KR20200052449A - Connected data architecture system to support artificial intelligence service and control method thereof - Google Patents

Abstract

The present invention relates to a connected data architecture system for an artificial intelligence (AI) service and a control method thereof. According to the present invention, the connected data architecture system for an AI service comprises: a data lake (10) for storing collected data; a data storage (20); cloud computing (30) for transmitting learning data to the data lake (10); and a micro storage (40), wherein the data lake (10) includes a data lake communication unit (11), a data collection unit (12), a learning request unit (13), and a data providing unit (14). According to the present invention, a customized AI service can be provided.

Description

CONNECTED DATA ARCHITECTURE SYSTEM TO SUPPORT ARTIFICIAL INTELLIGENCE SERVICE AND CONTROL METHOD THEREOF}

The present invention relates to a connected data architecture system for artificial intelligence services and a control method therefor, more specifically, while receiving structurally managed data through a software-defined network and receiving data learned through deep learning stably for each interface. It relates to a connected data architecture system for artificial intelligence services that can provide learned data according to a user's environment or state, and a control method therefor.

Artificial intelligence has become a key element of the 4th industrial revolution due to the rapid development of cloud computing environments and the support of big data, and deep learning and various learning policies have been implemented.

Such artificial intelligence is a technology that realizes human learning ability, reasoning ability, perception ability, and understanding ability of natural language through computer programs. In other words, artificial intelligence is a field of computer science and information technology that studies how computers can do thinking, learning, and self-development that human intelligence can do. Computers can imitate human intelligent behavior. To make it possible.

As a branch of artificial intelligence, neural networks exist, and the learning policies of neural networks are supervised learning, unsupervised learning, reinforcement learning, and transfer learning. And so on. Among the learning methods of neural networks, there are machine learning and deep learning among artificial intelligence learning methods.

Machine learning is a technology that a computer analyzes massive data by itself and predicts the future. It bundles similar things among vast amounts of data and recognizes the top and bottom structures of interrelated ones to predict future behavior based on this. It is a technique to do.

Deep Learning is a machine learning technology built on an artificial neural network (ANN) to enable computers to learn themselves as if they were using multiple data. It is a technique to tie them together and grasp the relationship between them.

Transfer learning uses deep learning only as feature extraction, and learns other models using the extracted features.How to speed up learning and increase prediction when creating a new model using an existing model to be. In other words, transfer learning is a convenient way to apply deep learning by learning by using fine-tuning for desired learning with a model that has already been trained.

As a technology using AI, in the AI audio device as described in Patent Publication No. 10-2015-0047803 (published date: 05/06/2015), the location of the current place where the AI audio device is currently located With detecting, at least one of the temperature and humidity of the current place, a sensing unit for detecting the presence or absence of a person, a first communication unit communicating with an external server via a network, a plurality of greeting voices and a plurality of greetings depending on the situation A storage unit for storing greeting data including at least one of text, music data including a plurality of music, a voice processing unit for processing and outputting voice and music, a display processing unit for processing and outputting text, and the sensing The weather information of the current place is collected from the external server by using the location information of the current place detected by the wealth, and When a person is detected in the current place by the sensing unit, it is determined whether it is a predetermined greeting output time, and when the determination result is the predetermined greeting output time, it is detected by the sensing unit from the current time and from the greeting data stored in the storage unit. A control unit that extracts at least one of the temperature and humidity and at least one of a greeting voice and a greeting text corresponding to the weather information of the current place collected from the external server and outputs it through at least one of the voice processing unit and the display processing unit There are included AI audio devices.

As the technology using artificial intelligence as in the above-mentioned patent, it is used in various fields, but most of them have a problem of relying on high-spec hardware, mass storage, and cloud. Accordingly, in a limited environment, various AI applications requiring edge devices independently or without cloud assistance and minimizing Internet connection are required without the help of the cloud.

The present invention has been proposed to solve the above-mentioned problems, while data is structurally managed through a software-defined network and stably learns data using various learning policies including deep learning for each interface to receive the learned data. An object of the present invention is to provide a connected data architecture system and control method for an artificial intelligence service that can provide the learned data according to the environment or state of the software as an artificial intelligence service using a software-defined network.

