KR101560726B1 - System and method for generating cluster using seed based link - Google Patents

Abstract

A cluster creation system and method using a seed according to a link is disclosed. A cluster generating system includes a seed determining unit for determining a seed composed of objects of similarity or more than a predetermined degree set on the basis of link information from a plurality of objects linked by a link and a cluster generating unit for generating a cluster using the determined seed . It is possible to shorten the initial execution time for generating a cluster by determining a seed having a high degree of similarity and creating a cluster.

Cluster, link, seed, similarity, tree, object, hash base

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system and a method for generating a cluster using a seed according to a link,

The present invention relates to a cluster creation system and method using a seed according to a link, and more particularly, to a cluster creation system and method for creating a cluster using a hash-based search method.

Clustering means grouping similar objects into clusters. Clustering research has been studied for a long time in various fields such as statistics, databases, and data mining. Through the clustering process, it can be applied to services such as advertisement and search to improve the accuracy of the service and to easily understand the tendency of similarity of objects.

In recent years, research has been focused on creating clusters for objects with high similarity using a large amount of data existing on the Internet during the clustering process. In particular, link-based clustering has been emerging that clusters objects considering links between objects.

However, as the use of the Internet increases, creating a cluster for a large number of objects on the Internet requires a lot of execution time. As the execution time increases, resources for clustering are consumed. That is, there is a demand for a method for rapidly performing processing while reducing the execution time while ensuring similarity in preparation for a large-capacity data environment.

As a result, it is necessary to accurately and efficiently cluster large volumes of data by improving the bottleneck that affects overall performance for clustering for large amounts of data.

The present invention provides a cluster generation system and method for shortening the execution time required for clustering by determining clusters using a seed and a few few objects with a very high similarity among a plurality of objects linked by a link.

The present invention creates a hash structure from transaction data for a plurality of objects linked by a link, extracts a pattern exceeding a minimum overflow while searching for a hash structure, determines the seed as a seed, A cluster creation system and method capable of reducing the number of cluster creation processes.

The present invention provides a cluster generation system and method capable of improving similarity of clusters to be generated by predicting an object having a small number of links as noise and removing it from transaction data through a preprocessing process for cluster creation.

The present invention provides a cluster creation system and method capable of easily grasping the similarity among a plurality of objects by generating a tree that hierarchically expresses the similarities between the objects through the clusters generated from the plurality of objects and the plurality of objects do.

A cluster generation system according to an embodiment of the present invention includes a seed determination unit for determining a seed composed of objects of similarity or more than a predetermined degree set on the basis of link information from a plurality of objects linked by a link, And a cluster generating unit for generating a cluster generating unit.

According to an aspect of the present invention, the seed determining unit may include a transaction data generating unit for generating transaction data using a plurality of objects linked by the link, a search unit for searching the transaction data to determine a hash structure composed of candidate patterns based on the pattern length And a seed extracting unit for extracting, from the candidate patterns constituting the hash structure, an object corresponding to a candidate pattern exceeding the lowest frequent occurrence, by a seed.

According to an aspect of the present invention, the hash structure determination unit may include a candidate pattern generation unit that generates a candidate pattern corresponding to the pattern length from the transaction data based on a pattern length according to the number of items, And a frequent determining unit that determines a frequent degree by counting the candidate patterns.

According to an aspect of the present invention, the seed determining unit may further include a noise removing unit for removing an object less than a predetermined number of links from the transaction data by determining the object as noise.

The cluster creation system according to an embodiment of the present invention sets the plurality of objects as low-level nodes for each type and sets a cluster generated from the plurality of objects as a high-level node to generate a structured tree And a tree generating unit.

A cluster generation method according to an embodiment of the present invention includes the steps of: determining a seed composed of a plurality of objects connected in a link and having a degree of similarity set in advance on the basis of link information, and generating a cluster using the determined seed .

According to an aspect of the present invention, the step of determining a seed composed of the objects having the similarity or greater in number is a step of generating transaction data using a plurality of objects linked by the link, And a step of extracting an object corresponding to a candidate pattern exceeding the lowest frequent occurrence among the candidate patterns constituting the hash structure by a seed.

