KR20230032259A - Method and system for estimating real estate information based on deep neural network - Google Patents

Abstract

A deep neural network-based real estate information estimation method according to an embodiment of the present invention is a method for a real estate application executed by at least one processor of a terminal to estimate real estate information based on a deep neural network, comprising the steps of: obtaining arbitrary time series data representing trends in real estate prices and transaction volumes for each real estate complex within an entire region over time; linking with a deep learning module including the deep neural network; inputting the arbitrary time series data into the linked deep learning module and obtaining predicted time series data, which is embedding data for real estate time series data for each real estate complex in the entire region for a predetermined target period; obtaining a spatiotemporal influence graph representing a spatiotemporal influence between real estate complexes in the entire region based on the obtained predicted time series data; obtaining target time series data estimating real estate prices and transaction volumes for each real estate complex in the entire region for a specific period based on the obtained spatiotemporal influence graph and the predicted time series data; and visualizing and providing the obtained target time series data.

Description

Real estate information estimation method and system based on deep neural network {METHOD AND SYSTEM FOR ESTIMATING REAL ESTATE INFORMATION BASED ON DEEP NEURAL NETWORK}

The present invention relates to a method and system for estimating real estate information based on a deep neural network. More specifically, it relates to a method and system for estimating real estate market price and transaction volume using a deep neural network.

The field of real estate valuation plays a very important role, from ordinary people who consume real estate to investors and banks.

The real estate market is an area that occupies a very large part in the modern financial market, and at the same time, it is an area in which many people directly and indirectly participate.

In particular, the housing market forms an important axis in which all citizens inevitably participate in any form, and in the reality that a significant amount of Korean household assets is concentrated in real estate, accurate information on real estate values, especially prices (market prices) and transaction volume, is essential. Information will be of great importance.

In general, in order to obtain such real estate information, it is necessary to collect various related information distributed in the relevant area and analyze it based on this result. After that, it is generated by manually estimating the sale price and deriving the results of regional analysis.

In particular, real estate prices are influenced by supply and demand in the real estate market, and the real estate market is influenced by complex variables such as the real economy, values of members of society, and real estate-related policies.

However, since the variables affecting real estate information change over time and have a complex effect, it is quite difficult to accurately predict fluctuations in real estate prices or transaction volumes.

On the other hand, conventionally, information related to a specific real estate market price is obtained by visiting a real estate agency individually or using magazines, newspapers, or the Internet related to real estate sales.

However, in the case of visiting each real estate brokerage agency, considerable time and money are consumed until the desired market price information is obtained, and there is a problem that a lot of money and effort are required. There is a limit that is virtually impossible to access.

Therefore, in most recent cases, people try to obtain real estate market price information using the Internet, but various data collected through Internet searches (e.g., publicly announced land price, real estate sale price, jeonse price, monthly rent price, auction information, and/or auction information) etc.) to estimate desired real estate market price information.

For the purpose of solving these difficulties, the Ministry of Land, Infrastructure and Transport is currently providing real estate transaction data.

In other words, the Ministry of Land, Infrastructure and Transport, etc., in accordance with Article 7 Paragraphs 1 and 2 (Public Publication of Administrative Information) of the "Act on Information Disclosure of Public Institutions" for real estate for which a real estate transaction report has been reported and a fixed date The actual transaction price for the granted real estate is disclosed, and this information is systematically disclosed through the "actual transaction processing disclosure system (http://rt.molit.go.kr/)" or "public data portal (https://www. data.go.kr/)" can be obtained for free.

However, since the above actual transaction price information provided by the Ministry of Land, Infrastructure and Transport is disclosed with the exact transaction date or exact address such as lot number deleted due to privacy issues, there is a limit to the use of useful information, and direct comparison cases are not available. Accurate price estimation is difficult in case of shortage.

KR 10-1930948 B1

The present invention has been made to solve the above problems, and an object of the present invention is to provide a method and system for estimating real estate market prices and transaction volumes using a deep neural network.

In detail, the present invention uses a deep neural network to measure the influence between real estate complexes in the entire region based on real estate time series data representing trends in real estate market price and transaction volume for each real estate complex in the entire region over time, Accordingly, it is intended to provide a method and system for estimating real estate information based on a deep neural network that estimates and provides real estate market prices and transaction volumes in a specific period for the entire region.

However, the technical problems to be achieved by the present invention and the embodiments of the present invention are not limited to the technical problems described above, and other technical problems may exist.

A method for estimating real estate information based on a deep neural network according to an embodiment of the present invention is a method in which a real estate application executed by at least one processor of a terminal estimates real estate information based on a deep neural network, and includes real estate complexes in an entire region. Obtaining random time-series data representing the trend of each real estate market price and transaction volume over time; interworking with a deep learning module including the deep neural network; inputting the arbitrary time-series data to the interlocked deep learning module, and acquiring predicted time-series data that is embedded data for real estate time-series data for each real estate complex in the entire region during a predetermined target period; obtaining a spatio-temporal influence graph representing the spatio-temporal influence between real estate complexes in the entire region based on the obtained predicted time-series data; obtaining target time-series data by estimating real estate market prices and transaction volumes for each real estate complex in the entire region in a specific period based on the obtained space-time influence graph and the predicted time-series data; and visualizing and providing the obtained target time series data.

At this time, the deep learning module includes a first deep learning model, and the first deep learning model is based on a Graph Convolution Network (GCN) and is a node for each deep learning operation stage according to a predetermined stride. ASPNet performing -wise convolution operation is included as the deep neural network.

In addition, the first deep learning model takes the random time-series data as an input and the predicted time-series data as an output.

In addition, the deep learning module includes a second deep learning model, and the second deep learning model combines time aggregation of an adaptive spatial graph and a spatio-temporal graph. A temporal aggregation-based spatio-temporal grpah convolution network (TAST) is included as the deep neural network.

In addition, the second deep learning model takes the predicted time-series data as an input and the space-time influence graph as an output.

In addition, the obtaining of the target time series data may include converting the predicted time series data and the feature vector length of the space-time influence graph to be the same using a predetermined convolution layer based on the interlocked deep learning module. inputting the predicted time series data and the space-time influence graph converted to have the same feature vector length into a fully connected layer (FCL), and the target time series based on the output data of the fully connected layer. It includes acquiring data.

In addition, the step of visualizing and providing the obtained target time series data may include formatting the target time series data into at least one form of a predetermined chart, graph, table, and data map. and generating the visualized real estate information visualization data.

