KR101221757B1 - Method for generating of 3D rule model using decision tree in OLAP and system thereof - Google Patents

Abstract

The present invention includes selecting a target DB for data mining, applying a decision tree technique to the target DB, obtaining a decision rule result of classifying the DB according to a plurality of rule conditions, and the decision rule. Generating a three-dimensional rule model corresponding to a result, wherein the decision rule result is represented by a tree shape of a binary tree type according to the classification, and each end node of the tree branch represents a specific rule condition. A three-dimensional rule model generation method using a decision tree technique in an OLAP system in which the three-dimensional rule model is classified into similar classes among the similar rule conditions Provide a system.
According to the three-dimensional rule model generation method using the decision tree method in the OLAP system according to the present invention, the data mining analysis result using the decision tree method is effectively visualized and represented as a three-dimensional rule model to analyze the general user. This has the advantage of enabling intuitive grasping of results and improving the convenience of data access.

Description

Method and generating of 3D rule model using decision tree in OLAP and system approx.

The present invention relates to a method and system for generating a three-dimensional rule model using a decision tree technique in an OLAP system. More specifically, the results of data mining analysis using a decision tree technique in an online analytical processing (OLAP) system are described. The present invention relates to a 3D rule model generation method and system using a decision tree technique in an OLAP system that effectively visualizes a dimensional rule model.

Online Analytical Processing (OLAP) generally refers to computer software that efficiently analyzes and processes large amounts of multi-dimensional data. Such OLAP applies to all areas of accumulating data (ex. General business, manufacturing processes, biotechnology). do.

On the other hand, the decision tree technique is a representative classification technique of data mining, and it divides the relationship of numerous data stored in the database into a pattern of binary tree and expresses it as a binary tree structure. This decision tree technique has the advantage that the classification structure is clearly distinguished by rules, and the time required to express the structure is short, so it is mainly used for data mining analysis work. However, since the length of the tree structure is determined by the complexity of the sample and the rule of data, the more correlation rules, the more difficult it is to grasp the rule correlation between the data.

Conventional visualization methods of data mining analysis results include complex texts and diagrams, and have been generally interpreted by a skilled analyst. The expression methods include two-dimensional autonomic curve graphs, and visualization of all the rule conditions required for data mining by cone-shaped three-dimensional modeling. However, this conventional case has a problem that many difficulties in interpretation without general description and knowledge about data mining.

According to the present invention, a decision tree in an OLAP system enables intuitive grasp of analysis results and improves data access convenience by effectively visualizing and expressing data mining analysis results using a decision tree method using a 3D rule model. Provided is a method and system for generating a 3D rule model using the method.

The present invention includes selecting a target DB for data mining, applying a decision tree technique to the target DB, obtaining a decision rule result of classifying the DB according to a plurality of rule conditions, and the decision rule. Generating a three-dimensional rule model corresponding to a result, wherein the decision rule result is represented by a tree shape of a binary tree type according to the classification, and each end node of the tree branch represents a specific rule condition. The 3D rule model is a method of generating a 3D rule model using a decision tree technique in an OLAP system in which the rule conditions are similar in color. to provide.

Here, the 3D rule model is displayed by connecting the same classes in the form of a 3D circular sphere, the individual size of the 3D circular sphere may be determined in proportion to the number of records that meet the rule conditions of the classes.

The 3D rule model generation method using the decision tree technique in the OLAP system includes extracting a color to be applied to each circular sphere of the 3D rule model from a YUV-based color coordinate system, and the 3D The method may further include applying the extracted color on each circular sphere of the rule model.

The extracting of the color may include setting a range of a circular trajectory of the YUV-based color coordinate system, initializing a minimum angle and a maximum angle for setting a color section, and setting the minimum and maximum angles. Setting a color section of each of the left and right child nodes of each layer constituting the decision rule result using the angle, and calculating a representative angle corresponding to the position of each child node on the corresponding color section, respectively. And calculating, for each of the child nodes, coordinate information on the color coordinate system corresponding to the representative angle by using the range of the circular orbit, and the child nodes corresponding to the calculated coordinates, respectively. Extracting the color of each.

In this case, the color of the parent node for the left child node and the right child node may correspond to the middle value of the color of the left child node and the color of the right child node.

In addition, setting the color sections of the left child node and the right child node, respectively, may use the following equation.