In addition, an object of the present invention is to provide a connected data architecture system and a control method for an artificial intelligence service that can be applied to low-end, low-power devices by distributing and parallelly learning data to be learned.

In addition, it is possible to learn a large amount of data to be learned using cloud computing or data lake, thereby reducing the learning computation amount, storage space, and learning time. There is a purpose to provide.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

Connected data architecture system using artificial intelligence according to the present invention and a control method therefor for achieving the above object, the data lake 10 and the data lake (10) for storing the collected data generated by collecting data ) The data storage 20 for sharing the result data generated by receiving the collected data from the transfer learning to the data lake 10, and learning generated by receiving the collected data from the data lake 10 to transfer learning A cloud computing 30 for transmitting data to the data lake 10 and a microcomputer for receiving and applying the learning data through the data lake 10 or receiving and using the result data and inputting the data Storage 40, the data lake 10, the data storage 20 and the cloud computing ( 30) data for communicating with the micro-storage 40, the data lake communication unit 11, and data for collecting the data received from the micro-storage 40 through the data lake communication unit 11 to generate the collected data A collection request unit 12 and a learning request unit 13 for judging and requesting learning to transmit the collected data generated in the data collection unit 12 to one of the data storage 20 and the cloud computing 30 ), A data providing unit 14 for providing the result data received from the data storage 20 through the learning request unit 13 or the learning data received from the cloud computing 30 to the micro storage 40. ).

In addition, the data storage 20, the data storage communication unit 21 for communicating with the data lake 10, and the data storage communication unit 21 through the data lake 10 transfers the collected data received Transfer learning unit 22 for learning, and learning completion unit 23 for transfer learning through the transfer learning unit 22 to generate result data, and result data generated in the learning completion unit 23 It may be configured to include a result sharing unit 24 for sharing.

At this time, the cloud computing 30, the cloud communication unit 31 for communicating with the data lake 10, and for backing up the collected data received from the data lake 10 through the cloud communication unit 31 The backup unit 32, the AI learning unit 33 for generating learning data by transfer learning the collected data backed up in the backup unit 32, and the learning data generated through the AI learning unit 33 It is preferably configured to include a data transfer unit 34 for delivery to the data lake 10.

In addition, the micro-storage 40 includes an information management unit 41 for storing and managing identification information for identification, a micro storage communication unit 42 for communicating with the data lake 10, and the micro storage communication unit. A data input unit 43 for inputting data to be transmitted to the data lake 10 through 42, and a data storage unit 44 for storing result data or learning data received from the data lake 10 ), A data application unit 45 for applying the learning data stored in the data storage unit 44, and learning data applied through the data application unit 45 or result data stored in the data storage unit 44 It may be configured to include a data using unit 46 for using.

In the present invention, the data collection step of collecting and accumulating the data in the data lake 10 for collecting data to generate the collected data, and the transfer learning of the collected data in the data lake 10 in parallel The data storage 20 for processing and the cloud computing 30 for back-up and transfer learning of the collected data are compared and determined by comparing the capacity of the collected data with a preset reference capacity to request learning of the collected data In the step, when the capacity of the collected data in the data lake 10 is less than or equal to a preset reference capacity, the collection data is transmitted to the data storage 20 to request learning, and the data lake 10 collects the data. When the data capacity exceeds a preset reference capacity, the cloud computing 30 collects the data. A learning request step for requesting learning by transmitting data, and a transfer learning step for transfer learning of the collected data received from the data lake 10 in the data storage 20, and transfer learning in the data storage 20 After generating result data for the result, a sharing step for sharing to the data lake 10, a backup step for backing up the collected data received from the data lake 10 in the cloud computing 30, and the cloud AI learning step of transferring learning data backed up by computing 30 to generate learning data, and delivery step for transferring learning data generated by cloud computing 30 to the data lake 10; The data lake 10 receives the result data of the data storage 20 or the learning data of the cloud computing 30 to the micro storage 40 The providing step for providing, the applying step of applying the learning data when receiving the learning data from the micro storage 40, and using the applied learning data or the result data in the micro storage 40 It comprises a use step of using the result data when receiving.

As described above, according to the present invention, it is possible to provide learned data according to a user's environment or condition by receiving data by learning data through deep learning for each interface stably while managing data structurally. , AI can operate independently on low-end devices such as wearable devices and IoT devices, and it is possible to provide AI services customized to users.

