KR20220037608A - SYSTEM AND METHOD FOR classification OF LIST AND COMPUTER-READABLE RECORDING MEDIUM THEREOF - Google Patents

Abstract

According to an embodiment of the present invention, a list classification system receives learning data including a name and corresponding attribute information, encodes the learning data into an ASCII code, and may extract a correlation between the name and the corresponding attribute information as output data using a machine learning algorithm for encoded learning data. Therefore, accurate and fast machine learning can be performed.

Description

List classification system and method, and a computer program for executing the method on a computer is recorded, a computer readable recording medium

The present invention relates to a list classification system and method, and to a computer-readable recording medium in which a computer program for executing the method in a computer is recorded. Ascii) a list classification system and method capable of converting into code and forming a relationship between a name and a property corresponding thereto as output data using a machine learning algorithm, a computer program for executing the method on a computer, computer-readable It relates to possible recording media.

An artificial intelligence (AI) system is a computer system that implements human-level intelligence, and unlike the existing rule-based smart system, the machine learns and makes decisions on its own. The more the AI system is used, the better the recognition rate and the more accurate understanding of user preferences, and the existing rule-based smart systems are gradually being replaced by deep learning-based AI systems.

Artificial intelligence technology consists of machine learning and element technologies using machine learning. Machine learning is an algorithm technology that categorizes/learns the characteristics of input data by itself, and element technology uses machine learning algorithms such as deep learning to simulate functions such as cognition and judgment of the human brain. It consists of technical fields such as understanding, reasoning/prediction, knowledge expression, and motion control.

The various fields where artificial intelligence technology is applied are as follows. Linguistic understanding is a technology for recognizing and applying/processing human language/text, and includes natural language processing, machine translation, dialogue system, question and answer, and speech recognition/synthesis. Visual understanding is a technology for recognizing and processing objects like human vision, and includes object recognition, object tracking, image search, human recognition, scene understanding, spatial understanding, image improvement, and the like. Inference prediction is a technology for logically reasoning and predicting by judging information, and includes knowledge/probability-based reasoning, optimization prediction, preference-based planning, and recommendation. Knowledge expression is a technology that automatically processes human experience information into knowledge data, and includes knowledge construction (data generation/classification) and knowledge management (data utilization). Motion control is a technology for controlling autonomous driving of a vehicle and movement of a robot, and includes motion control (navigation, collision, driving), manipulation control (action control), and the like.

In general, in order to apply a machine learning algorithm to real life, learning is performed in a trial and error method due to the characteristics of the basic methodology of machine learning. In particular, deep learning requires hundreds of thousands of iterations. It is impossible to execute this in the actual physical external environment, so instead, the actual physical external environment is implemented on a computer and learning is performed through simulation.

On the other hand, in general, in the manufacturing process of a ship, the material details for each ship are classified and managed into attribute groups according to the function when manufacturing the ship. there is. Due to this, there is a problem in that an error of arranging the material in the wrong position in the subsequent process occurs, and unnecessary time is wasted to correct it.

Accordingly, the present invention provides a list classification system and method capable of converting a name input into a list and attribute information corresponding thereto into an ASCII code and forming a relationship between a name and a corresponding attribute as output data using a machine learning algorithm; There is provided a computer-readable recording medium in which a computer program for executing the method in a computer is recorded.

A list classification method according to an embodiment of the present invention includes: receiving learning data including a name and corresponding attribute information; encoding the training data into an ASCII code; and extracting a correlation between the name and the corresponding attribute information as output data using a machine learning algorithm for the encoded learning data.

In addition, the step of receiving a predetermined name input; and extracting attribute information corresponding to the input name by using the output data.

In addition, receiving a list including a predetermined name and corresponding attribute information as input; and detecting an error in the list in which the predetermined name and the corresponding attribute information are erroneously matched by using the output data.

In addition, the step of extracting the output data may include the step of receiving the ASCII code array of the encoded learning data as time series signal data and extracting the correlation between the name and the corresponding attribute information as output data. can

In addition, the name includes the material details for each ship used in the shipbuilding industry, and the corresponding attribute information may include an attribute group corresponding to the material details.