In addition, the step of visualizing and providing the obtained target time series data may include obtaining influence relationship data between real estate complexes in the entire region based on the space-time influence graph, and converting the acquired influence relationship data to the generated real estate and displaying the real estate information visualization data in a form corresponding to the format of the information visualization data.

Meanwhile, a system for estimating real estate information based on a deep neural network according to an embodiment of the present invention includes at least one display outputting real estate information visualization data; at least one memory; and at least one processor; and at least one application that is stored in the memory and executed by the processor to estimate real estate information based on a deep neural network. Obtain random time-series data representing according to the lapse of time, interwork with a deep learning module including the deep neural network, input the random time-series data into the interlocked deep learning module, and Obtain forecast time series data, which is embedding data for real estate time series data for each real estate complex in the region, obtain a spatio-temporal influence graph representing the spatio-temporal influence between real estate complexes in the entire region based on the obtained forecast time series data, and obtain the Target time series data obtained by estimating the real estate market price and transaction volume for each real estate complex in the entire region in a specific period based on the space-time influence graph and the predicted time series data, and visualizing the obtained target time series data in a predetermined format Create and provide real estate information visualization data.

At this time, the deep learning module of the real estate information estimation system based on a deep neural network according to an embodiment of the present invention includes a first deep learning model and a second deep learning model, and the first deep learning model includes Graph Convolution ASPNet, which is based on Network (GCN) and performs a node-wise convolution operation for each deep learning operation stage according to a predetermined stride, is included as the deep neural network, and the second deep learning model is adapted A temporal aggregation-based spatio-temporal grpah convolution network (TAST), which combines an adaptive spatial graph and a spatio-temporal graph, is used as the deep neural network. include as

A method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention measure the influence between real estate complexes in the entire region based on real estate time series data for the entire region using the deep neural network, and measure the influence between real estate complexes in the entire region By reflecting and estimating and providing the real estate market price and transaction volume in a specific period for the entire region, it is possible to derive the influence between real estate complexes in the entire region based on reliable ground data rather than human heuristic judgment. There is an effect of providing more accurately estimated real estate market price and transaction volume information.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention improve the accuracy and reliability of real estate information provided by estimating and providing real estate information based on reliable ground data as described above. At the same time, it has the effect of enhancing the fairness of real estate transactions.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention provide visualized data based on the estimated real estate information, so that users can individually search for real estate agencies or find real estate for sale. There is an effect that can reduce the effort or cost consumed to acquire the desired real estate market price information, such as performing analysis by searching magazines, newspapers, or the Internet.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention provide the visualized real estate information as above, thereby intuitively providing real estate information about the real estate market price and/or transaction volume intended by the user. There is an effect that can be quickly provided in a form that is easy to recognize.

However, the effects obtainable in the present invention are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood from the description below.

1 is a conceptual diagram of a real estate information estimation system based on a deep neural network according to an embodiment of the present invention.
2 is an internal block diagram of a terminal according to an embodiment of the present invention.
3 is an example of a block flow diagram for explaining a deep learning module according to an embodiment of the present invention.
4 is an example of a block flow diagram for explaining a first deep learning model according to an embodiment of the present invention.
5 is an example of a block flow diagram for explaining a second deep learning model according to an embodiment of the present invention.
6 is a flowchart illustrating a method for estimating real estate information based on a deep neural network according to an embodiment of the present invention.
7 and 8 are examples of real estate information visualization data according to an embodiment of the present invention.

Since the present invention can apply various transformations and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. Effects and features of the present invention, and methods for achieving them will become clear with reference to the embodiments described later in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various forms. In the following embodiments, terms such as first and second are used for the purpose of distinguishing one component from another component without limiting meaning. Also, expressions in the singular number include plural expressions unless the context clearly dictates otherwise. In addition, terms such as include or have mean that features or elements described in the specification exist, and do not preclude the possibility that one or more other features or elements may be added. In addition, in the drawings, the size of components may be exaggerated or reduced for convenience of description. For example, since the size and thickness of each component shown in the drawings are arbitrarily shown for convenience of description, the present invention is not necessarily limited to the illustrated bar.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, and when describing with reference to the drawings, the same or corresponding components are assigned the same reference numerals, and overlapping descriptions thereof will be omitted. .

1 is a conceptual diagram of a real estate information estimation system based on a deep neural network according to an embodiment of the present invention.

Referring to FIG. 1, a deep neural network-based real estate information estimation system (10: hereinafter, a real estate information estimation system) according to an embodiment of the present invention is a deep neural network that estimates real estate market prices and transaction volumes using a deep neural network. based real estate information estimation service (hereinafter referred to as real estate information estimation service) may be provided.

In detail, the real estate information estimation system 10 according to the embodiment uses a deep neural network to provide real estate time series data (hereinafter, for the entire region A real estate information estimation service that measures the influence between real estate complexes in the entire region based on real estate time series data) and estimates and provides real estate market prices and transaction volumes in a specific period for the entire region by reflecting the influence can be implemented.

Here, the real estate time-series data according to the embodiment may refer to time-series data representing trends in real estate market prices and transaction volumes for a specific real estate complex within the entire region over a predetermined period of time.

In an embodiment, the real estate information estimation system 10 providing the above real estate information estimation service may include a terminal 100, a real estate information estimation server 200, and a network 300 (Network).

At this time, the terminal 100 and the real estate information estimation server 200 may be connected through the network 300 .

Here, the network 300 according to the embodiment refers to a connection structure capable of exchanging information between nodes such as the terminal 100 and/or the real estate information estimation server 200, and such a network 300 An example of 3GPP (3rd Generation Partnership Project) network, LTE (Long Term Evolution) network, WIMAX (World Interoperability for Microwave Access) network, Internet (Internet), LAN (Local Area Network), Wireless LAN (Wireless Local Area Network) ), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth (Bluetooth) network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc. are included, but are not limited thereto.

Hereinafter, the terminal 100 and the real estate information estimation server 200 implementing the real estate information estimation system 10 will be described in detail with reference to the accompanying drawings.

- Terminal (100: Terminal)

The terminal 100 according to an embodiment of the present invention may be a predetermined computing device in which a real estate application providing a real estate information estimation service is installed.

In detail, from a hardware point of view, the terminal 100 may include a mobile type computing device 100-1 and/or a desktop type computing device 100-2 in which a real estate application is installed.

Here, the mobile type computing device 100 - 1 may be a mobile device such as a smart phone or a tablet PC in which a real estate application is installed.