In addition, through the interface between the data lake and the data storage that collects the data to be learned, the learning of the data to be learned can be processed in parallel, so sharing and scalability can be secured. It is possible to provide AI services efficiently because learning and data classification are possible.

In addition, because it provides scalability through the interface between data lake and cloud computing that collects the data to be learned, it has an effect of shortening the learning and learning time even if the data to be learned is large.

1 is a connected data architecture system for artificial intelligence services according to an embodiment of the present invention,
2 is a data lake of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention,
3 is a data storage of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention,
4 is a cloud computing of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention,
5 is a micro storage of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention,
6 is a method of controlling a connected data architecture system for artificial intelligence services according to an embodiment of the present invention.

Hereinafter, a connected data architecture system for an artificial intelligence service and a control method therefor will be described in detail with reference to the accompanying drawings.

1 is a connected data architecture system for artificial intelligence services according to an embodiment of the present invention, Figure 2 is a data lake of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention, Figure 3 is It is a data storage of a connected data architecture system for artificial intelligence service according to an embodiment of the present invention.

In addition, Figure 4 is a cloud computing of a connected data architecture system for an artificial intelligence service according to an embodiment of the present invention, Figure 5 is a micro storage of a connected data architecture system for an artificial intelligence service according to an embodiment of the present invention And, Figure 6 is a control method of a connected data architecture system for artificial intelligence services according to an embodiment of the present invention.

When adding quotation marks to the components of the drawings, the same components have the same reference numerals as possible even though they are displayed on different drawings, and a known function determined to unnecessarily obscure the subject matter of the present invention And detailed description of the configuration is omitted. In addition, directional terms such as 'top', 'bottom', 'front', 'back', 'leading', 'front', 'end', etc. are used in connection with the orientation of the disclosed drawing (s). Since the components of the embodiments of the present invention can be positioned in various orientations, the directional terms are used for illustrative purposes and are not limiting.

Connected data architecture system for an artificial intelligence service according to an exemplary embodiment of the present invention, as shown in Figure 1, the data lake 10 for storing the collected data generated by the data, and the data By receiving the collected data from the lake 10, transfer the learning data generated by the transfer to the data lake 10 to share the result data 20, and the data lake 10 receives the collected data from the transfer learning Cloud computing 30 for transmitting the generated learning data to the data lake 10 and micro storage for receiving and applying the learning data through the data lake 10 or receiving and using the result data ( 40).

The data lake 10, as shown in Figure 2, the data storage 20 and the cloud computing 30 and the data storage communication unit 11 for communicating with the micro storage 40, the The data collection unit 12 for collecting data received from the micro storage 40 through the data lake communication unit 11 and generating the collected data, and the data collected in the data collection unit 12 are the data Results received from the data storage 20 through the learning request unit 13 and the learning request unit 13 for judging a learning request by determining to transmit to one of the storage 20 and the cloud computing 30 It may be configured to include a data providing unit 14 for providing data or learning data received from the cloud computing 30 to the micro storage 40.

The data lake 10 captures, processes, and analyzes large amounts of data and provides it to a micro storage 40 that consumes users or data. Distributed Storage Layer, Security Layer, Data Acquisition Layer, Messaging Layer, Ingestion Layer, Lambda Architecture, , It may be configured to include a service layer (Serving Layer).

In addition, the data lake 10 provides the cloud computing 30 and cloud bursting, and provides the micro storage 40 and cloud spanning.

The cloud bursting is an application distribution model used in a hybrid cloud (mixed cloud) environment, and when the capacity of the data lake 10 is exceeded, it is automatically transferred to the public cloud due to excessive demand, so that the application can be continuously executed. .

The cloud spanning is a delivery model that allows application components that require a lot of computing resources to be simultaneously deployed in multiple cloud environments, and can connect multiple computers to cooperate with each other.

The data lake communication unit 11 communicates with the data storage 20, the cloud computing 30, and the micro storage 40 through a wireless network. Wi-Fi, Internet, etc. can be used as the wireless network.

The data collection unit 12 receives the data from the micro-storage 40, accumulates and collects the data, and generates the collected data.

In this case, the data collection unit 12 may periodically receive data of the micro storage 40.