은 상기 최소각도,

는 상기 최대각도이고,

와

는 상기 왼쪽 자식 노드의 색상 구간의 설정을 위한 최소각도 및 최대각도이고,

와

는 상기 오른쪽 자식 노드의 색상 구간의 설정을 위한 최소각도 및 최대각도이다.here,

Is the minimum angle,

Is the maximum angle,

Wow

Is the minimum angle and the maximum angle for setting the color section of the left child node,

Wow

Is the minimum angle and the maximum angle for setting the color section of the right child node.

In addition, the representative angle may use the following equation.

는 상기 왼쪽 자식 노드의 위치에 대응되는 대표각도이고,

는 상기 오른쪽 자식 노드의 위치에 대응되는 대표각도이다.here,

Is a representative angle corresponding to the position of the left child node,

Is a representative angle corresponding to the position of the right child node.

In addition, the coordinate information may use the following equation.

은 상기 원형 궤도의 범위이고,

와

는 상기 왼쪽 자식 노드의 경우

와

값, 상기 오른쪽 자식 노드의 경우

와

값을 각각 나타낸다.here,

Is the range of the circular orbit,

Wow

Is the above left child node

Wow

The value, for the right child node above

Wow

Each value is represented.

값은 노드의 길이에 비례하여 증가하도록 설정될 수 있다.In addition, in the circular color coordinate system, as the length of the node increases so that the brightness information changes according to the length of the node, the type of color is further subdivided.

The value may be set to increase in proportion to the length of the node.

The present invention also provides a computer-readable recording medium having recorded thereon a program for executing a three-dimensional rule model generation method using a decision tree technique in the OLAP system.

In addition, the present invention, a data selection unit for selecting a target DB for data mining, a result acquisition unit for obtaining a decision rule result of classifying the DB by a plurality of rule conditions by applying a decision tree technique to the target DB, And a model generation unit for generating a 3D rule model corresponding to the decision rule result, wherein the decision rule result is expressed in the shape of a tree branch of a binary tree type according to the classification, and each end of the tree branch. Nodes are divided into individual classes representing a set of target DBs that satisfy a specific rule condition. The three-dimensional rule model uses a decision tree technique in an OLAP system in which the rule conditions are expressed in similar colors. Provide a dimensional rule model generation system.

In addition, the three-dimensional rule model generation system using a decision tree technique in the OLAP system, a color extraction unit for extracting a color to be applied to each circular sphere of the three-dimensional rule model from the YUV-based color coordinate system, and the The apparatus may further include a color applying unit that applies the extracted color on each circular sphere of the 3D rule model.

According to the three-dimensional rule model generation method using the decision tree method in the OLAP system according to the present invention, the data mining analysis result using the decision tree method is effectively visualized and represented as a three-dimensional rule model to analyze the general user. This has the advantage of enabling intuitive grasping of results and improving the convenience of data access.

1 is a flowchart of a 3D rule model generation method using a decision tree technique in an OLAP system according to an embodiment of the present invention.
2 is a system configuration diagram for FIG. 1.
3 is an exemplary diagram of a target DB for step S110 of FIG.
4 is an exemplary diagram of a result of applying the decision tree technique to FIG. 3 as step S120 of FIG. 1.
5 is an exemplary diagram of storing the result of FIG. 4 as an XML-based script.
FIG. 6 is an exemplary diagram of a 3D rule model corresponding to the result of FIG. 4, in operation S130 of FIG. 1.
7 is an exemplary diagram of a symbiotic and competitive relationship representation of a three-dimensional rule model according to an embodiment of the present invention.
8 is a diagram illustrating a YUV-based color palette for step S140 of FIG. 1.
9 is a detailed flowchart for step S140 of FIG.
10 is an exemplary view illustrating a color extraction result according to step S140 of FIG. 1.
FIG. 11 is a conceptual diagram illustrating a color extraction process of nodes ①, ②, and ③ of FIG. 10.
12 is an exemplary view showing a similarity relationship between rule conditions for color spheres represented according to an embodiment of the present invention.

DETAILED DESCRIPTION Embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention.

The present invention relates to a three-dimensional rule model generation method using a decision tree technique in an OLAP system, and in particular, by applying statistical and mathematical analysis methods from a large amount of data, various and valuable values such as relationships, propensities, patterns, etc. Present information effectively to users. The present invention can effectively visualize and express the decision tree-based data mining results supporting decision making.