According to an aspect of the present invention, the step of determining the seed may further include removing a predetermined number or less of the objects from the transaction data by judging the number of the links as noise.

A cluster generation method according to an embodiment of the present invention sets the plurality of objects as low-level nodes for each type and sets a cluster generated from the plurality of objects as a high-level node to generate a structural tree Step < / RTI >

According to an embodiment of the present invention, a small number of objects having a very high degree of similarity among a plurality of objects linked by a link are determined as a seed, and a cluster is created using a seed, thereby shortening the execution time required for clustering.

According to an embodiment of the present invention, a hash structure is generated from transaction data for a plurality of objects connected by a link, a pattern exceeding a minimum frequency is extracted while searching for a hash structure, It is possible to reduce the cluster creation time.

According to an embodiment of the present invention, a similarity of a cluster to be generated can be improved by eliminating an object having a small number of links as noise and removing it from transaction data through a preprocessing process for cluster creation.

According to an embodiment of the present invention, a similarity degree among a plurality of objects can be easily grasped by generating a tree that hierarchically expresses a degree of similarity between objects through a cluster generated from a plurality of objects and a plurality of objects have.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to or limited by the embodiments. Like reference symbols in the drawings denote like elements. A cluster creation method according to an embodiment of the present invention may be performed by a cluster creation system.

FIG. 1 is a diagram for explaining an entire process performed by a cluster creation system according to an embodiment of the present invention.

The cluster creation system 100 may cluster a plurality of linked objects 101 and 102 according to the link-based similarity to generate a structured tree 103. [ Clustering can mean grouping similar objects into specific clusters. At this time, the object 101 and the object 102 connected by a link may be data representing different types. For example, object 101 may be a blog, and object 102 may be a post linked by a link through a scrap. In FIG. 1, two kinds of objects representing different types are shown, but the types are not limited.

It has become important to effectively cluster large volumes of data while the amount of objects (data) that are the object of the cluster continuously increases. Since each object is linked through a link, the amount of computation that must be processed for clustering increases as the amount of data increases.

In contrast, the cluster creation system 100 according to an embodiment of the present invention can create a hash structure using transaction data composed of objects linked by a link, and generate a cluster through a hash-based seed exploration method . Specifically, the cluster creation system 100 determines seeds of a very small number of highly similar objects in consideration of the fact that clusters are created by a small number of highly similar objects 101 and 102, By including seeds 101 and 102 in the seed, the initial clustering can be efficiently and quickly processed.

In addition, the cluster creation system 100 may shorten the clustering execution time by repeatedly removing the objects (101, 102) corresponding to the noise from the transaction data through preprocessing before creating the cluster. The cluster creation system 100 generates clusters for the objects 101 and 102 through the above process and generates the trees 103 and 104 for each type using the generated clusters, thereby reducing the accuracy Link-based clustering can be performed to reduce the execution time without having to perform link-based clustering.

2 is a block diagram illustrating the overall configuration of a cluster creation system according to an embodiment of the present invention.

Referring to FIG. 2, the cluster creation system 100 may include a seed determination unit 201, a cluster generation unit 202, and a tree generation unit 203.

The seed determining unit 201 can determine a seed composed of objects of similarity or more than a predetermined degree set on the basis of the link information from the plurality of objects 101 and 102 linked by a link. According to an embodiment of the present invention, the seed may be a reference object for generating an extended cluster. At this time, the plurality of objects 101 and the plurality of objects 102 have different types, and the seed determining unit 201 can determine the seed for each of the objects 101 and 102 of different types.

The cluster creation system 100 according to an embodiment of the present invention can reduce the time for generating clusters from the objects 101 by roughly constructing an initial cluster corresponding to the seed. In this case, the seed may include a small number of objects having a very high degree of similarity among the plurality of objects 101.

For example, the seed determining unit 201 may include a transaction data generating unit 204, a noise removing unit 205, a hash structure determining unit 206, and a seed extracting unit 207. Specifically, the seed determining unit 201 can determine a seed from a plurality of objects 101 and 102 of different types through a hash-based seed exploration method.