Also, the machine learning algorithm may include at least one of a K-Nearest Neighbor (KNN) algorithm, a weighted KNN algorithm, an ensemble bagging tree algorithm, and an ensemble subspace KNN algorithm.

A list classification system according to an embodiment of the present invention includes: a database for storing learning data including a name and corresponding attribute information; Receive the training data, encode the name included in the training data into an ASCII code, and output the correlation between the name and the corresponding attribute information using the encoded name using a machine learning algorithm It may include a server that extracts to .

Also, the server may operate to receive a predetermined name and extract attribute information corresponding to the inputted name using the output data.

In addition, the server may operate to receive a list including a predetermined name and corresponding attribute information, and use the output data to detect an error in the list in which the predetermined name and the corresponding attribute information are erroneously matched. there is.

Also, the server may operate to receive the ASCII code arrangement of the encoded learning data as time series signal data and extract a correlation between the name and the corresponding attribute information as output data.

In addition, the name includes the material details for each ship used in the shipbuilding industry, and the corresponding attribute information may include an attribute group corresponding to the material details.

Also, the machine learning algorithm may include at least one of a K-Nearest Neighbor (KNN) algorithm, a weighted KNN algorithm, an ensemble bagging tree algorithm, and an ensemble subspace KNN algorithm.

Meanwhile, in the computer-readable recording medium according to an embodiment of the present invention, a computer program for executing the above-described list classification method in a computer may be recorded.

According to the present invention, machine learning can be performed more easily by using an ASCII code as a time-series signal that is easy to classify, so that a correlation between a name included in a list and an attribute corresponding thereto can be easily grasped. In addition, more accurate and faster machine learning can be performed using relatively simple machine learning algorithms such as KNN (K-Nearest Neighbor) or ensemble, rather than complex algorithms such as natural language processing or deep learning. .

1 is an exemplary diagram showing the configuration of a list classification system according to an embodiment of the present invention.
2 is an exemplary diagram showing the configuration of a server according to an embodiment of the present invention.
3 is a flowchart illustrating a procedure of a list classification method according to an embodiment of the present invention.
4 is a diagram showing the probability distribution of learning data for features when the material details are expressed in the form of signals converted to ASCII codes without natural language processing.

Hereinafter, with reference to the accompanying drawings, the embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them. The present invention may be embodied in several different forms and is not limited to the embodiments described herein.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

1 is an exemplary diagram showing the configuration of a list classification system according to an embodiment of the present invention.

1, the list classification system 100 includes a plurality of user terminals 110-1, ..., 110-n, a server 120, a database 130, and a network (N). can According to an embodiment, the database 130 may be implemented in a cloud environment separately from the server 120 , but is not limited thereto, and the database 130 may be provided in the server 120 . For example, the plurality of user terminals 110-1, ..., 110-n, the server 120, and the database 130 may be connected to each other to be able to communicate with each other through the network N. According to an embodiment, the list classification system 100 receives the material details (names) and corresponding attribute group information for each line as a list, and uses a machine learning algorithm to extract the correlation between the name and the corresponding attribute. can

For example, the rate of abnormality by material category by ship line in the field can be shown as shown in Table 1.

all normal More than ratio(%) iron fittings 27,525 25,533 1,992 7.8 electrical equipment 713 491 222 45.2 plumbing material 29,107 24,273 4,834 19.9 cabin material 608 549 59 10.7 lagging 187 140 47 33.6 products 1,700 1,685 15 0.9

A plurality of user terminals (110-1, ..., 110-n) may form a list of material details by receiving information on material details (names) for each ship and attribute groups corresponding thereto from users in shipyards, etc. . In addition, a plurality of user terminals (110-1, ..., 110-n), to transmit the material details list formed in real time or non-real time to the server 120 and / or the database 130 via the network (N) The server 120 may extract a list of material details from a plurality of user terminals 110-1, ..., 110-n as learning data. According to one embodiment, the server 120, as shown in Table 2, may extract a list of material details for iron fittings of a specific line as learning data, but is not limited thereto.