1 is a flowchart of a 3D rule model generation method using a decision tree technique in an OLAP system according to an embodiment of the present invention. 2 is a system configuration diagram for FIG. 1. The system 100 includes a data selecting unit 110, a result obtaining unit 120, a model generating unit 130, a color extracting unit 140, and a color applying unit 150.

Hereinafter, a method of generating a 3D rule model using the decision tree method will be described in detail with reference to FIGS. 1 and 2.

First, a target DB for data mining is selected through the data selector 110 (S110). The target DB may correspond to an analysis target DB in various fields.

3 is an exemplary diagram of a target DB for step S110 of FIG.

3 is an example of a target DB having main data including personal ID information, gender, occupation, place of residence, marital status, salary, hobby, religion, total number of accounts, credit balance, received balance, and the like. Of course, the present invention is not necessarily limited to this example.

Here, among the plurality of analysis items (personal ID information, gender, occupation, place of residence, marital status, salary, hobby, religion, total number of accounts, credit balance, received balance), an item to be analyzed is determined as a class. For example, if you want to mine data based on your residence, set Guro-gu as Class A, Mapo-gu as Class B, Gangseo-gu as Class C, and Gangnam-gu as Class D.

After the step S110, the decision tree technique is applied to the target DB to obtain a decision rule result of classifying the DB according to a plurality of rule conditions (S120). This step S120 is performed through the result acquisition unit 120.

4 is an exemplary diagram of a result of applying the decision tree technique to FIG. 3 as step S120 of FIG. 1. This is the result of a decision tree analysis made up of a series of rule patterns. At this time, the rule conditions used are salary, receiving balance, credit balance, occupation, place of residence, number of accounts, religious presence, hobby and the like.

From this, it can be seen that the decision rule result obtained by analyzing a large target DB based on a decision tree-based data mining technique is represented as a tree branch of a binary tree type according to the classification according to the rule condition. Can be.

In addition, it can be seen that each end node of the tree branch is divided into individual classes representing a set of target DBs that satisfy specific rule conditions. In other words, each end node of the decision tree can be viewed as a set of rule conditions, which are divided into classes.

5 is an exemplary diagram of storing the result of FIG. 4 as an XML-based script. The set of class condition rules is stored as an XML-based script with efficient data management and structural generality. It consists of unique number, attribute name, instance value, and weight value.

After the step S120, the model generator 130 generates a 3D rule model corresponding to the decision rule result (S130).

FIG. 6 is an exemplary diagram of a 3D rule model corresponding to the result of FIG. 4, in operation S130 of FIG. 1. Here, the three-dimensional rule model is displayed connected in the form of a three-dimensional circular sphere clustered among the same classes. In this case, each class is represented by a set of circular spheres in which the same classes are connected in a row. In addition, the three-dimensional rule model is displayed in a similar color between the classes similar to the rule condition.

The individual size of the three-dimensional circular spheres is determined in proportion to the number of records meeting the rule conditions of the corresponding classes. In FIG. 6, the larger the diameter size of the circular sphere is, the larger the number of records are. In this case, an individual size (ex, diameter) of the circular spheres may be determined by adding a specific gravity value proportional to the number of records.

7 is an exemplary diagram of a symbiotic and competitive relationship representation of a three-dimensional rule model according to an embodiment of the present invention. That is, the three-dimensional rule model generated from decision tree-based rule data can be expressed in a three-dimensional visualization symbiosis form or a three-dimensional visualization competition form, which is an ecosystem-based life form, for each class to have a specific direction. have.

This FIG. 7 adds a lifelike dynamic representation to the analyzed data, allowing the data analyst, i.e., users, to feel the lively interaction between the multidimensional / large data. These creatures are similar to migratory birds that are grouped in a specific direction, and are expressed in complex shapes such as tadpoles swimming in the water with large heads and thin tails.

After the step S130, the color to be applied to each circular sphere of the three-dimensional rule model from the YUV-based color coordinate system is extracted through the color extraction unit 140 (S140). This step S140 is to measure the degree of similarity of the rule data of the circular sphere visualized by the three-dimensional rule model and to express it later in color.