For example, the mobile type computing device 100 - 1 includes a smart phone, a mobile phone, a digital broadcasting terminal, personal digital assistants (PDA), a portable multimedia player (PMP), a tablet PC, and the like. can be included

In addition, the desktop-type computing device 100-2 provides a real estate information estimation service based on wired/wireless communication, such as a personal computer such as a fixed desktop PC, a laptop computer, and an ultrabook in which a real estate application is installed. It may include a device in which a program for executing is installed.

Also, according to embodiments, the terminal 100 may further include a predetermined server computing device providing a real estate information estimation service environment.

2 is an internal block diagram of a terminal 100 according to an embodiment of the present invention.

Meanwhile, referring to FIG. 2 , from a functional point of view, the terminal 100 includes a memory 110, a processor assembly 120, a communication module 130, an interface module 140, an input system 150, and a sensor system 160. ) and a display system 170 . These components may be configured to be included in the housing of the terminal 100 .

In detail, the real estate application 111 is stored in the memory 110 , and the real estate application 111 may store any one or more of various application programs, data, and commands for providing a real estate information estimation service environment.

That is, the memory 110 may store commands and data that may be used to create a real estate information estimation service environment.

Also, the memory 110 may include a program area and a data area.

Here, the program area according to the embodiment may be linked between an operating system (OS) that boots the terminal 100 and functional elements, and the data area may be data generated according to the use of the terminal 100. can be stored.

In addition, the memory 110 may include at least one or more non-transitory computer-readable storage media and temporary computer-readable storage media.

For example, the memory 110 may be various storage devices such as ROM, EPROM, flash drive, hard drive, etc., and web storage that performs the storage function of the memory 110 on the Internet can include

The processor assembly 120 may include at least one processor capable of executing instructions of the real estate application 111 stored in the memory 110 to perform various tasks for creating a real estate information estimation service environment.

In an embodiment, the processor assembly 120 may control the overall operation of components through the real estate application 111 of the memory 110 to provide a real estate information estimation service.

The processor assembly 120 may be a system on chip (SOC) suitable for the terminal 100 including a central processing unit (CPU) and/or a graphics processing unit (GPU), and an operating system stored in the memory 110. (OS) and/or application programs may be executed, and components mounted in the terminal 100 may be controlled.

In addition, the processor assembly 120 may communicate internally with each component through a system bus, and may include one or more predetermined bus structures including a local bus.

In addition, the processor assembly 120 includes application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), and controllers. ), micro-controllers, microprocessors, and electrical units for performing other functions.

The communication module 130 may include one or more devices for communicating with external devices. This communication module 130 may communicate through a wireless network.

In detail, the communication module 130 may communicate with the terminal 100 storing a content source for implementing a real estate information estimation service environment, and may communicate with various user input components such as a controller that receives a user input.

In an embodiment, the communication module 130 may transmit/receive various data related to the real estate information estimation service to/from other terminals 100 and/or external servers.

This communication module 130, technical standards or communication schemes for mobile communication (eg, LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G NR (New Radio), WIFI) Alternatively, data may be transmitted and received wirelessly with at least one of a base station, an external terminal 100, and an arbitrary server on a mobile communication network constructed through a communication device capable of performing a short-distance communication method.

The sensor system 160 may include various sensors such as an image sensor 161, a position sensor (IMU) 163, an audio sensor 165, a distance sensor, a proximity sensor, and a contact sensor.

Here, the image sensor 161 may capture an image and/or video of a physical space around the terminal 100 .

In an embodiment, the image sensor 161 may capture and obtain an image (eg, a map image, etc.) related to the real estate information estimation service.

In addition, the image sensor 161 may be disposed on the front or/or rear side of the terminal 100 to acquire an image by photographing the direction side of the terminal 100, and may acquire a physical image through a camera disposed toward the outside of the terminal 100. space can be photographed.

The image sensor 161 may include an image sensor device and an image processing module. In detail, the image sensor 161 may process still images or moving images obtained by an image sensor device (eg, CMOS or CCD).

In addition, the image sensor 161 may extract necessary information by processing a still image or moving image obtained through an image sensor device using an image processing module, and transmit the extracted information to a processor.

The image sensor 161 may be a camera assembly including one or more cameras. The camera assembly may include a general camera that captures a visible light band, and may further include a special camera such as an infrared camera and a stereo camera.

In addition, the image sensor 161 as described above may be included in the terminal 100 and operated according to embodiments, or may be included in an external device (eg, an external server, etc.) to perform the above-described communication module 130 and/or Alternatively, it may operate through interworking based on the interface module 140 .

The position sensor (IMU) 163 may detect at least one of motion and acceleration of the terminal 100 . For example, it may be made of a combination of various position sensors such as an accelerometer, a gyroscope, and a magnetometer.

In addition, the location sensor (IMU) 163 may recognize spatial information about a physical space around the terminal 100 by interworking with the location communication module 130 such as GPS of the communication module 130 .

The audio sensor 165 may recognize sounds around the terminal 100 .

In detail, the audio sensor 165 may include a microphone capable of detecting a user's voice input using the terminal 100 .

In an embodiment, the audio sensor 165 may receive voice data necessary for a real estate information estimation service from a user.

The interface module 140 may communicatively connect the terminal 100 with one or more other devices. Specifically, interface module 140 may include wired and/or wireless communication devices compatible with one or more different communication protocols.

Through this interface module 140, the terminal 100 can be connected to various input/output devices.

For example, the interface module 140 may output audio by being connected to an audio output device such as a headset port or a speaker.

Although it has been described that the audio output device is connected through the interface module 140 as an example, an embodiment installed inside the terminal 100 may also be included.

Also, for example, the interface module 140 may obtain a user input by being connected to an input device such as a keyboard and/or a mouse.

Although it has been described that the keyboard and/or mouse are exemplarily connected through the interface module 140, an embodiment installed inside the terminal 100 may also be included.

The interface module 140 connects a device having a wired/wireless headset port, an external charger port, a wired/wireless data port, a memory card port, and an identification module. Ports, audio I/O (Input/Output) ports, video I/O (Input/Output) ports, earphone ports, power amplifiers, RF circuits, transceivers and other communication circuits It may be configured to include at least one of.

The input system 150 can detect a user's input related to the real estate information estimation service (eg, a gesture, voice command, button operation, or other type of input).

In detail, the input system 150 may include a predetermined button, a touch sensor, and/or an image sensor 161 that receives a user motion input.

In addition, the input system 150 may be connected to an external controller through the interface module 140 to receive a user's input.

The display system 170 may output various information related to the real estate information estimation service as a graphic image.

As an embodiment, the display system 170 may display real estate time-series data (in an embodiment, real estate information visualization data, etc.) visualized in various ways.