A B One STANDARD DOOR CUSHION AJDC 2 COSMO STANDARD STOPPER SET AJDS ... ... ... 27524 BOTTOM PLUG PACKING LEAD 4.0T(+0.2~-0.2) OD63XID55 FRBP 27525 BOTTOM PLUG SOCKET D-GRADE 35T DSE-D2074S FRBP

For example, the training data of Table 2 may include the number of 27,520 pieces of data in which 5 pieces of noise information are deleted out of a total of 27,525 pieces of data, and the classification attribute group may include 219 pieces. In addition, the server 120 may encode a material detail (name) and a corresponding attribute group included in the extracted learning data into an ASCII code. An attempt was made to use the index value by extracting the keywords of the material description and attribute group using the existing encoding method, but the encoding method using the ASCII code is used for the following reasons.

First, there is no need to update the index when new words are added, natural language processing is excluded to minimize development resources, but this is to preserve the time-series characteristics of the language. Because it is advantageous to do (Machine Learning).

In addition, as shown in FIG. 4 , when the learning data is expressed in the form of a signal converted into an ASCII code without natural language processing, it can be confirmed that the feature map is distributed in a form advantageous for classification by machine learning.

Table 3 is an example diagram showing the result of encoding material details and corresponding attribute groups into ASCII codes.

A B C D E ... AA AB AC AD AE One 89 94 65 77 43 ... 54 2 47 78 83 74 56 ... 96 87 3 32 78 45 58 97 ... 64 58 72 4 68 32 79 42 52 ... 23 37 94 45 5 45 24 76 85 34 ... 65 87 92 23 38 ... ... ... ... ... ... ... ... ... ... ... ...

In addition, the server 120 may extract the correlation between the material details (name) and the attribute information corresponding thereto as output data by using the encoded learning data and a machine learning algorithm. According to an embodiment, the server 120 receives the ASCII code arrangement of the encoded learning data as input as time series signal data, and extracts the correlation between the material details (name) and the corresponding attribute information as output data. However, the present invention is not limited thereto. Table 4 shows the relationship between the machine learning algorithm used and the learning time required to complete the training and the prediction accuracy.

algorithm study time Prediction Accuracy Dense KNN 20 minutes (1,202 seconds) 96.3% weighted KNN 21 minutes (1,265 seconds) 96.1% Ensemble (Bagging Tree) 27 minutes (1,649 seconds) 95.9% Ensemble (subspace KNN) 28.6 minutes (1,717 seconds) 95.9%

According to the results shown in Table 4, it can be confirmed that there are a number of problems in which a machine learning model with a prediction accuracy of 95% or more can be developed only by data encoding without deep learning among the actual data in the field. The database 130 is the It can store various data for implementation. Data stored in the database 130, a plurality of user terminals (110-1, ..., 110-n), and data obtained, processed, or used by at least one component of the server 120 As such, it may include software (eg: a program). The database 130 may include volatile and/or non-volatile memory. In addition, the database 130 can be implemented in a cloud environment, so that the capacity of the server 120 is limited when data is accumulated, but the database 130 is not limited thereto, and cloud infrastructure and managed ) service-based, it can include various systems with high availability and high scalability. As an embodiment, the database 130 may be stored in the form of a mapping table by mapping the material details for each line and information on the attribute group corresponding thereto, but is not limited thereto.

In the present invention, the program is software stored in the database 130, a plurality of user terminals (110-1, ..., 110-n), and an operating system for controlling the resources of the server 120, applications and / Alternatively, the application may include middleware that provides various functions to the application so that the application can utilize the resources of the plurality of user terminals 110-1, ..., 110-n, and the server 120, and the like.