8 is a diagram illustrating a YUV-based color palette for step S140 of FIG. 1. The upper left is an example of a YUV color palette, and the upper right is a coordinate system representing brightness information (Y) and circular orbit (r) in the left palette. In the present invention, the color is extracted using a color coordinate system based on YUV, which is one of the methods of expressing color difference information. The YUV-based color coordinate system is divided into color sections along a circular orbit.

The YUV method has a lower color quality deterioration than RGB having information about a primary color signal. Since the human eye is more sensitive to brightness information than color information, the present embodiment selects a YUV method that can express the brightness information Y and the color information U and V separately. In the present invention, the brightness information (Y) value is changed (more subdivided) according to the node length of the decision tree (see the figure below in FIG. 8), and after determining the size (r) of each circular trajectory of the YUV-based color coordinate system, It was set as the color section. This may increase the combination of colors.

According to step S140, the color sections of the child nodes are separated for each layer according to the characteristics of the decision tree stratified to each end node, and the classes similar to the rule conditions are extracted to similar colors along the circular trajectory. . This principle is also confirmed in the lower picture of Fig. 8, that is, the decision tree picture to which the color expression is applied, and it can be seen that each class of the decision tree is extracted color in the circular orbit.

8, it can be seen that the color of the 'parent node' for the left child node and the right child node in the decision tree corresponds to the middle value of the color of the left child node and the right child node.

After the step S140, the extracted color is applied to each circular sphere of the three-dimensional rule model (S150). This is performed through the color applying unit 150. According to this step S150, the color of each circular sphere expresses a variety of colors, similar classes are represented by similar colors.

Hereinafter, the step of extracting the color will be described in more detail. 9 is a detailed flowchart for step S140 of FIG. First, the range of the circular trajectory of the YUV-based color coordinate system is set (S910). In the present embodiment, in the case of the brightness information Y of the YUV-based color coordinate system, the left child node of the decision node is increased and the right child node is decreased. Then, the range of the circular orbit r increases from the origin (r = 0) to a range of up to 0.5, so that the value of r gradually increases by "0.5 / the total node length" each time the length of the nodes increases by one. Set.

은 90°, 최대각도

는 450°로 초기화하였다.After the step S910, in order to set the color section, the minimum and maximum angles are initialized (S920). Minimum angle in this embodiment

Is 90 °, maximum angle

Was initialized to 450 °.

Next, using the minimum angle and the maximum angle, the color sections of the left child node and the right child node of each layer constituting the decision rule result are respectively set (S930). This S930 step uses Equation 1.

은 상기 최소각도,

는 상기 최대각도이다. here,

Is the minimum angle,

Is the maximum angle.

와

는 상기 왼쪽 자식 노드의 색상 구간의 설정을 위한 최소각도 및 최대각도이다. 또한,

와

는 상기 오른쪽 자식 노드의 색상 구간의 설정을 위한 최소각도 및 최대각도이다.And,

Wow

Is the minimum angle and the maximum angle for setting the color section of the left child node. Also,

Wow

Is the minimum angle and the maximum angle for setting the color section of the right child node.

Thereafter, the representative angles corresponding to the positions of the respective child nodes in the corresponding color section are respectively calculated (S940). The representative angle is calculated using Equation 2.

는 상기 왼쪽 자식 노드의 위치에 대응되는 대표각도이고,

는 상기 오른쪽 자식 노드의 위치에 대응되는 대표각도이다.here,

Is a representative angle corresponding to the position of the left child node,

Is a representative angle corresponding to the position of the right child node.

Thereafter, coordinate information on the color coordinate system corresponding to the representative angle is calculated for each of the child nodes using the range of the circular trajectory (S950). The calculation of the coordinate information uses equation (3).

은 상기 원형 궤도의 범위이고,

와

는 상기 왼쪽 자식 노드의 경우

와

값, 상기 오른쪽 자식 노드의 경우

와

값을 각각 나타낸다. 이때, 상기 원형 궤도의 범위는 상기 원형 좌표계의 반지름 값에 대응된다.