The learning request unit 13 is for requesting learning by transmitting the collected data generated in the data collection unit 12 to one of the data storage 20 and the cloud computing 30. It can be judged according to the capacity.

That is, the learning request unit 13 determines the target for the learning request from the data storage 20 and the cloud computing 30 according to the capacity of the collected data, and the determined data storage 20 and It can be transmitted to any one of the cloud computing 30 to be requested to learn.

At this time, the learning request unit 13 compares the capacity of the collected data and a preset reference capacity to determine one of the data storage 20 and the cloud computing 30 for requesting learning.

When the capacity of the collected data is less than or equal to the preset reference capacity, the collected data is sent to the data storage 20 to request learning, and when the capacity of the collected data exceeds the preset reference capacity, the cloud computing 30 ) To request the learning by sending the collected data.

The learning request unit 13 transmits the collected data to any one of the data storage 20 and the cloud computing 30 determined according to the situation, and requests the learning to efficiently transfer and learn to provide rapid AI service. It becomes possible.

The data providing unit 14 receives the result data from the data storage 20 when the collected data is transmitted to the data storage 20 by the learning request unit 13, and the learning request unit ( 13) When the collected data is transmitted to the cloud computing 30, the learning data is received from the cloud computing 30.

The data providing unit 14 provides the result data received from the data storage 20 or the learning data received from the cloud computing 30 to the corresponding micro storage 40.

At this time, the data providing unit 14 is preferably stored in advance the identification information of the micro storage 40 according to the type of the result data or learning data, the micro storage corresponding to the type of the result data or learning data (40) may be extracted to provide the result data or learning data corresponding thereto.

Also, the data providing unit 14 may provide the result data or the learning data to the requested micro storage 40.

The data storage 20, as shown in Figure 3, a data storage communication unit 21 for communicating with the data lake 10, and the data lake 10 through the data storage communication unit 21 A transfer learning unit 22 for transfer learning of the collected data received from, a learning completion unit 23 for transfer learning through the transfer learning unit 22 to generate result data, and the learning completion unit 23 ) May include a result sharing unit 24 for sharing the result data generated in.

The data storage communication unit 21 may use a wireless network to communicate with the data lake 10. In other words, the data storage communication unit 21 may communicate using the same wireless network as the data lake 10.

The transfer learning unit 22 supports transfer learning to provide a distributed AI service using FPGA or GPU resources.

The transfer learning is a section of machine learning that transfers past models to a new domain without excessive computing power. In addition, in the transfer learning, the fundamental distribution of the embedding vector of the domain is different from the original domain, and thus it is possible to reuse the existing model and domain knowledge, thereby avoiding repetitive training on a long data set.

The transfer learning unit 22 receives the collected data received from the data lake 10 using a fixed feature extractor, a fine-tuning, a pretrained model, and the like. Transfer learning.

In the learning completion unit 23, the collected data is transferred from the transfer learning unit 22 to generate result data.

The result data is a result of the completion of transfer learning of the collected data, and is preferably an AI service. Accordingly, the result data may be directly used in the micro storage 40.

The result sharing unit 24 is for sharing the result data with the data lake 10, and can also be used by sharing with other data lakes.

Since the data storage 20 is configured, it is possible to provide the same level of collaboration and scalability as the data lake 10, and the physical network, storage and computing resources of the data lake 10 for AI service. Through sharing, inheritance of collaboration and extensibility enables efficient transfer learning.

On the other hand, the cloud computing 30, as shown in Figure 4, the cloud communication unit 31 for communicating with the data lake 10, and the data lake 10 through the cloud communication unit 31 The backup unit 32 for backing up the collected data received from, and the AI learning unit 33 and the AI learning unit for generating learning data by transferring and learning the collected data backed up in the backup unit 32 33) is preferably configured to include a data transmission unit 34 for transmitting the learning data generated through the data lake (10).

The cloud communication unit 31 communicates with the data lake 10 using a wireless network such as Wi-Fi or the Internet.

The backup unit 32 receives the collected data from the data lake 10 through the cloud communication unit 31 and backs it up.

By backing up the collected data of the data lake 10 due to the backup unit 32, even if the original is damaged or deleted due to insufficient capacity or errors of the data lake 10, it is prevented from affecting AI learning in the future. I can do it.

The AI learning unit 33 generates learning data by transfer learning the backed up collected data.