The network N may perform wireless or wired communication between a plurality of user terminals 110 - 1 , ... , 110 -n, the server 120 , and the database 130 . For example, the network N is a long-term evolution (LTE), LTE Advanced (LTE-A), code division multiple access (CDMA), wideband CDMA (WCDMA), Wireless BroadBand (WiBro), wireless fidelity (WiFi) , Bluetooth (Bluetooth), near field communication (NFC), GPS (Global Positioning System), or GNSS (global navigation satellite system) according to a method such as wireless communication can be performed. For example, the network N may perform wired communication according to a method such as universal serial bus (USB), high definition multimedia interface (HDMI), recommended standard 122 (RS-122), or plain old telephone service (POTS). may be

In the present invention, artificial intelligence (AI) refers to a technology that imitates human learning ability, reasoning ability, perceptual ability, etc., and implements it with a computer, and concepts such as machine learning and symbolic logic may include Machine Learning is an algorithm technology that classifies or learns characteristics of input data by itself. Artificial intelligence technology is an algorithm of machine learning that can analyze input data, learn the results of the analysis, and make judgments or predictions based on the results of the learning. In addition, technologies that use machine learning algorithms to simulate functions such as cognition and judgment of the human brain can also be understood as a category of artificial intelligence. For example, technical fields such as verbal comprehension, visual comprehension, reasoning/prediction, knowledge expression, and motion control may be included.

Machine learning may refer to the processing of training a neural network model using the experience of processing data. With machine learning, computer software could mean improving its own data processing capabilities. The neural network model is constructed by modeling the correlation between data, and the correlation may be expressed by a plurality of parameters. A neural network model extracts and analyzes features from given data to derive correlations between data, and repeating this process to optimize parameters of a neural network model can be called machine learning. For example, the neural network model may learn a mapping (correlation) between an input and an output with respect to data given as an input/output pair. Alternatively, the neural network model may learn the relationship by deriving regularity between the given data even when only input data is given.

The artificial intelligence learning model or neural network model may be designed to implement a human brain structure on a computer, and may include a plurality of network nodes that simulate neurons of a human neural network and have weights. A plurality of network nodes may have a connection relationship with each other by simulating a synaptic activity of a neuron through which a neuron sends and receives a signal through a synapse. In the AI learning model, a plurality of network nodes can exchange data according to a convolutional connection relationship while being located in layers of different depths. The artificial intelligence learning model may be, for example, an artificial neural network model, a convolutional neural network model, or the like. As an embodiment, the artificial intelligence learning model may be machine-learned according to methods such as supervised learning, unsupervised learning, reinforcement learning, ensemble learning, and the like. Machine learning algorithms for performing machine learning include K-Nearest Neighbor (KNN) algorithm, Decision Tree, Bayesian Network, Support Vector Machine, Artificial Neural Network, Ada-boost, Perceptron, Genetic Programming, Clustering, etc. may be used.

Among them, the KNN algorithm is a type of instance-based learning or lazy learning, in which functions are approximated only locally and all computations can be deferred until function evaluation. Because KNN algorithms rely on distance for classification, normalizing the training data can significantly improve accuracy. A useful technique for both classification and regression is to assign weights to the contributions of neighbors, such that close neighbors contribute more to the mean than more distant neighbors. For example, a common weight scheme could consist of giving each neighbor a weight of 1/d. where d is the distance to the neighbor. Neighbors can come from a set of objects whose classes (for KNN classification) or object attribute values (for KNN regression) are known. It does not require an explicit training step, but can be thought of as a training set for the algorithm. A characteristic of the KNN algorithm is that it is sensitive to the local structure of the data.

CNNs are a type of multilayer perceptrons designed to use minimal preprocessing. CNN consists of one or several convolutional layers and general artificial neural network layers on top of it, and additionally utilizes weights and pooling layers. Thanks to this structure, CNN can fully utilize the input data of the two-dimensional structure. Compared with other deep learning structures, CNN shows good performance in both video and audio fields. CNNs can also be trained through standard back-passing. CNNs are easier to train than other feed-forward neural network techniques and have the advantage of using fewer parameters.

Convolutional networks are neural networks that contain sets of nodes with bound parameters. The increasing size of available training data and the availability of computational power, coupled with advances in algorithms such as piecewise linear units and dropout training, have greatly improved many computer vision tasks. For huge data sets, such as those available for many tasks today, overfitting is not important, and increasing the size of the network improves test accuracy. Optimal use of computing resources becomes a limiting factor. To this end, a distributed, scalable implementation of deep neural networks may be used.