Such displays include a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), and a flexible display. , a 3D display, and an e-ink display.

The components may be disposed in the housing of the terminal 100, and the user interface may include a touch sensor 173 on the display 171 configured to receive a user touch input.

In detail, the display system 170 may include a display 171 that outputs an image and a touch sensor 173 that detects a user's touch input.

For example, the display 171 may be implemented as a touch screen by forming a mutual layer structure or integrally with the touch sensor 173 . Such a touch screen may function as a user input unit providing an input interface between the terminal 100 and the user, and may provide an output interface between the terminal 100 and the user.

Meanwhile, the terminal 100 according to an embodiment of the present invention may perform deep learning required for a real estate information estimation service in conjunction with a deep neural network.

Here, the deep neural network according to the embodiment may include a Graph Convolution Network (GCN), an adaptive spatio-temporal graph network, and/or a temporal aggregation-based spatio-temporal grpah convolution network.

In addition, depending on the embodiment, the terminal 100 may further perform at least a part of the functional operation performed by the real estate information estimation server 200 to be described later.

- Real estate information estimation server (200: Real estate information estimation server)

Meanwhile, the real estate information estimation server 200 according to an embodiment of the present invention may perform a series of processes for providing a real estate information estimation service.

In detail, in the embodiment, the real estate information estimation server 200 provides the real estate information estimation service by exchanging data necessary for driving the real estate information estimation process in an external device such as the terminal 100 with the external device. can provide

More specifically, in the embodiment, the real estate information estimating server 200, in an external device (in the embodiment, the mobile type computing device 100-1 and/or the desktop type computing device 100-2, etc.) 111) can provide an environment in which it can operate.

To this end, the real estate information estimation server 200 may include applications, data, and/or commands for the real estate application 111 to operate, and may transmit/receive data based thereon with the external device.

In addition, in the embodiment, the real estate information estimation server 200 may obtain arbitrary time-series data for the entire region based on interworking with an external server and/or user input.

Here, the random time-series data according to the embodiment may mean real estate time-series data for the entire area for a random period.

In addition, in an embodiment, the real estate information estimation server 200 may obtain target time series data based on the obtained random time series data.

Here, the target time series data according to the embodiment may mean real estate time series data for the entire region estimated for a specific period.

At this time, in the embodiment, the real estate information estimation server 200 may perform deep learning necessary for the real estate information estimation service in conjunction with a predetermined deep neural network, and through this, the target based on the arbitrary time series data Time series data can be obtained.

In detail, in the embodiment, the real estate information estimation server 200 reads a predetermined deep neural network driving program built to perform the deep learning from a memory module, and according to the read predetermined deep neural network system, the following Deep learning can be performed.

Here, the deep neural network according to the embodiment may include a Graph Convolution Network (GCN), an adaptive spatio-temporal graph network, and/or a temporal aggregation-based spatio-temporal grpah convolution network.

At this time, depending on the embodiment, the above deep neural network is directly included in the real estate information estimation server 200 or implemented as a device and/or server separate from the real estate information estimation server 200 to provide the real estate information estimation service. Deep learning can be performed for

In the following description, it is described that the deep neural network is included and implemented in the real estate information estimation server 200, but is not limited thereto.

In more detail, in the embodiment, the real estate information estimation server 200 may perform deep learning with the random time series data as an input and the target time series data as an output in conjunction with a predetermined deep neural network as described above. .

3 is an example of a block flow diagram for explaining a deep learning module according to an embodiment of the present invention.

More specifically, referring to FIG. 3, in the embodiment, the real estate information estimation server 200 includes a deep learning module 210 that performs deep learning to provide a real estate information estimation service based on the deep neural network. can do.

In detail, the deep learning module 210 according to the embodiment takes the above-described arbitrary time series data (ie, real estate time series data for the entire region for a certain period) as an input and aims to estimate the real estate time series data. It may include a deep neural network that outputs a result of embedding real estate time-series data (hereinafter, predicted time-series data) for the entire region during the period.

In addition, the deep learning module 210 is a deep neural network that takes the predicted time series data as an input and outputs a graph representing spatiotemporal influence between real estate complexes in the entire region (hereinafter referred to as a spatiotemporal influence graph). can include

In addition, in an embodiment, the deep learning module 210 generates time series data (hereinafter referred to as target time series data) representing the real estate market price and transaction volume for the entire region in a specific period based on the predicted time series data and the space-time influence graph. It is possible to perform deep learning to estimate and output.

In more detail, in an embodiment, the deep learning module 210 may include a first deep learning model and a second deep learning model.

Specifically, the first deep learning model 211 according to the embodiment may be a deep neural network that receives the random time-series data as an input and the predicted time-series data as an output.

In detail, in the embodiment, the first deep learning model 211 is a W-net backbone network based on a Graph Convolution Network (GCN) and performing node-wise convolution. ) may be included.

In general, a conventional Graph Convolution Network (GCN) has an over-smoothing problem in which performance decreases as more deep learning layers are accumulated.

Thus, in an embodiment of the present invention, the first deep learning model 211 may be implemented based on a plurality of layers that perform convolution for each node to avoid the above-described problem.

That is, in the embodiment, the first deep learning model 211 stacks a plurality of layers that perform convolution for each node and performs deep learning based thereon, thereby minimizing inter-node convolution. In this way, the problem of excessive uniformity can be avoided.

In addition, in the embodiment, the first deep learning model 211 expands a receptive field by performing an operation according to a predetermined stride in the case of an image task and increases the feature size and operation cost. Focusing on the fact that the computation cost can be reduced, it can be formed in a structure in which deep learning operation stages are divided according to predetermined intervals and layers are stacked based on them.

In addition, the form of the w-block constituting the above first deep learning model 211 (ie, ASPNet consisting of node-wise convolution operation) can be implemented as shown in FIG.

In detail, as an embodiment, the w-block of the first deep learning model 211 converts the first operation value obtained by applying the result of the convolution operation for each node as the Tanh activation function and the result of the convolution operation for each node as the Sigmoid activation function. Output data of a corresponding layer may be provided by synthesizing the applied second operation value.

In summary, the first deep learning model 211 can perform a node-wise convolution operation for each deep learning operation stage based on the w-block form by taking the random time-series data as an input, As a result, the predicted time series data may be output.

At this time, each real estate time series data (ie, each node) in the random time series data input to the first deep learning model 211 based on the Graph Convolution Network (GCN) determines the real estate market price and transaction volume of the corresponding real estate complex. It is time-series data represented over time, and since the convolution per node as described above performs feature extraction for each node without inter-node operation, the W-net backbone network is a deep neural network that eventually extracts temporal features for each node. could be a network.