The learning data is a transfer learning of the collected data, and can be used after being applied to the micro storage 40 in the future.

At this time, the AI learning unit 33 supports Fast Training based on large-capacity learning data with AI using GPU cluster or GPU Cloud and computing resources of the public cloud to transfer the collected data.

By supporting Fast Training in the AI learning unit 33, transfer learning of the collected data can be quickly performed, and even if the collected data is large, errors in learning can be prevented as much as possible.

The data transfer unit 34 transfers the learning data generated through the AI learning unit 33 to the data lake 10.

By configuring the cloud computing 30, it is possible to secure the scalability of storage through the public cloud by limiting the physical resources of the data lake 10, and learning and learning the collected data through computing resources of the public cloud. This will shorten the time.

On the other hand, algorithms for applying transfer learning in the data storage 20 and the cloud computing 30 are collected data previously learned and newly received collection data rather than randomly or intuitively applied by a developer / user or a service provider. By measuring the correlation of, the feasibility of transfer learning and time and computing resources for transfer learning can be effectively applied.

The correlation measurement is divided into a parametric method and a non-parametric method, and the parametric correlation uses a Pearson product moment correlation coefficient technique, and the nonparametric method is used. The correlation can be Kendall's Tau technique or Spearman's rank order correlation coefficient technique.

The correlation coefficient, which is a result of measuring the correlation, has a value between -1 and +1, and 0 means that there is no correlation. At this time, a value close to -1 or +1 means that the relevance is high, + indicates both directions of the relationship, and-indicates reverse direction of the relationship.

The micro-storage 40, as shown in FIG. 5, the information management unit 41 for storing and managing identification information for identification, and the micro-storage communication unit 42 for communicating with the data lake 10 ), A data input unit 43 for inputting data to be transmitted to the data lake 10 through the micro-storage communication unit 42, and result data or learning data received from the data lake 10 are stored. Data storage unit 44 for applying, a data application unit 45 for applying the learning data stored in the data storage unit 44, and learning data or the data storage unit applied through the data application unit 45 It may be configured to include a data use unit 46 for using the result data stored in (44).

The micro-storage 40 is a low-spec, low-power device such as a healthcare, wearable device, IoT device, and the like.

The information management unit 41 generates, stores, and manages the identification information so that the micro storage 40 can be distinguished from other micro storage.

The identification information may include an identification number for identifying the micro-storage 40 and the type of the micro-storage 40, and may be generated by further including information of a user who uses the micro-storage 40. have.

The micro storage communication unit 42 may use a wireless network such as the Internet or Wi-Fi to communicate with the data lake 10.

The data input unit 43 inputs data for learning and transmits it to the data lake 10 through the micro storage communication unit 42.

The data for learning may include a user environment or a state through a user's pattern analysis.

The data is initial data learned through the data storage 20 or the cloud computing 30 in order to provide AI services in the future, and may be directly input and generated by a user.

Preferably, the data input unit 43 transmits the data to the data lake 10 according to a preset transmission time.

The data storage unit 44 receives result data or learning data from the data lake 10 that receives the data, and stores the received result data or learning data.

When the learning data is stored in the data storage unit 44, the learning data is applied by the data application unit 45.

The data application unit 45 applies the learning data so that the user can use the AI service, and applies the result of the learning data.

The data use unit 46 uses the result data when learning data applied to the data application unit 45 is used or when result data is stored in the data storage unit 44. That is, the data usage unit 46 can provide AI service by using the learning data applied to the data application unit 45 or the result data according to the environment or status of the user using the micro storage 40. There will be.

The micro storage 40 configured as described above can provide the AI service to the user by overcoming the characteristics of the micro storage and the disadvantages of low-spec computing through the data lake 10.

In addition, between the micro-storage 40 and the data lake 10 is an embedded AI, suitable for a low-end embedded system, and may be an independent artificial intelligence system or an artificial intelligence system that minimizes the connection of cloud services. Accordingly, it is possible to use a learning function that minimizes the use of external communication data in products necessary for real life.

The embedded AI will support personalized services such as smart objects, natural language processing, and personalization.

The control method of the connected data architecture system using artificial intelligence configured as described above, as shown in FIG. 6, first, collects and accumulates the data in the data lake 10 to generate collected data (S10). ).