The transaction data generation unit 204 may generate transaction data using a plurality of objects 101 linked by a link. For example, the transaction data generation unit 204 may set the transaction type of the objects 102 of the different types indicated by the plurality of objects 101 on the link.

The transaction data generation unit 204 may generate the transaction data by classifying the plurality of objects 101 into items constituting the set transaction according to the link. That is, the transaction data generation unit 204 may convert the objects 101 and 102 linked by the link into transaction data. A specific example of generating transaction data is described in FIG.

The noise removing unit 205 may remove noise from the transaction data by determining that the number of links is equal to or less than the predetermined number. Certain objects 101 may be associated with other types of objects 102 through a small number of links. The cluster is made up of highly similar objects 101, and the similarity is related to the number of links. If the number of links is small, the reliability of the similarity between objects can not be guaranteed. Thus, when an object having a small number of links is included in the cluster, the accuracy of the cluster can be reduced. In addition, objects with a small number of links increase the number of data to be processed for clustering, which increases clustering execution time.

Therefore, the cluster generation system 100 according to an embodiment of the present invention determines the objects 101 having a small number of links as noise, and removes the objects 101 in advance through the preprocessing process before generating the clusters, The similarity between objects can be improved and the clustering time can be shortened.

For example, the noise remover 205 may repeatedly remove objects of different types on a per type basis. For example, it is assumed that the object 101 (A) corresponding to the type 1 is linked to the objects 102 (a, b) corresponding to the type 2. The noise removing unit 205 can remove objects having one link by judging them as noise.

At this time, when the number of links of the object a 102 is one, the noise removing unit 203 can remove the object a 102 from the cluster. As object a 102 is removed, the number of links of object A 102 may also be reduced. Then, since the number of links of the object A 101 corresponding to the type 1 is changed from two to one, the noise removing unit 203 can also remove the object A 101 from the cluster by judging it as noise. As a result, the cluster creation system 100 according to an embodiment of the present invention can repeatedly remove the noise corresponding to the noise until the object corresponding to the noise does not exist in the cluster.

The hash structure determination unit 206 can determine the hash structure composed of the candidate patterns based on the pattern length by searching the generated transaction data. At this time, the transaction data can search through the noise removal unit 205 for the transaction data from which objects judged as noise are removed. The pattern length may be the number of objects included in the object pair included in the transaction data. For example, the hash structure determination unit 206 may include a candidate pattern generation unit and a frequentness determination unit.

The candidate pattern generator may generate a candidate pattern corresponding to the pattern length from the transaction data based on the pattern length according to the number of items. That is, the candidate pattern generator may search the transaction data to generate a candidate pattern corresponding to the pattern length.

The frequent occurrence determining unit may determine a frequency of occurrence of the candidate pattern in the transaction data by counting the candidate pattern from the transaction data. As a result, the hash structure determination unit 206 can determine the hash structure using the frequentities corresponding to the candidate pattern and the candidate pattern, respectively.

The seed extracting unit 207 can extract an object corresponding to a candidate pattern exceeding the lowest frequen- cy among the candidate patterns constituting the hash structure as a seed. That is, the seed extractor 207 sets the lowest frequen- cies that can guarantee a degree of similarity equal to or higher than a predetermined level, and extracts an object corresponding to a candidate pattern exceeding the lowest frequen- cy among the plurality of objects 101 as a seed .

Therefore, the seed extracting unit 207 can quickly determine a small number of highly similar objects from the plurality of objects 101 by searching for a hash structure composed of one candidate pattern.

The cluster generating unit 202 may generate a cluster using the seed determined through the seed determining unit 201. [ Clusters can be created for each type of object.

For example, the cluster generating unit 202 may initially generate a cluster composed of objects corresponding to a seed. The cluster generating unit 202 may extract the objects that are frequently included in the same transaction as the seed for the objects that are not included in the cluster among the plurality of objects and may add them to the cluster. Specifically, the cluster generating unit 202 may perform a process of adding an object to a cluster until the number of objects constituting the cluster becomes a preset number.