In addition, in the embodiment, the predicted time series data obtained through the first deep learning model 211 is time series data representing the real estate market price and transaction volume for the entire region in a specific period for which real estate time series data is to be estimated (ie, , may serve as base data for estimating target time series data) in the embodiment.

In this way, the first deep learning model 211 implemented based on the above-described structure avoids the problem of excessive standardization of the Graph Convolution Network (GCN) and at the same time minimizes the cost of deep learning operation, thereby efficiently Forecast time series data can be output.

In addition, the first deep learning model 211, by providing the predicted time-series data as described above, enables estimation of the target time-series data, which will be described later, and can improve its accuracy and reliability.

Meanwhile, the second deep learning model 212 according to an embodiment may be a deep neural network having the predicted time-series data as an input and the space-time influence graph as an output.

In an embodiment, the space-time influence graph may be a graph representing the space-time influence (eg, price increase and decrease effect and/or transaction volume increase and decrease effect, etc.) between real estate complexes within the entire region.

5 is an example of a block flow diagram for explaining the second deep learning model 212 according to an embodiment of the present invention.

In detail, referring to FIG. 5 , the second deep learning model 212 according to the embodiment is based on time aggregation of a method combining an adaptive spatial graph and a spatio-temporal graph. It may include a graph convolution network (Temporal aggregation-based spatio-temporal grpah convolution network, TAST).

More specifically, in general, a conventional Graph Convolution Network (GCN) usually uses an adaptive spatial graph or a static spatio-temporal graph.

Specifically, data in a general deep neural network structure provides connectivity information (adjacency graph) between nodes.

For example, in real estate time series data for the entire region, connection information between real estate complexes within the entire region may be implemented as an adjacency matrix.

However, as in the example, predefined connectivity information has limitations in identifying hidden dependencies between nodes.

Therefore, conventionally, the above problem is solved using a method of automatically generating a spatial adjacency graph structure using a deep neural network (adaptive spatial graph).

However, since the real estate time series data as in the embodiment of the present invention has spatiotemporal characteristics including not only spatial characteristics but also temporal characteristics, performing the GCN operation with the spatial tangent graph properly changes the shape of the real estate time series data. may not be reflected.

On the other hand, conventionally, attempts have been made to solve the above-described problem using a spatio-temporal graph, but this is not an adaptive graph, but a predefined graph based on a feature manually created by a person. There is a limit to generating it, which can degrade the expressive ability of the deep neural network model.

In an embodiment of the present invention, in order to solve the limitations of the conventional method as described above, the second deep learning model 212 uses an adaptive spatial adjacency graph and a spatio-temporal graph. It can be implemented as a time-aggregation-based space-time graph convolution network (TAST) of a combining method.

That is, the second deep learning model 212 according to an embodiment of the present invention may be a deep neural network model that generates the space-time graph by reflecting the characteristics of real estate time-series data having spatio-temporal characteristics.

In more detail, referring to (a) of FIG. 5, the second deep learning model 212 performs a spatio-temporal graph convolution (STGC) operation to handle the spatio-temporal graph. can be performed.

Here, the space-time graph convolution (STGC) may be an operation of applying and aggregating a temporal weighted sum for each feature in consideration of temporal connectivity between nodes.

At this time, in the embodiment, the second deep learning model 212 goes further than the existing space-time graph convolution algorithm, reconstructs the nodes based on temporal connectivity between the nodes, and integrates temporal and spatial operations through this. It can perform time-aggregation-based space-time graph convolution (TAST) operations.

That is, the second deep learning model 212 according to the embodiment uses a time aggregation-based space-time graph convolution (TAST) algorithm developed from the existing space-time graph convolution (STGC) algorithm to obtain the predicted time-series data. It is possible to perform deep learning with an input and the space-time influence graph as an output.

In this case, the time aggregation-based space-time graph convolution network (TAST) according to the embodiment may be implemented as shown in (b) of FIG. 5 .

As such, the second deep learning model 212 is a time aggregation-based space-time graph convolution network (TAST) combining an adaptive spatial graph and a spatio-temporal graph. By being implemented, it is possible to generate and output a space-time influence graph that changes according to the time provided by the predicted time-series data.

In addition, the second deep learning model 212 provides a space-time influence graph as above, so that the path on which the target time-series data described later is estimated (eg, the first real estate complex in the entire area is second to third) affected by the real estate complex, etc.), can improve its reliability, and can provide objectively verifiable ground data.

On the other hand, the deep learning module 210 including the first deep learning model 211 and the second deep learning model 212 as described above in the embodiment, the above information obtained from the first deep learning model 211 Based on the predicted time series data and the space-time influence graph obtained from the second deep learning model 212, real estate time series data for the entire region in a specific period (ie, real estate market price for the entire region in a specific period) and time series data indicating transaction volume) may perform deep learning (hereinafter, target deep learning) that outputs target time series data.

In detail, in the embodiment, the deep learning module 210 includes node-wise convolution operation result values for each deep learning operation stage of the first deep learning model 211 (ie, the predicted time series data), To generate and output the target time-series data by performing the target deep learning that synthesizes the time-aggregation-based space-time graph convolution (TAST) operation result of the second deep learning model 212 (ie, the space-time influence graph) can

In an embodiment, the deep learning module 210 may perform the target deep learning and convert the predicted time series data and the feature vector length of the space-time influence graph to be the same using a predetermined convolution layer. there is.

In addition, the deep learning module 210 may pass the predicted time-series data and the space-time influence graph converted to have the same feature vector length through a fully connected layer (FCL).

Thus, the deep learning module 210 may generate and output the target time-series data based on the output data of the fully connected layer.

Therefore, the deep learning module 210 includes output data (in an embodiment, predicted time series data) of the first deep learning model 211 and output data (in an embodiment, predicted time series data) of the second deep learning model 212. The target time series data can be generated and output based on a space-time influence graph), and through this, real estate market price and transaction volume information estimated through deep learning based on reliable ground data rather than human heuristic judgments can be provided. there is.

Returning again, in another embodiment, the real estate information estimation server 200 may visualize and provide target time series data obtained as described above.

In an embodiment, the real estate information estimation server 200 may generate and provide real estate information visualization data obtained by visualizing the target time series data according to a predetermined method.

In addition, in the embodiment, the real estate information estimation server 200 may store and manage various application programs, commands, and/or data for implementing the real estate information estimation service.