In order to transmit the collected data from the data lake 10 to either the data storage 20 or the cloud computing 30, it is determined by comparing the collected data capacity with a preset reference capacity (S20).

That is, the data lake 10 measures the capacity of the collected data to determine whether the capacity of the collected data exceeds a preset reference capacity.

When the capacity of the collected data in the data lake 10 is determined to be less than or equal to a preset reference capacity, the collected data is transmitted to the data storage 20 in order to process the transfer learning of the collected data in parallel to request learning. Is done. On the other hand, when it is determined in the data lake 10 that the collected data has exceeded a preset reference capacity, the collected data is transmitted to the cloud computing 30 to request learning by backing up the collected data and learning the transfer. Will be (S30).

When the collected data is transmitted from the data lake 10 to the data storage 20, the collected data is transferred to the data storage 20 to learn (S40).

The data storage 20 generates result data for the result of transfer learning the collected data, and shares the result data with the data lake 10 (S41).

As described above, by collaborating with the data storage 20 around the data lake 10, collecting data for learning, transfer learning, and applying the learning results support sharing and scalability.

By using the data storage 20, the limitation of physical resources for data collection can be secured and shared by SDI technology, and transfer learning can be processed in parallel by SDI technology and virtualization technology. It becomes possible to share the result data and provide AI services.

 When the collected data is transmitted from the data lake 10 to the cloud computing 30, the collected data is backed up by the cloud computing 30 (S50).

The cloud computing 30 generates learning data by transfer learning the backed up collected data (S51).

The learning data generated in the cloud computing 30 is transferred to the data lake 10 (S52).

As described above, AI learning through the cloud computing 30 around the data lake 10 has scalability of storage and computing resources through the public cloud of the cloud computing 30 to collect data, It supports the scalability of the stage of transfer learning. In other words, the limit of the physical resources of the data lake 10 secures scalability to the storage of the public cloud of the cloud computing 30, and the transfer learning of the collected data of the data lake 10 is the cloud. It is possible to shorten learning and learning time through computing resources of the public cloud of computing 30.

The data lake 10 receiving the result data of the data storage 20 or the learning data of the cloud computing 30 provides the result data or the learning data to the micro storage 40 (S60). At this time, the data lake 10, the identification information of the micro-storage 40 must be stored in advance, extract the micro-storage 40 corresponding to the result data or the learning data to extract the result data or the learning data Should be provided.

When receiving the learning data from the micro storage 40, the result of the learning data may be applied to use the AI service through the micro storage 40 (S70).

When the result data is received from the micro storage 40, it can be used in the micro storage 40. In addition, when the learning data is applied to the micro storage 40, the applied learning data may be used (S80).

The micro-storage 40 collects data by receiving the learning data or the result data through the data lake 10 and transfers the learning through the data lake 10 to perform the characteristics of low-spec computing. Will be able to overcome it.

The connected data architecture system using artificial intelligence controlled as described above, while structurally managing data, stably learns data through deep learning for each interface, receives training data and result data, and receives learning data according to the user's environment or condition. B. Since it can provide result data, artificial intelligence can operate independently on low-end devices such as healthcare, wearable devices, IoT devices, etc., and it is effective in providing AI services customized to users.

In addition, through the interface between the data lake 10 and the data storage 20 that collects the data to be learned, the learning of the data to be learned can be processed in parallel, so sharing and scalability can be secured, so that the data is integrated into one. By reducing the capacity of, it is possible to provide AI services efficiently by enabling optimal learning and data classification.

In addition, because it provides scalability through the interface between the data lake 10 and cloud computing 30, which collects the data to be learned, there is an effect to shorten the learning and learning time even if the data to be learned is large.

The embodiments described above and illustrated in the drawings should not be interpreted as limiting the technical spirit of the present invention. The protection scope of the present invention is limited only by the matters described in the claims, and a person having ordinary knowledge in the technical field of the present invention can improve and modify the technical spirit of the present invention in various forms. Therefore, such improvements and modifications will fall within the protection scope of the present invention if it is apparent to those skilled in the art.