The tree generation unit 203 sets a plurality of objects 101 and 102 as low-level nodes for each type, sets clusters generated from a plurality of objects as high-level nodes, Lt; / RTI > That is, a tree X 103 may be generated for a plurality of objects 101, and a tree Y 103 may be generated for a plurality of objects 102.

At this time, each of the plurality of objects 101 and 102 is determined as the end node in the tree 103 and 104, and the cluster can be determined as the non-end node. In addition, the cluster creation system 100 can set the initially determined cluster as an object through the cluster creation process, and create a new cluster from the corresponding cluster. That is, the cluster creation system 100 can generate the structured tree 103 and 104 according to the level for the plurality of objects 101 and 102 and the cluster by repeating the cluster creation process.

3 is a diagram for explaining transaction data based on objects linked by a link according to an embodiment of the present invention.

Referring to FIG. 3, it can be seen that a plurality of objects 301 and 302 of different types are linked by a link. For example, in FIG. 3, it is assumed that a plurality of objects 301 refers to a blog and a plurality of objects 302 refers to a post.

The cluster creation system can determine a seed composed of objects of similarity or more based on link information from a plurality of objects 301 and 302 linked by a link. Such similarity can be determined according to the link of the object. Then, the seed can be extracted from each of a plurality of objects 301 and 302 of different types.

The cluster creation system can generate transaction data using a plurality of linked objects (301, 302). In one example, the cluster creation system may set up a transaction as a different type of objects 302, each of which is indicated by a plurality of objects 301 through a link. The cluster generating system can classify a plurality of objects 301 into items constituting a transaction according to a link to generate transaction data.

In FIG. 3, when a plurality of post data is set as a transaction, the cluster creation system can classify the blog data linked to each post data as items 303 for the transaction. For example, since the blog data B1, B3, and B4 are linked to the post data P3, the items for the transaction P3 can be B1, B3, and B4.

Through this process, the blog data linked to each of the post data P1 to P8 set in the transaction is classified as the item 303, and transaction data can be generated. Conversely, when a plurality of blog data is set as a transaction, the cluster creation system may classify the post data linked to each blog data as an item for a transaction.

The cluster generation system can generate the transaction data using a plurality of linked objects to grasp the patterns of the objects that occur frequently. For example, the blog data B1 and B3 can be said to be similar to each other because the post data P1 and P3 are commonly pointed through the link. Also, because B1 and B3 are frequently included in the same transactions P1 and P3, the cluster creation system can set B1 and B3 in a pattern that occurs frequently.

In addition, the cluster creation system can remove noise from the generated transaction data by judging that the number of links is less than the predetermined number. For example, the cluster creation system can repeatedly remove objects determined to be noise by each type. The noise removal process can be repeated until there is no object corresponding to the noise in the transaction data.

As a result, the cluster creation system can create clusters more quickly because it does not need to create clusters for objects judged to be noise by removing objects with low links before creating clusters. And, the similarity is related to the number of links. That is, the cluster generation system can improve the similarity between the objects included in the cluster and shorten the clustering time by eliminating the objects that are less than the preset number by judging the noise as the noise.

4 is a diagram illustrating a process of determining a hash structure from transaction data according to an embodiment of the present invention.

Referring to FIG. 4, the cluster creation system may search the transaction data 401 to generate a hash structure. In one example, the cluster creation system may search the transaction data 401 to determine a hash structure 402 configured with a candidate pattern based on the pattern length. At this time, the transaction data 401 may indicate that the object determined as noise has been removed. That is, the cluster creation system can search the transaction data 401 from which the object judged as noise has been removed to generate the hash structure.

Specifically, the cluster generation system may generate a candidate pattern corresponding to the pattern length from the transaction data based on the pattern length according to the number of items constituting the transaction data 401. [ The cluster generation system can determine the frequent occurrence by counting candidate patterns from the transaction data.