은 최대 0.5의 값을 갖는다. 또한, 상기 노드의 길이 하나씩 증가할 때마다

값이 0에서 최대 0.5까지 증가한다. 만약, 데이터 마이닝된 결과, 총 노드의 길이가 7개인 경우에는, 노드의 길이가 하나씩 증가할 때마다

값은 0.07(=0.5/7) 만큼 증가한 값을 수학식 3에 적용한다. 즉, 노드의 길이가 0인 최초 부모 노드의 경우

=0(따라서 (0,0)의 좌표에 위치), 다음의 자식 노드에 대해서는 노드의 길이가 1이므로

=0.07, 그 다음의 자식 노드에 대해서는 노드의 길이가 2이므로

=0.14 값이 적용된다.here,

Is the range of the circular orbit,

Wow

Is the above left child node

Wow

The value, for the right child node above

Wow

Each value is represented. At this time, the range of the circular orbit corresponds to the radius value of the circular coordinate system.

Has a maximum value of 0.5. In addition, each time the length of the node increases by one

The value increases from 0 to a maximum of 0.5. If, as a result of data mining, the total node length is seven, each time the node length increases by one

The value increases by 0.07 (= 0.5 / 7) to Equation 3. That is, for the first parent node with zero length node

= 0 (and thus located at the coordinates of (0,0)), so that the next child node has a node length of 1

= 0.07, for the next child node the length of the node is 2

The value = 0.14 applies.

Next, colors of the respective child nodes corresponding to the calculated coordinates are extracted, respectively (S960). Accordingly, the individual colors of all nodes in the decision tree are determined through respective coordinate values located in the circular trajectory of the YUV-based color coordinate system.

10 is an exemplary view illustrating a color extraction result according to step S140 of FIG. 1. A series of rule conditions that descend from the start node to the end node are defined as one class each. Each node is extracted from the corresponding section of the YUV-based color coordinate system shown in FIG. 10 through the process of FIG.

According to each rule, the first two child nodes ① and ② are present, and the child nodes ① and ② are each different in the corresponding interval of the YUV-based color coordinate system shown in FIG. 10 through the process of FIG. 9. Extracted and assigned to individual colors.

Node ① is classified into child nodes ③ and ④ according to each rule. At this time, node ① corresponds to parent node for nodes ③ and ④. Child nodes ③ and ④ are also extracted from the corresponding sections of the YUV-based color coordinate system shown in FIG. 10 through the process of FIG.

That is, as shown in the example of the decision tree to which the color of FIG. 10 is applied, the terminal node is divided into class A, class B, class C, etc. according to a series of rules. It includes.

값은 노드의 길이에 비례하여 증가한다.Referring to the lower left figure of FIG. 10, the circular color coordinate system shows that the type of color is further subdivided as the length of the node increases so that the brightness information Y changes according to the length of the node. Can be. At this time, as described above

The value increases in proportion to the length of the node.

The circular coordinate system of FIG. 10 is a diagram representing the circular orbit of the YUV coordinate system when the total depth of the node is 7. FIG. That is, the orbits are divided into seven stages from the center of the coordinate system toward the outside (marks 2 and 4 of the circular coordinate system do not mean nodes (②, ④) at the top of FIG. 10, but the length of the node ( Depth) shows the corresponding coordinate system coverage area by star). This is to express a more simplified example for convenience of description as the total length of the node is smaller than the right figure of FIG. 8. For reference, the decision tree at the top of FIG. 10 simply shows up to the length of 4 nodes.

In the lower right figure of FIG. 10, the brightness information (Y) values of the left child node and the right child node of the decision tree are added or subtracted at a predetermined ratio to extract the actual color in the left circular trajectory. That is, as described above, the brightness value of the left child node gradually increases, and the brightness value of the right child node gradually decreases based on the parent node. This can be seen by comparing the allocated color information of each node (①, ②, ③, etc.) of the upper decision tree of FIG. 10 and the corresponding node (①, ②, ③, etc.) of the lower right figure. As the length of the node increases, the left child node increases in brightness information Y slightly, and the right child node decreases in brightness information Y slightly.

FIG. 11 is a conceptual diagram illustrating a color extraction process of nodes ①, ②, and ③ of FIG. 10. This will be described in connection with Equation and FIG. 10 above.

= 90°, 최대각도

= 450°를 이용하여, 상기 왼쪽 자식 노드 ①에 대한

값과

값을 수학식 1을 통해 우선 산출한다(S930).

는 90°,

는 270°로 산출되고, 이에 따라 상기 왼쪽 자식 노드 ①에 대한 색상 구간은 상기 색상좌표계 상에서 90°~ 270°범위로 설정된다.First, the color extraction process of the first left child node ① will be described in FIG. 10. In this case, the beginning of FIG. 11 corresponds to the Yearly Income part of FIG. 10, and the length of the node is 0 and has a value of r = 0. First, the initial minimum angle

= 90 °, maximum angle

= 450 °, for the left child node ①

Value and

A value is first calculated through Equation 1 (S930).