In an embodiment, the real estate information estimation server 200 may store and manage at least one or more arbitrary time series data, predicted time series data, a space-time influence graph, target time series data, real estate information visualization data, and/or a predetermined deep learning model. can

On the other hand, further referring to FIG. 1, in the embodiment, the real estate information estimation server 200 as described above includes at least one processor module (220) for data processing and at least one processor module for exchanging data with an external device. A predetermined computing device including one or more communication modules (230: Communication Module) and at least one memory module (240: Memory Module) storing various application programs, data and/or instructions for providing a real estate information estimation service. can be implemented as

Here, the memory module 240 may store any one or more of an operating system (OS), various application programs, data, and instructions for providing a real estate information estimation service.

Also, the memory module 240 may include a program area and a data area.

Here, the program area according to the embodiment may be linked between an operating system (OS) that boots the terminal 100 and functional elements, and the data area may be data generated according to the use of the terminal 100. can be stored.

In an embodiment, the memory module 240 may be a variety of storage devices such as ROM, RAM, EPROM, flash drive, hard drive, etc., and a web device that performs the storage function of the memory module 240 on the Internet. It may also be web storage.

Also, the memory module 240 may be a recording medium detachable from the terminal 100 .

Meanwhile, the processor module 220 may control the overall operation of each unit described above in order to implement the real estate information estimation service.

The processor module 220 may be a system-on-a-chip (SOC) suitable for the terminal 100 including a central processing unit (CPU) and/or a graphics processing unit (GPU), and is stored in the memory module 240. It can execute an operating system (OS) and/or an application program, and can control each component mounted on the terminal 100 .

In addition, the processor module 220 may communicate internally with each component through a system bus, and may include one or more predetermined bus structures including a local bus.

In addition, the processor module 220 includes application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), and controllers. ), micro-controllers, microprocessors, and electrical units for performing other functions.

In the above description, it has been described that the real estate information estimating server 200 according to the embodiment of the present invention performs the functional operations as described above, but at least some of the functional operations performed by the real estate information estimating server 200 according to the embodiment may be performed in an external device (eg, terminal 100, etc.), and at least a part of the functional operation performed in the external device may be further performed in the real estate information estimation server 200. It could be possible.

- Real estate information estimation method based on deep neural network

Hereinafter, a method in which the real estate application 111 executed by at least one processor of the terminal 100 according to an embodiment of the present invention estimates real estate information based on a deep neural network will be described in detail with reference to the accompanying drawings. .

In an embodiment of the present invention, at least one processor of the terminal 100 may execute at least one real estate application 111 stored in at least one memory 110 or operate in a background state.

Hereinafter, the at least one or more processors operate to execute commands of the real estate application 111 to perform the method of providing the above-described real estate information estimation service, which is shortened to that the real estate application 111 performs. do.

6 is a flowchart illustrating a method for estimating real estate information based on a deep neural network according to an embodiment of the present invention.

Referring to FIG. 6 , in an embodiment, the real estate application 111 executed by at least one processor of the terminal 100 or operating in a background state may obtain arbitrary time-series data. (S101)

Here, the random time-series data according to an embodiment of the present invention may mean real estate time-series data for the entire area for an arbitrary period.

That is, the arbitrary time-series data may be real estate time-series data representing trends in real estate market price and transaction volume for each of at least one real estate complex in the entire region for an arbitrary period of time.

Here, the real estate time series data according to the embodiment indicates trends in real estate market prices and transaction volume for a predetermined period of time for a specific real estate complex (eg, a specific apartment and / or officetel, etc.) in the entire region over time It can mean time series data.

In detail, in the embodiment, the real estate application 111 collects and obtains the arbitrary time-series data based on external servers (eg, KB Real Estate, Korea Appraisal Board and / or Ministry of Land, Infrastructure and Transport servers, etc.) and / or user input, etc. can

For example, the real estate application 111 may provide a first real estate complex (eg, a first apartment) to a fifth real estate complex (eg, from March 2003 to March 2021, etc.) 5) Arbitrary time-series data including a plurality of real estate time-series data indicating trends in real estate market price and transaction volume for each apartment can be obtained from an external server and/or user input.

Also, in an embodiment, the real estate application 111 may obtain predicted time series data based on the obtained random time series data. (S103)

Here, the predicted time series data according to an embodiment of the present invention is a real estate time series for the entire region during a target period for estimating real estate time series data (eg, a period from a predetermined past time to a predetermined future time, etc.) It may refer to data resulting from embedding data.

In detail, in an embodiment, the real estate application 111 may acquire the predicted time series data in conjunction with the deep learning module 210 described above.

In more detail, the real estate application 111 may acquire the predicted time-series data in conjunction with the first deep learning model 211 of the deep learning module 210 .

Specifically, the real estate application 111 may input arbitrary time-series data obtained as described above to the first deep learning model 211 .

At this time, the first deep learning model 211 receiving the random time-series data takes the random time-series data as an input and performs deep learning within the first deep learning model 211 based on the above-described w-block form. A node-wise convolution operation may be performed for each operation stage, and as a result, the predicted time series data may be output.

In the embodiment, a detailed description of how the first deep learning model 211 outputs the predicted time series data based on the random time series data is the first deep learning model of the above-described real estate information estimation server 200 ( 211) is followed.

In addition, the first deep learning model 211 may provide the predicted time-series data output as above to the real estate application 111 .

Thus, the real estate application 111 may obtain the predicted time series data based on the random time series data.

In this way, the real estate application 111 performs deep learning in conjunction with the first deep learning model 211 according to the embodiment, thereby avoiding the excessive standardization problem of the Graph Convolution Network (GCN) and at the same time deep learning It is possible to efficiently obtain the predicted time series data by minimizing computational cost.

In addition, the real estate application 111 obtains the predicted time series data as described above using a deep neural network, so that the above-described target time series data can be estimated based on objective data rather than subjective judgment of a person later At the same time, its accuracy and reliability can be improved.

Also, in an embodiment, the real estate application 111 may obtain a space-time influence graph based on the obtained predicted time-series data. (S105)

Here, the space-time influence graph according to the embodiment may be a graph representing spatiotemporal influence between real estate complexes in the entire region.

For example, the space-time influence graph may be a graph showing an influence relationship between price increases and/or increases and decreases in transaction volume between real estate complexes within the entire region.

In an embodiment, the spatio-temporal influence graph may be converted into predetermined influence relationship data (eg, influence direction and/or influence index) according to the corresponding spatio-temporal influence level, and may be visualized and provided.