2 is an exemplary diagram showing the configuration of a server according to an embodiment of the present invention.

As shown in FIG. 2 , the server 120 may include a receiver 121 , a processor 122 , a transmitter 123 , a system bus 124 , a digital packet 125 , and a database 130 . As an embodiment, the receiver 121 , the processor 122 , the transmitter 123 , the digital packet 125 , and the database 130 may be connected to each other so as to be able to communicate using the system bus 124 , and the server 120 . ), at least one of these components may be omitted, or another component may be added to the server 120 . In addition, additionally (additionally) or alternatively (alternatively), some of the components may be implemented as integrated, or may be implemented as a singular or a plurality of entities.

The receiving unit 121, a list of material details including information on material details (names) and corresponding attribute groups from a plurality of user terminals 110-1, ..., 110-n through a network N may be received in the form of a digital packet 125 and transmitted to the processor 122 and the database 130 .

The processor 122 may operate to perform encoding in ASCII codes using the received material detail list as training data. According to an embodiment, the processor 122 may receive an ASCII code arrangement of the encoded learning data as time series signal data, but is not limited thereto.

In addition, the processor 122 may operate to extract the correlation between the material description (name) and the attribute group information corresponding thereto as output data using the encoded learning data using a machine learning algorithm. According to an embodiment, the processor 122 receives the ASCII code arrangement of the encoded learning data as time series signal data, and extracts the correlation between the material details (name) and the corresponding attribute group information as output data. can, but is not limited thereto. Here, the machine learning algorithm may include, but is not limited to, a dense KNN (K-Nearest Neighbor) algorithm, a weighted KNN algorithm, an ensemble bagging tree algorithm, an ensemble subspace KNN algorithm, and the like. Here, the ensemble machine learning algorithm is a support vector machine, a hidden Markov model, a regression, a neural network, a naive Bayes classification, etc. of supervised learning algorithms and k-means clustering, hierarchical clustering, Bayesian network, Linde-Buzo-Gray algorithm, principal components analysis, auto An algorithm for performing machine learning by combining various algorithms among unsupervised learning algorithms such as an encoder (autoencoder) may be represented.

The processor 122 receives a predetermined name from any one user terminal (eg, 110-1) among the plurality of user terminals 110-1, ..., 110-n, and attributes corresponding to the input name. It may operate to extract attribute information corresponding to the input name by using the output data indicating the correlation between the information.

In addition, the processor 122 receives a list including a name and corresponding attribute information, and uses the output data indicating a correlation between the name and the corresponding attribute information, a list in which the name and the corresponding attribute information are erroneously matched can operate to detect errors in

Transmission unit 123, the attribute information corresponding to the input name in the form of the digital packet 125 formed in the processor 122, the name and the corresponding attribute information are erroneously matched the list error through the network (N) database ( 130) and/or a plurality of user terminals 110-1, ..., 110-n.

3 is a flowchart illustrating a procedure of a list classification method according to an embodiment of the present invention. Although process steps, method steps, algorithms, etc. are described in a sequential order in the flowchart of FIG. 3 , such processes, methods, and algorithms may be configured to operate in any suitable order. In other words, the steps of the processes, methods, and algorithms described in various embodiments of the invention need not be performed in the order described herein. Also, although some steps are described as being performed asynchronously, in other embodiments some of these steps may be performed concurrently. Further, the exemplification of a process by description in the drawings does not imply that the exemplified process excludes other changes and modifications thereto, and that the illustrated process or any of its steps may be used in any of the various embodiments of the present invention. It is not meant to be essential to one or more, nor does it imply that the illustrated process is preferred.

As shown in FIG. 3 , in step S310 , learning data including a name and corresponding attribute information is received. For example, referring to FIGS. 1 and 2 , a plurality of user terminals 110-1, ..., 110-n of the list classification system 100 includes material details for each line and attribute group information corresponding thereto. It is possible to input a list of material details including

In step S320, the training data is encoded in ASCII code. For example, referring to FIGS. 1 and 2 , the server 120 of the list classification system 100 may encode a name and corresponding attribute information included in the learning data into an ASCII code.