It is assumed that the pattern length is set to 2 in Fig. Then, the cluster creation system searches the transaction data 401 to generate a candidate pattern having a pattern length of 2 for each transaction. At this time, since the transactions P1, P3, P4, P6, and P8 include two or more items, a candidate pattern having a pattern length of 2 can be generated. A candidate pattern with a pattern length of 2 means that two objects are paired.

The cluster creation system searches the transaction data 401 to generate candidate patterns {B1, B2}, {B1, B3}, {B1, B4}, {B3, B4}, {B4, B5} having a pattern length of 2 can do. The cluster creation system can determine the frequent occurrence by counting candidate patterns in the transaction data 401. In the transaction data 401, the frequency of each candidate pattern is 1, 2, 1, 1, 2.

If the lowest frequency is set to 1, the cluster generating system can determine the seed pattern as a candidate pattern exceeding the minimum frequency. That is, the cluster generation system can determine the candidate patterns {B1, B3} and {B4, B5} as seeds. The pattern length and the lowest frequent pattern set in FIG. 4 can be changed according to the configuration of the system.

5 is a diagram illustrating a process of creating a cluster and adding an object according to an embodiment of the present invention.

The cluster generation system can extract an object corresponding to a candidate pattern exceeding the lowest frequen- cy among the candidate patterns constituting the hash structure as a seed. Then, the cluster creation system can create the clusters using the seed, which is an object determined to be similar to each other through the link information. Also, the cluster creation system can add to the cluster objects that are not included in the cluster and appear frequently in the same transaction as the seed.

Referring to the example of FIG. 4, the candidate patterns {B1, B3}, {B4, B5} exceeding the lowest frequent fly 1 can be determined as seeds, respectively. Then, as shown in the reference numeral 501 in FIG. 5, B1 and B3 can be determined as a cluster X, which is a group, and B4 and B5 can be determined as a cluster Y, which is another group. Then, as can be seen at 502 in FIG. 5, the cluster creation system may add B2 contained in the same transaction as B1, B3 to cluster X.

As a result, the cluster generation system can determine a seed of a few objects having a very high degree of similarity among a plurality of objects by exploring a hash structure based on transaction data. In addition, the cluster generating system can perform clustering using a plurality of objects by creating clusters using a seed, and adding objects frequently occurring in the same transaction as the objects included in the clusters to the clusters.

6 is a diagram illustrating a tree composed of objects and clusters according to an embodiment of the present invention.

FIG. 6 shows a tree in which clusters are generated from a plurality of objects of different types, and a plurality of objects and clusters are constructed. In the case of level 0, which is the end node of the tree, it consists of a plurality of objects. In the case of the non-terminal node of the level 1, which is a higher level, the cluster is formed of a plurality of objects of level 0. The non-terminal node of the upper level 2 is composed of the clusters generated by considering the cluster of level 1 as an object again. Through this process, the tree can be expanded.

A plurality of objects corresponding to level 0 in the trees of different types can be connected to each other via links. The similarity between nodes can be determined through such a link. For example, a tree can store the degree of similarity between child nodes belonging to the same parent node. The similarity between child nodes not belonging to the same parent node can be calculated through the similarity between the parent nodes of the child nodes.

That is, similarity between all the nodes in the tree is not calculated but is calculated through the hierarchical structure of the tree, so that the similarity of the nodes included in the tree can be calculated quickly. Based on the finally calculated similarity, the tree can be refined to nodes with similarity.

FIG. 7 is a flowchart illustrating an entire process of a cluster generation method according to an embodiment of the present invention.

The cluster creation system may determine a seed composed of objects that are similar or higher than a predetermined degree set on the basis of link information from a plurality of objects linked by a link (S701).

For example, the cluster creation system can determine the seed through a hash structure based search method. Specifically, the cluster creation system may generate transaction data using a plurality of linked objects (S704). At this time, the cluster creation system sets other types of objects indicated by links of the plurality of objects as transactions, classifies the plurality of objects as items constituting the set transaction according to the link, and generates transaction data can do.