Is 90 °,

Is calculated as 270 °, and thus, the color section for the left child node ① is set in the range of 90 ° to 270 ° on the color coordinate system.

값과

값을 수학식 2에 대입한 결과,

값은 180°로 산출된다(S940). 이는 곧 상기 왼쪽 자식 노드 ①의 위치에 대응되는 대표각도에 해당된다.Next, obtained above

Value and

As a result of substituting the value into Equation 2,

The value is calculated at 180 ° (S940). This corresponds to a representative angle corresponding to the position of the left child node ①.

와

값을 수학식 3에 대입하되, 상기 왼쪽 자식 노드 ①은 노드의 길이(Depth)가 1인 경우므로

=0.07을 적용하면, 상기 왼쪽 자식 노드 ①의 좌표 정보(U,V)가 (-0.07,0)으로 산출된다(S950). Then, the above

Wow

A value is substituted into Equation 3, but since the left child node ① has a node length of 1,

When = 0.07 is applied, the coordinate information (U, V) of the left child node ① is calculated as (-0.07,0) (S950).

Accordingly, the color corresponding to the position of (U, V) = (-0.07,0) in the color coordinate system of FIG. 10 (the color of position ①) is extracted (S960), and this is the circular sphere color for the node ①. Apply.

= 90°, 최대각도

= 450°를 이용하여, 상기 오른쪽 자식 노드 ②에 대한

값과

값을 수학식 1을 통해 산출한다(S930).

는 270°,

는 450°로 산출되고, 이에 따라 상기 오른쪽 자식 노드 ②에 대한 색상 구간은 상기 색상좌표계 상에서 270°~ 450°범위로 설정된다.Next, the process of extracting the color of the first right child node ② will be described in FIG. 10. Initialized minimum angle

= 90 °, maximum angle

= 450 °, for the right child node ②

Value and

The value is calculated through Equation 1 (S930).

270 °,

Is calculated as 450 °, and thus the color section for the right child node ② is set in the range of 270 ° to 450 ° on the color coordinate system.

값과

값을 수학식 2에 대입한 결과,

값은 360°로 산출된다(S940). 이는 곧 상기 오른쪽 자식 노드 ②의 위치에 대응되는 대표각도에 해당된다.Next, obtained above

Value and

As a result of substituting the value into Equation 2,

The value is calculated at 360 ° (S940). This corresponds to a representative angle corresponding to the position of the right child node ②.

값과

값을 수학식 3에 대입하되, 상기 오른쪽 자식 노드 ② 또한 노드의 길이(Depth)가 1인 경우므로

=0.07을 적용하면, 상기 오른쪽 자식 노드 ②의 좌표 정보(U,V)가 (0.07,0)으로 산출된다(S950).Then, the above

Value and

A value is substituted into Equation 3, but the right child node ② also has a node length of 1,

When = 0.07 is applied, coordinate information (U, V) of the right child node ② is calculated as (0.07,0) (S950).

Accordingly, the color corresponding to the position of (U, V) = (0.07,0) in the color coordinate system of FIG. 10 (the color of position ②) is extracted (S960), and applied as a circular sphere color for the node ②. do.

과 최대각도

는 ①번 노드에 대하여 산출되었던 최소각도와 최대각도 값 즉, 90°와 180°를 각각 사용한다. Of course, the same method as the above-described method is used for the left child node ③ with respect to ①. At this time, the minimum angle applied to the equation (1)

And max angle

Uses the minimum and maximum angle values calculated for node ①, that is, 90 ° and 180 °, respectively.

This result of FIG. 11 leads to the same result as the bottom right result of FIG. 10.

In this way, the colors of all three-dimensional circular spheres corresponding to each end node of the decision tree are extracted within the color interval of the color coordinate system, and the circular trajectory range is determined according to the complexity of the decision tree structure. Accordingly, the angle between the minimum angle and the maximum angle, that is, the color interval determined for the node becomes narrower toward the bottom of the decision tree. Of course, the series of steps described above recursively calculates the separation process of the color section or the application of color to the circular sphere until the last end node is found.