In step S330 , the correlation between the name and the corresponding attribute information is extracted as output data by using the encoded learning data with a machine learning algorithm. For example, referring to FIGS. 1 and 2 , the server 120 of the list classification system 100 uses a machine learning algorithm to convert the encoded name and the corresponding attribute information into a correlation between the name and the corresponding attribute information. Relationships can be extracted as output data. According to one embodiment, the server 120 receives the ASCII code arrangement of the encoded learning data as time series signal data, and extracts the correlation between the material details (name) and the corresponding attribute group information as output data. can, but is not limited thereto.

Through the list classification method and system according to an embodiment of the present invention, machine learning can be performed by receiving the ASCII code arrangement as time series signal data so that it is possible to learn information about language continuity without a complex natural language processing algorithm, When material details are expressed as a time series signal with ASCII code values, it can be a time series signal that is intuitively easy to classify, so it is advantageous for machine learning. In addition, machine learning with a high level of prediction accuracy is possible without special feature engineering in the KNN algorithm and ensemble algorithm, which are basic machine learning algorithms.

In the above, the present invention has been described with reference to the embodiments shown in the drawings. However, the present invention is not limited thereto, and various modifications or other embodiments within the scope equivalent to the present invention are possible by those of ordinary skill in the art to which the present invention pertains. Accordingly, the true scope of protection of the present invention should be defined by the following claims.

100: List classification system 110-1, … ,110-n: user terminal
120: server 130: database
121: receiver 122: processor
123: transmitter 124: system bus
125: digital packet N: network

Claims (13)

receiving learning data including a name and corresponding attribute information;
encoding the training data into an ASCII code; and
extracting the correlation between the name and the corresponding attribute information as output data by using the encoded learning data with a machine learning algorithm,
How to classify a list. The method of claim 1,
receiving a predetermined name; and
Further comprising the step of extracting attribute information corresponding to the input name by using the output data,
How to classify a list. The method of claim 1,
receiving a list including a predetermined name and corresponding attribute information; and
detecting an error in the list in which the predetermined name and the corresponding attribute information are incorrectly matched by using the output data;
How to classify a list. The method of claim 1,
The step of extracting the output data is,
Including the step of receiving the ASCII code array of the encoded learning data as input as time series signal data and extracting the correlation between the name and the corresponding attribute information as output data,
How to classify a list. The method of claim 1,
The name includes the material details for each ship used in the shipbuilding industry,
The corresponding attribute information includes an attribute group corresponding to the material details,
How to classify a list. The method of claim 1,
The machine learning algorithm is
Including any one of a dense KNN (K-Nearest Neighbor) algorithm, a weighted KNN algorithm, an ensemble bagging tree algorithm, and an ensemble subspace KNN algorithm,
How to classify a list. a database for storing learning data including names and corresponding attribute information;
Receive the learning data, encode a name included in the learning data into an ASCII code, and output a correlation between the name and the corresponding attribute information using the encoded learning data using a machine learning algorithm Including a server to extract data,
List classification system. 8. The method of claim 7,
The server is
receiving an input of a predetermined name and operating to extract attribute information corresponding to the inputted name using the output data;
List classification system. 8. The method of claim 7,
The server is
receiving a list including a predetermined name and corresponding attribute information, and operating to detect an error in the list in which the predetermined name and the corresponding attribute information are incorrectly matched by using the output data;
List classification system. 8. The method of claim 7,
The server is
receiving the ASCII code array of the encoded learning data as time series signal data and operating to extract the correlation between the name and the corresponding attribute information as output data,
List classification system. 8. The method of claim 7,
The name includes the material details for each ship used in the shipbuilding industry,
The corresponding attribute information includes an attribute group corresponding to the material details,
List classification system. 8. The method of claim 7,
The machine learning algorithm is
Including any one of a dense KNN (K-Nearest Neighbor) algorithm, a weighted KNN algorithm, an ensemble bagging tree algorithm, and an ensemble subspace KNN algorithm,
List classification system. A computer program for executing the list classification method according to any one of claims 1 to 6 on a computer is recorded, a computer readable recording medium.

Images