquire an operation state such as a time when the refrigerator is used or a number of times the refrigerator is used.

In this manner, the analysis module 120 can obtain the operating states of the home appliances of a plurality of households.

Meanwhile, in order to obtain the distribution data of the assumptions in the specific area, the processor 110 classifies the usage patterns obtained from the assumptions in the plurality of areas by region and obtains the usage patterns obtained from the assumptions 530 in the specific one of the plurality of regions The usage pattern can be entered into the learning model.

On the other hand, the learning model 150 may output the results obtained by dividing the assumptions in the specific region according to the detailed elements of the type, using the usage pattern obtained from the assumptions 530 in the specific region.

Specifically, when the learning model 150 is generated in a manner of learning the learning model 150, the learning model 150 obtains the usage pattern obtained from the assumptions 530 in the specific area from the assumptions 530 in the specific area The result of classifying the obtained usage patterns according to the detailed elements of the type can be output.

For example, let's assume that the type is the number of household members, and the type of detail is more than one, two, and three. In this case, the first learning model corresponding to the number of household members can output the results obtained by classifying the usage patterns obtained from the assumptions 530 in the specific area into one family, two family, and three or more families have.

For another example, let's assume that the type is the number of household members, the sub-element of type 1, and the multinational household. In this case, the first learning model corresponding to the number of the household members can output the result of classifying the usage patterns obtained from the assumptions 530 in the specific area into classes of one-person household and multi-household household.

For another example, suppose the type is an income type, and the sub-types of the type are external and double-income. In this case, the third learning model corresponding to the income type can output the result of classifying the usage patterns obtained from the assumptions 530 in the specific region into the classes of the external bills and the double-income classes.

On the other hand, when the learning model 150 is generated in a non-learning mode, the learning model 150 may generate the learning pattern 150 from the assumptions 530 in the specific region The result of clustering the obtained usage patterns according to the detailed elements of the type can be output.

For example, let's assume that the type is the number of household members, and the type of detail is more than one, two, and three. In this case, the first learning model corresponding to the number of household members can output the result obtained by grouping the usage patterns obtained from the assumptions 530 in the specific area into one household, two households, and three households.

Meanwhile, the processor 110 may obtain the distribution data on the sub-elements of the type of the assumptions 530 in the specific region using the output values output from the learning model 150. [

One example of the distribution data will be described with reference to FIG.

The processor 110 may divide the assumptions in a specific region according to the detailed elements of the type using the resultant value obtained by dividing the usage pattern of the assumptions within the specific region according to the detailed elements of the type.

The processor 110 may store in the memory 130 information that divides the assumptions within a particular area according to the type of detail.

Referring to Figure 6, the processor 110 may generate a distribution diagram that illustrates the assumptions within a particular region 610 divided into a single household 621, a two-person household 622, and three or more households 623 .

On the other hand, the distribution chart is only one example of the distribution data, and the distribution data may mean information obtained by dividing the assumptions within the specific region as well as the distribution chart according to the detailed elements of the type.

Meanwhile, the processor 110 may transmit distribution data on the sub-elements of the types of households within the specific area to the business entity 200.

Meanwhile, the processor 110 may acquire the sales strategy information based on the distribution data on the detailed elements of the types of the assumptions in the specific area, and may transmit the acquired sales strategy information to the business entity 200. [

In particular, the processor 110 may obtain information such as goods, services, marketing, etc. suitable for providing within a particular area, based on distribution data on the sub-elements of the types of assumptions within the particular area.

For example, if there is a crowd of one person in a particular area, the processor 110 may send information about the ready-to-eat food (e.g., a microwaveable dumplings) to the business 200 have.

In another example, when furniture made up of women in a particular area is densely packed (for example, in front of women), the processor 110 may send information about the cosmetics-related goods or services to the carrier 200 to the carrier 200 .

In another example, if three or more households are concentrated in a particular area and the operator 200 is a consumer electronics store, the processor 110 may provide the merchant 200 with marketing information such as a large refrigerator or a large washing machine in a prominent place The method can be transmitted to the business entity 200.

As described above, according to the present invention, the artificial intelligence apparatus 100 can derive furniture distribution by type based on usage data of home appliances operating in conjunction with the artificial intelligence apparatus 100.

Therefore, according to the present invention, it is possible to generate business revenues by selling furniture distribution type data to a business nearby, for example, a convenience store, a bakery, a coffee shop, and a delivery service.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention.

Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments.

The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

Claims

A communication unit for receiving usage data from at least one home appliance disposed in the home;

An analysis module for obtaining a usage pattern in the assumption based on the usage data;

A learning model for outputting results obtained by dividing the assumptions in the specific region according to detailed elements of the type using the usage pattern obtained from the assumptions within the specific region; And

And a processor for providing distribution data on sub-elements of types of assumptions in the specific area to the business entity using the resultant value,

The learning model includes:

A neural network trained using multiple home usage patterns

Artificial intelligence devices.
The method according to claim 1,

Wherein the usage data includes:

Comprising at least one of identification data of a home appliance and operation data of a home appliance

Artificial intelligence devices.
The method according to claim 1,

In the usage pattern,

Including the home appliances 'hold status and the home appliances'

Artificial intelligence devices.
The method according to claim 1,

In this type,

Number of household members, household members or income type

Artificial intelligence devices.
The method according to claim 1,

The learning model includes:

Based on learning of non-background, by training the neural network using usage patterns of a plurality of assumptions

Artificial intelligence devices.
6. The method of claim 5,

The learning model includes:

When the usage pattern obtained from the assumptions in the specific region is input, the usage patterns obtained from the assumptions in the specific region are output as clustered results according to the type of detail elements

Artificial intelligence devices.
The method according to claim 1,

The learning model includes:

Based on map learning, is generated by training the neural network with usage patterns and sub-elements of the type labeled on the usage patterns

Artificial intelligence devices.
8. The method of claim 7,

The learning model includes:

When the use pattern obtained from the assumptions in the specific region is inputted, the use pattern obtained from the assumptions in the specific region is sorted according to the sub-elements of the type,

Artificial intelligence devices.
The method according to claim 1,

The processor comprising:

Classifying the usage pattern obtained from the assumptions in a plurality of regions by region and inputting a usage pattern obtained from the assumptions in the specific region among the plurality of regions into the learning model

Artificial intelligence devices.
The method according to claim 1,

The processor comprising:

Acquires sales strategy information based on distribution data on the detailed elements of types of households in a specific area, and transmits the acquired sales strategy information to the business entity

Artificial intelligence devices.
Generating training models by training neural networks using usage patterns of a plurality of assumptions;

Receiving usage data from at least one home appliance disposed in the home;

Obtaining a usage pattern in the home based on the usage data;

Outputting a result obtained by dividing the assumptions in the specific region according to the detailed elements of the type using the usage pattern obtained from the assumptions within the specific region; And

And providing distribution data on the types of assumptions in the specific area to the business entity using the resultant value

Operation method of artificial intelligence device.