10: Data Lake 11: Data Lake Communication Department
12: Data collection unit 13: Learning request unit
14: data providing unit 20: data storage
21: Data storage communication department 22: Transfer learning department
23: Learning completion part 24: Result sharing part
30: cloud computing 31: cloud communication department
32: backup unit 33: AI learning unit
34: data transfer unit 40: micro storage
41: Information Management Department 42: Micro Storage Communication Department
43: data input unit 44: data storage unit
45: data application unit 46: data application unit

Claims (5)

Data lake (10) for storing the collected data generated by collecting data, and data for sharing the result data generated by receiving and receiving the collected data from the data lake (10) to the data lake (10) The storage 20 and the cloud computing 30 for receiving the collected data from the data lake 10 and transferring the learning data generated by transfer learning to the data lake 10 and the data lake 10 It includes a micro-storage 40 for receiving and applying the learning data or receiving and using the result data and inputting the data,
The data lake 10 is a data lake communication unit 11 for communicating with the data storage 20, the cloud computing 30 and the micro storage 40, and the data lake communication unit 11 through the Data collection unit 12 for collecting data received from the micro storage 40 and generating the collected data, and the data collected in the data collection unit 12 in the data storage 20 and the cloud computing ( 30) from the learning request unit 13 and the result data received from the data storage 20 through the learning request unit 13 and the learning request unit 13 for determining to transmit to any one of the cloud computing (30) Connected data architecture system for artificial intelligence services, including a data providing unit (14) for providing the received learning data to the micro storage (40).
The method according to claim 1,
The data storage 20 transfers the data storage communication unit 21 for communicating with the data lake 10 and the collected data received from the data lake 10 through the data storage communication unit 21. For the transfer learning unit 22, and the transfer learning through the transfer learning unit 22, learning completion unit 23 for generating result data and the learning completion unit 23 to share the result data generated Connected data architecture system for artificial intelligence services, including a result sharing unit 24 for.
The method according to claim 1 or claim 2,
The cloud computing 30 includes a cloud communication unit 31 for communicating with the data lake 10 and a backup unit for backing up the collected data received from the data lake 10 through the cloud communication unit 31. (32), the AI learning unit 33 for generating learning data by transfer learning the collected data backed up in the backup unit 32, and the learning data generated through the AI learning unit 33 to the data Connected data architecture system for artificial intelligence services, including a data delivery unit (34) for delivery to the lake (10).
The method according to claim 3,
The micro storage 40 includes an information management unit 41 for storing and managing identification information for identification, a micro storage communication unit 42 for communicating with the data lake 10, and the micro storage communication unit 42 ) And a data input unit 43 for inputting data to be transmitted to the data lake 10 and a data storage unit 44 for storing result data or learning data received from the data lake 10. , Data applying unit 45 for applying the learning data stored in the data storage unit 44 and learning data applied through the data application unit 45 or result data stored in the data storage unit 44 are used. Connected data architecture system for artificial intelligence services, including a data use unit 46 for.
A data collection step of collecting the data in the data lake 10 for collecting data and storing the accumulated data to generate collected data;
In order to request learning of the collected data among the data storage 20 for parallelly processing the transfer learning of the collected data in the data lake 10 and the cloud computing 30 for backing up the collected data and learning the transfer of the collected data A determination step of comparing and determining the capacity of the collected data and a preset reference capacity;
When the capacity of the collected data in the data lake 10 is less than or equal to a preset reference capacity, the collected data is transmitted to the data storage 20 to request learning, and the capacity of the collected data in the data lake 10 A learning request step for requesting learning by transmitting the collected data to the cloud computing 30 when the preset reference capacity is exceeded;
A transfer learning step of transfer learning the collected data received from the data lake 10 in the data storage 20;
A sharing step for generating result data for the result of transfer learning in the data storage 20 and sharing it with the data lake 10;
A backup step of backing up the collected data received from the data lake 10 in the cloud computing 30;
AI learning step to transfer the collected data backed up in the cloud computing (30) to generate learning data,
A delivery step for delivering learning data generated in the cloud computing 30 to the data lake 10;
A providing step for receiving the result data of the data storage 20 or the learning data of the cloud computing 30 from the data lake 10 and providing it to the micro storage 40;
An application step of applying the learning data when the learning data is received from the micro storage 40;
Control method of a connected data architecture system for an artificial intelligence service, comprising using the applied learning data in the micro-storage (40) or using the result data when receiving the result data.

Images