The cluster creation system determines that the number of links is equal to or less than a predetermined number, and removes the objects from the transaction data through a preprocessing step (S705). The similarity between objects increases as the number of common links increases. However, when an object having a small number of links is included in a cluster, the degree of similarity of clusters can be reduced. Therefore, the cluster generation system can determine the object having a small number of links as noise through the preprocessing process before creating the cluster. And, the cluster creation system can improve the similarity of clusters generated later by removing objects judged as noise from the transaction data. At this time, the number of links can be changed according to the configuration of the system.

For example, the cluster creation system can iteratively remove objects of different types on a per type basis. That is, since objects are linked with objects of different types, when one object is removed, the number of links of other types of objects connected to the object and the link can also be reduced.

Then, the cluster creation system searches the transaction data to determine a hash structure composed of candidate patterns based on the pattern length (S706). Here, the transaction data may be one in which the object determined as noise is removed. In one example, the cluster creation system may generate a candidate pattern corresponding to the pattern length from the transaction data based on the pattern length according to the number of items. Thereafter, the cluster creation system may determine the frequency of occurrence by counting candidate patterns from the transaction data. Finally, a hash structure composed of candidate patterns and frequent patterns for candidate patterns can be determined.

The cluster creation system can extract an object corresponding to a candidate pattern exceeding the lowest frequen- cy among the candidate patterns constituting the hash structure by seeding (S707).

The cluster creation system may create a cluster using the seed (S702). For example, a cluster creation system may create clusters using seeds, and may add objects that are not included in the cluster and appear frequently in the same transaction as the seed, to the cluster. That is, in the cluster creation system according to an embodiment of the present invention, a seed, which is a small number of highly similar objects, is determined to create a cluster, and an initial execution time for generating a cluster can be shortened by expanding the cluster through the seed have.

In step S703, the cluster creation system sets up a plurality of objects as low-level nodes for each type, and sets a cluster generated from the plurality of objects as high-level nodes in step S703.

In this case, in the case of the level 0 which is the end node of the tree, it is composed of a plurality of objects. In the case of the non-terminal node of the level 1, which is a higher level, the cluster is formed of a plurality of objects of level 0. The non-terminal node of the upper level 2 is composed of the clusters generated by considering the cluster of level 1 as an object again. Through this process, the tree can be expanded.

The parts not described in FIG. 7 can be referred to the description of FIG. 1 to FIG.

The cluster creation method according to an exemplary embodiment of the present invention includes a computer readable medium including program instructions for performing various computer-implemented operations. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The media may be program instructions that are specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, Modification is possible. Accordingly, the spirit of the present invention should be understood only in accordance with the following claims, and all equivalents or equivalent variations thereof are included in the scope of the present invention.

FIG. 1 is a diagram for explaining an entire process performed by a cluster creation system according to an embodiment of the present invention.

2 is a block diagram illustrating the overall configuration of a cluster creation system according to an embodiment of the present invention.

3 is a diagram for explaining transaction data based on objects linked by a link according to an embodiment of the present invention.

4 is a diagram illustrating a process of determining a hash structure from transaction data according to an embodiment of the present invention.

5 is a diagram illustrating a process of creating a cluster and adding an object according to an embodiment of the present invention.

6 is a diagram illustrating a tree composed of objects and clusters according to an embodiment of the present invention.

FIG. 7 is a flowchart illustrating an entire process of a cluster generation method according to an embodiment of the present invention.

Description of the Related Art

100: Cluster creation system

101: Object <Type 1> 102: Object <Type 2>

103: Tree <Type 1> 104: Tree <Type 2>

Claims (19)