In detail, in an embodiment, the real estate application 111 may acquire the space-time influence graph in conjunction with the deep learning module 210 .

In more detail, the real estate application 111 may acquire the space-time influence graph in association with the second deep learning model 212 of the deep learning module 210 .

Specifically, the real estate application 111 may input the predicted time-series data obtained by interworking with the first deep learning model 211 to the second deep learning model 212 .

At this time, the second deep learning model 212 receiving the predicted time series data may perform a time-aggregation-based space-time graph convolution (TAST) operation based on the input predicted time-series data to output the space-time influence graph. there is.

In the embodiment, a detailed description of how the second deep learning model 212 outputs the space-time influence graph based on the predicted time series data is the second deep learning model of the above-described real estate information estimation server 200 ( 212) is followed.

In addition, the second deep learning model 212 may provide the space-time influence graph output as above to the real estate application 111 .

Thus, the real estate application 111 may acquire the space-time influence graph based on the predicted time-series data.

In this way, the real estate application 111 acquires a space-time influence graph that changes in response to the time provided by the predicted time series data by performing deep learning in conjunction with the second deep learning model 212 according to the embodiment. can do.

In addition, the real estate application 111 acquires the space-time influence graph as described above using a deep neural network, so that the estimated path for the target time series data described later (eg, the first real estate complex in the entire area is affected by the second to third real estate complexes, etc.) can be provided, reliability thereof can be improved, and objective basis data for estimating the target time series data can be provided.

Also, in an embodiment, the real estate application 111 may obtain target time series data based on the obtained prediction time series data and the space-time influence graph. (S107)

Here, the target time series data according to the embodiment, based on the predicted time series data and the space-time influence graph, in a specific period (eg, the current time or a period including a predetermined period before and after the current time) It may refer to time series data obtained by estimating real estate time series data for a region (ie, time series data representing real estate prices and transaction volumes for the entire region in a specific period).

In an embodiment, since the target time series data is generated based on the predicted time series data and the spatio-temporal influence graph as described above, real estate for the entire area in a specific period estimated by considering all spatio-temporal characteristics of the real estate time series data. It can be time series data.

In detail, in an embodiment, the real estate application 111 may acquire the target time series data in conjunction with the deep learning module 210 .

In more detail, the real estate application 111 interworks with the deep learning module 210, and the predicted time series data obtained from the first deep learning model 211 of the deep learning module 210 and the deep learning module ( Target deep learning based on the space-time influence graph obtained from the second deep learning model 212 of (210) may be performed.

Here, the target deep learning according to the embodiment is a deep learning in which the predicted time series data and the space-time influence graph are used as input data, and target time series data representing the real estate market price and transaction volume for the entire region in a specific period are output. can mean

In detail, in the embodiment, the real estate application 111 interworks with the deep learning module 210, and the node-wise convolution operation result value for each deep learning operation stage of the first deep learning model 211 ( That is, to perform the target deep learning that synthesizes the predicted time series data) and the time-aggregation-based space-time graph convolution (TAST) operation result of the second deep learning model 212 (ie, the space-time influence graph). can

As an embodiment, the real estate application 111 may use a predetermined convolution layer in conjunction with the deep learning module 210 to convert the predicted time series data and the feature vector length of the space-time influence graph to the same length. can

In addition, the real estate application 111 may pass the predicted time-series data and the space-time influence graph converted to have the same feature vector length through a fully connected layer (FCL).

Thus, the real estate application 111 may obtain the target time series data based on the output data of the fully connected layer.

That is, the real estate application 111 performs target deep learning as described above in conjunction with the deep learning module 210 according to the embodiment, thereby targeting a target time series representing the real estate market price and transaction volume for the entire region in a specific period. Data can be obtained, and through this, real estate market price and transaction volume information more accurately estimated based on reliable ground data rather than human heuristic judgment can be provided.

Also, in an embodiment, the real estate application 111 may visualize and provide the obtained target time series data. (S109)

7 and 8 are examples of real estate information visualization data according to an embodiment of the present invention.

In detail, referring to FIGS. 7 and 8 , in the embodiment, the real estate application 111 converts the target time series data obtained as above (ie, time series data representing the real estate market price and transaction volume for the entire area in a specific period). Real estate information visualization data visualized in a predetermined manner can be created.

In an embodiment, the real estate application 111 converts the target time series data into a predetermined chart (Chart, eg, bar chart, line chart, area chart, scatter plot) plot, pie chart, bubble chart, radar chart, frame diagram, funnel plot and/or word cloud chart, etc.) , in the form of a graph, table, and/or data map (eg, dot density map, choropleth map, field map, symbol map, bubble map, connection map, and/or flow map, etc.) Visualization may generate the real estate information visualization data.

At this time, in the embodiment, the real estate application 111 may visualize the spatiotemporal influence between real estate complexes in the entire region obtained from the above-described space-time influence graph and include it in the real estate information visualization data.

In an embodiment, the real estate application 111 provides data on the relationship of influence between real estate complexes in the entire region (eg, influence) calculated according to a predetermined method (eg, a numerical value proportional to the degree of temporal and spatial influence) based on the temporal and spatial influence. It is possible to obtain an influence direction indicating the direction in which the effect is applied and/or an influence index indicating the degree of influence as a numerical value).

In addition, the real estate application 111 may visualize the acquired data on the influence relationship between real estate complexes in a predetermined manner and include it in the real estate information visualization data.

In detail, in an embodiment, the real estate application 111 may visualize the influence relationship data in a form corresponding to the method of visualizing the target time series data and include it in the real estate information visualization data.

For example, referring to FIG. 7 , when the real estate application 111 generates real estate information visualization data by visualizing the target time series data in a data map format, a form corresponding to the data map format ( For example, a form of displaying influence relationship data between a plurality of real estate complexes (A, B, C, ..., I) in the data map with a predetermined symbol (eg, arrow symbol, etc.) and / or text on the data map, etc. ), the influence relationship data may be visualized and included in the real estate information visualization data.

In addition, the real estate application 111 may transmit and provide the generated real estate information visualization data to a display output and/or an external device (eg, another terminal 100, etc.).

In this way, the real estate application 111 can visualize and provide the real estate market price and transaction volume time series data for a specific period estimated through the deep neural network in various ways, thereby providing the real estate information that the user wants to check (in the embodiment, real estate Market price and trading volume, etc.) can be implemented and provided in a diversified form that is easy to intuitively recognize.