By repeating the process of the above principle, it is possible to calculate the results of Figure 10, that is, the coordinates for each node, respectively, and to extract the color from each of the three-dimensional rule model, that is, to apply the color to each circular sphere It becomes possible.

For reference, referring to the decision tree illustrated in FIG. 10, even if the classification of the application rules is different, the types of classes may be the same, and the applied colors may be similar even if the types of classes are different.

12 is an exemplary view showing a similarity relationship between rule conditions for color spheres represented according to an embodiment of the present invention. 12 illustrates a three-dimensional rule model for classes A, B, and C, and individual colors of the three-dimensional rule model represent colors of end nodes corresponding to classes of decision trees.

Referring to this, the color of the circular sphere in a specific class represents the degree of similarity between each rule condition, which may be expressed similarly to a part of rule conditions corresponding to another class.

In Fig. 12, the circular spheres of (1) and (2) are both class A and the same kind of class but different colors are assigned. This means that the similarity between rules falls. In contrast, the circular spheres of (3) and (4) differ in class C and A, but are expressed in almost similar colors due to the high similarity between the rule conditions.

According to the present invention as described above, by expressing the ecosystem-based three-dimensional model generated by data mining in various colors according to the similarity of rule conditions, it is possible to more clearly reveal the correlation between classes. This is notable because it is not a simple representation of color, but a feature that the color changes without overlap in proportion to the similarity of rule conditions between nodes.

If the similarity of rule condition falls even though it is the end node of the same class, the color may be different.If the similarity of rule condition is great even if it is the end node belonging to different class, the color is similar to the user. To help people understand

That is, the present invention supports efficient and appropriate decision making by visualizing a large amount of data and information in a three-dimensional form having excellent visual effects without restricting two-dimensional data and information in the expression of data mining results. In addition, it improves the accessibility of the analysis results, thus extending the scope of analysis and utilization even for general users who do not have the skill of interpretation.

The three-dimensional rule model generation method using the decision tree technique as described above can be implemented as computer-readable codes on a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data is stored as a medium that can be read by a computer system.

Examples of computer-readable recording media include ROM, RAM, CO-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which are also implemented in the form of carrier waves (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

Although the present invention has been described with reference to the embodiments shown in the drawings, these are merely exemplary and those skilled in the art will understand that various modifications and equivalent other embodiments are possible. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

100: 3D Rule Model Generation System Using Decision Tree Technique
110: data selection unit 120: result acquisition unit
130: model generation unit 140: color extraction unit
150: color application

Claims (18)

Selecting a target DB for data mining;
Obtaining a decision rule result of classifying the DB by a plurality of rule conditions by applying a decision tree technique to the target DB; And
Generating and visualizing a three-dimensional rule model corresponding to the decision rule result;
The decision rule result is
According to the classification, it is expressed in the shape of a tree branch of binary tree type, and each end node of the tree branch is divided into individual classes representing a set of target DBs satisfying specific rule conditions.
The three-dimensional rule model,
The rule conditions are displayed in similar colors between the similar classes, the same classes are displayed connected to the three-dimensional shape,
The three-dimensional rule model generation method using a decision tree technique in the OLAP system is determined in proportion to the number of records that meet the rule conditions of the classes. The method according to claim 1,
The three-dimensional shape is a three-dimensional rule model generation method using a decision tree method in the OLAP system of the three-dimensional circular sphere shape. The method according to claim 2,
Extracting a color to be applied on each circular sphere of the 3D rule model from a YUV-based color coordinate system; And
3. The method of claim 3, further comprising applying the extracted color to each circular sphere of the 3D rule model. The method according to claim 3,
Extracting the color,
Setting a range of a circular trajectory of the YUV-based color coordinate system;
Initializing a minimum angle and a maximum angle for setting a color section;
Setting color sections of the left child node and the right child node of each layer constituting the decision rule result using the minimum angle and the maximum angle;
Calculating representative angles corresponding to positions of respective child nodes on a corresponding color section, respectively;
Calculating coordinate information on the color coordinate system corresponding to the representative angle for each of the child nodes using the range of the circular trajectory; And
3. A method of generating a 3D rule model using a decision tree technique in an OLAP system comprising extracting colors of the respective child nodes corresponding to the calculated coordinates. The method of claim 4,
The color of the parent node for the left child node and the right child node is,
3D rule model generation method using a decision tree technique in the OLAP system corresponding to the middle value of the color of the left child node and the color of the right child node. The method of claim 4,
The step of setting the color sections of the left child node and the right child node, respectively, is a method of generating a 3D rule model using a decision tree technique in an OLAP system using the following equation:

here,

Is the minimum angle,

Is the maximum angle,

Wow

Is the minimum angle and the maximum angle for setting the color section of the left child node,