A seed determining unit for determining a seed composed of objects from a plurality of objects linked by a link based on link information based on the number of linked links, the objects having similarities set in advance to a degree of similarity set in advance; And A cluster generating unit for generating a cluster using the determined seed, &Lt; / RTI &gt; The method according to claim 1, The seed determination unit A transaction data generation unit for generating transaction data using a plurality of objects linked by the link; A hash structure determination unit for searching the transaction data to determine a hash structure composed of candidate patterns based on a pattern length; And A seed extracting unit for extracting, from the candidate patterns constituting the hash structure, an object corresponding to a candidate pattern exceeding the lowest frequen- cy, &Lt; / RTI &gt; 3. The method of claim 2, Wherein the transaction data generation unit comprises: And sets the plurality of objects to a transaction in accordance with the link and classifies the plurality of objects into items constituting the set transaction to generate the transaction data Cluster creation system. The method of claim 3, The hash structure determination unit may determine, A candidate pattern generator for generating a candidate pattern corresponding to the pattern length from the transaction data based on a pattern length according to the number of items; And A frequent determining unit for determining the frequent occurrence rate by counting the generated candidate patterns from the transaction data, &Lt; / RTI &gt; 5. The method of claim 4, The hash structure determination unit may determine, Wherein the hash structure is determined using the candidate patterns and the frequentities of the candidate patterns. 3. The method of claim 2, Wherein, A cluster is created using the extracted seed, and objects added to the cluster that are frequently included in the same transaction as the seed among the objects not included in the cluster are added to the cluster. 3. The method of claim 2, The seed determination unit And a noise removing unit for removing a predetermined number or less of the objects from the transaction data, Further comprising: The hash structure determination unit may determine, And searches the transaction data from which the object determined as the noise is removed. 8. The method of claim 7, Wherein the noise eliminator comprises: And repeatedly removing objects of different types for each type. The method according to claim 1, A tree generating unit configured to set the plurality of objects as lower level nodes for each type and to set a cluster generated from the plurality of objects as higher level nodes, Further comprising: &lt; / RTI &gt; A cluster creation method performed by a cluster creation system, Determining a seed comprising a plurality of objects connected to a seed determination attaching link included in the cluster creation system and including objects having a degree of similarity higher than a degree of similarity based on the number of linked links based on link information; And Wherein the cluster generating unit included in the cluster generating system generates a cluster using the determined seed &Lt; / RTI &gt; 11. The method of claim 10, Wherein the step of determining a seed, which is composed of objects having the predetermined similarity or more, Generating transaction data using a plurality of objects linked by the link; Searching the transaction data to determine a hash structure composed of candidate patterns based on the pattern length; And Extracting an object corresponding to a candidate pattern exceeding a minimum frequen- cy among the candidate patterns constituting the hash structure with a seed; &Lt; / RTI &gt; 12. The method of claim 11, Wherein the step of generating the transaction data comprises: Wherein each of the plurality of objects sets other types of objects indicated by links on the transaction as a transaction and generates the transaction data by classifying the plurality of objects into items constituting the set transaction according to the link How to create a cluster. 13. The method of claim 12, Wherein the step of determining a hash structure composed of candidate patterns based on the pattern length comprises: Generating a candidate pattern corresponding to the pattern length from the transaction data based on a pattern length according to the number of the items; And Counting the generated candidate patterns from the transaction data and determining a frequent occurrence rate &Lt; / RTI &gt; 14. The method of claim 13, Wherein the step of determining a hash structure composed of candidate patterns based on the pattern length comprises: Wherein the hash structure is determined using the candidate patterns and the frequentities of the candidate patterns. 12. The method of claim 11, Wherein the generating the cluster comprises: Generating a cluster using the extracted seed; And Adding to the cluster objects that are frequently included in the same transaction as the seed among the objects not included in the cluster &Lt; / RTI &gt; 13. The method of claim 12, The step of determining the seed may comprise: Determining that the number of links is equal to or less than a predetermined number as noise, and removing the objects from the cluster Further comprising: Wherein the determining the hash structure comprises: And searching for transaction data from which the object determined as the noise is removed. 17. The method of claim 16, Wherein the step of determining that the number of links is equal to or less than a predetermined number as noise and removing the objects from the cluster,  And repeatedly removing objects of different types for each type. 11. The method of claim 10, Setting the plurality of objects as low-level nodes for each type and setting a cluster generated from the plurality of objects as high-level nodes to generate a structured tree &Lt; / RTI &gt; 18. A computer-readable recording medium on which a program for executing the method of any one of claims 10 to 18 is recorded.

Images