In summary, in the embodiment, the real estate application 111 measures the influence between real estate complexes in the entire area by using time-series data representing trends in real estate market price and transaction volume for each real estate complex in the entire area over time. Real estate prices and transaction volumes in a specific period for at least one real estate complex in the entire region may be estimated, visualized, and provided by reflecting this.

As described above, the deep neural network-based real estate information estimation method and system according to an embodiment of the present invention measure the influence between real estate complexes in the entire region based on real estate time series data for the entire region using the deep neural network, , By estimating and providing the real estate market price and transaction volume in a specific period for the entire region by reflecting this, the influence between real estate complexes in the entire region can be derived based on reliable ground data rather than human heuristic judgment, Through this, there is an effect of providing more accurately estimated real estate market price and transaction volume information.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention improve the accuracy and reliability of real estate information provided by estimating and providing real estate information based on reliable ground data as described above. At the same time, it has the effect of enhancing the fairness of real estate transactions.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention provide visualized data based on the estimated real estate information, so that users can individually search for real estate agencies or find real estate for sale. There is an effect that can reduce the effort or cost consumed to acquire the desired real estate market price information, such as performing analysis by searching magazines, newspapers, or the Internet.

In addition, the method and system for estimating real estate information based on a deep neural network according to an embodiment of the present invention provide the visualized real estate information as above, thereby intuitively providing real estate information about the real estate market price and/or transaction volume intended by the user. There is an effect that can be quickly provided in a form that is easy to recognize.

In addition, the embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer readable recording medium. The computer readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. medium), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like as well as machine language codes generated by a compiler. A hardware device may be modified with one or more software modules to perform processing according to the present invention and vice versa.

Specific implementations described in the present invention are examples and do not limit the scope of the present invention in any way. For brevity of the specification, description of conventional electronic components, control systems, software, and other functional aspects of the systems may be omitted. In addition, the connection of lines or connecting members between the components shown in the drawings are examples of functional connections and / or physical or circuit connections, which can be replaced in actual devices or additional various functional connections, physical connection, or circuit connections. In addition, if there is no specific reference such as “essential” or “important”, it may not be a component necessarily required for the application of the present invention.

In addition, the detailed description of the present invention described has been described with reference to preferred embodiments of the present invention, but those skilled in the art or those having ordinary knowledge in the art will find the spirit of the present invention described in the claims to be described later. And it will be understood that the present invention can be variously modified and changed without departing from the technical scope. Therefore, the technical scope of the present invention is not limited to the contents described in the detailed description of the specification, but should be defined by the claims.

Claims (10)

A method in which a real estate application executed by at least one processor of a terminal estimates real estate information based on a deep neural network,
Obtaining random time-series data representing trends in real estate market prices and transaction volume for each real estate complex in the entire region over time;
interworking with a deep learning module including the deep neural network;
inputting the arbitrary time-series data to the interlocked deep learning module, and acquiring predicted time-series data that is embedded data for real estate time-series data for each real estate complex in the entire region during a predetermined target period;
obtaining a spatio-temporal influence graph representing the spatio-temporal influence between real estate complexes in the entire region based on the obtained predicted time-series data;
obtaining target time-series data by estimating real estate market prices and transaction volumes for each real estate complex in the entire region in a specific period based on the obtained space-time influence graph and the predicted time-series data; and
Visualizing and providing the obtained target time series data
Real estate information estimation method based on deep neural network. According to claim 1,
The deep learning module includes a first deep learning model,
The first deep learning model,
ASPNet, which is based on a Graph Convolution Network (GCN) and performs a node-wise convolution operation for each deep learning operation stage according to a predetermined stride, is included as the deep neural network
Real estate information estimation method based on deep neural network. According to claim 2,
The first deep learning model,
The arbitrary time series data as input and the predicted time series data as output
Real estate information estimation method based on deep neural network. According to claim 3,
The deep learning module includes a second deep learning model,
The second deep learning model,
A temporal aggregation-based spatio-temporal grpah convolution network (TAST), which combines an adaptive spatial graph and a spatio-temporal graph, is used in the deep neural network. including as a network
Real estate information estimation method based on deep neural network. According to claim 4,
The second deep learning model,
The predicted time series data as input and the space-time influence graph as output
Real estate information estimation method based on deep neural network. According to claim 5,
Obtaining the target time series data,
Converting the prediction time-series data and the feature vector length of the space-time influence graph to be the same using a predetermined convolution layer based on the interlocked deep learning module;
Inputting the predicted time-series data and the space-time influence graph converted to the same feature vector length into a fully connected layer (FCL);
Acquiring the target time series data based on output data of the fully connected layer
Real estate information estimation method based on deep neural network. According to claim 1,
Visualizing and providing the obtained target time series data,
Generating real estate information visualization data that visualizes the target time series data in at least one format of a predetermined chart, graph, table, and data map
Real estate information estimation method based on deep neural network. According to claim 7,
Visualizing and providing the obtained target time series data,
Obtaining influence relationship data between real estate complexes in the entire region based on the space-time influence graph;
Further comprising displaying the acquired influence relationship data on the real estate information visualization data in a form corresponding to the format of the generated real estate information visualization data
Real estate information estimation method based on deep neural network. at least one display outputting real estate information visualization data;
at least one memory; and
at least one processor; including,
At least one application stored in the memory and executed by the processor to estimate real estate information based on a deep neural network, the at least one application comprising:
Obtain random time-series data representing trends in real estate market prices and transaction volume for each real estate complex in the entire region over time,
In conjunction with a deep learning module including the deep neural network,
Inputting the arbitrary time series data to the interlocked deep learning module and obtaining predicted time series data, which is embedded data for real estate time series data for each real estate complex in the entire region during a predetermined target period,
Obtaining a spatiotemporal influence graph representing the spatiotemporal influence between real estate complexes in the entire region based on the obtained predicted time series data;
Obtaining target time series data by estimating the real estate market price and transaction volume for each real estate complex in the entire region in a specific period based on the obtained space-time influence graph and the predicted time series data;
Creating and providing the real estate information visualization data obtained by visualizing the obtained target time series data in a predetermined format
Real estate information estimation system based on deep neural network. According to claim 9,
The deep learning module,
Including a first deep learning model and a second deep learning model,
The first deep learning model,
ASPNet, which is based on a Graph Convolution Network (GCN) and performs a node-wise convolution operation for each deep learning operation stage according to a predetermined stride, is included as the deep neural network,
The second deep learning model,
A temporal aggregation-based spatio-temporal grpah convolution network (TAST), which combines an adaptive spatial graph and a spatio-temporal graph, is used in the deep neural network. including as a network
Real estate information estimation system based on deep neural network.

Images