Wow

Is the minimum angle and the maximum angle for setting the color section of the right child node. The method of claim 6,
The representative angle is a three-dimensional rule model generation method using a decision tree method in an OLAP system using the following equation:

here,

Is a representative angle corresponding to the position of the left child node,

Is a representative angle corresponding to the position of the right child node. The method of claim 7,
The coordinate information is a 3D rule model generation method using a decision tree method in an OLAP system using the following equation:

here,

Is the range of the circular orbit,

Wow

Is the above left child node

Wow

The value, for the right child node above

Wow

Each value is represented. The method according to claim 8,
In the circular color coordinate system,
As the length of the node increases so that the brightness information changes according to the length of the node, the type of color is further subdivided.

A 3D rule model generation method using a decision tree technique in an OLAP system configured to increase in proportion to the length of a node. A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 1 to 9. A data selection unit for selecting a target DB for data mining;
A result obtaining unit obtaining a decision rule result of classifying the DB by a plurality of rule conditions by applying a decision tree technique to the target DB; And
And a model generator for generating and visualizing a 3D rule model corresponding to the decision rule result.
The decision rule result is
According to the classification, it is expressed in the shape of a tree branch of binary tree type, and each end node of the tree branch is divided into individual classes representing a set of target DBs satisfying specific rule conditions.
The three-dimensional rule model,
The rule conditions are displayed in similar colors between the similar classes, the same classes are displayed connected to the three-dimensional shape,
And a three-dimensional rule model generation system using a decision tree technique in an OLAP system in which the individual sizes of the three-dimensional shapes are determined in proportion to the number of records meeting the rule conditions of the corresponding classes. The method of claim 11,
3D rule model generation system using the decision tree method in the OLAP system of the three-dimensional shape is a three-dimensional circular sphere. The method of claim 12,
A color extraction unit for extracting a color to be applied on each circular sphere of the 3D rule model from a YUV-based color coordinate system; And
3D rule model generation system using a decision tree technique in the OLAP system further comprises a color applying unit for applying the extracted color on each circular sphere of the three-dimensional rule model. The method according to claim 13,
The color extraction unit,
After setting the range of the circular orbit of the YUV-based color coordinate system, for setting the color section, initialize the minimum angle and the maximum angle,
Using the minimum and maximum angles, the color sections of the left child node and the right child node of each layer constituting the decision rule result are respectively set, and then a representative corresponding to the position of each child node on the corresponding color section. Calculate each angle,
Using the range of the circular trajectory, the coordinate information on the color coordinate system corresponding to the representative angle is calculated for each of the child nodes, and then the colors of the respective child nodes corresponding to the calculated coordinates are respectively extracted. 3D Rule Model Generation System Using Decision Trees in OLAP System. The method according to claim 14,
The color extraction unit,
A 3D rule model generation system using a decision tree method in an OLAP system using the following equation when setting the color sections of the left child node and the right child node:

here,

Is the minimum angle,

Is the maximum angle,

Wow

Is the minimum angle and the maximum angle for setting the color section of the left child node,

Wow

Is the minimum angle and the maximum angle for setting the color section of the right child node. The method according to claim 15,
The color extraction unit,
3D rule model generation system using the decision tree method in the OLAP system using the following equation when calculating the representative angle:

here,

Is a representative angle corresponding to the position of the left child node,

Is a representative angle corresponding to the position of the right child node. 18. The method of claim 16,
The color extraction unit,
3D rule model generation system using the decision tree method in the OLAP system using the following equation when calculating the coordinate information:

here,

Is the range of the circular orbit,

Wow

Is the above left child node

Wow

The value, for the right child node above

Wow

Each value is represented. 18. The method of claim 17,
In the circular color coordinate system,
As the length of the node increases so that the brightness information changes according to the length of the node, the type of color is further subdivided.

3D rule model generation system using decision tree method in OLAP system, whose value is increased in proportion to node length.

Images