JP2004246622A - Outlier detection supporting program, outlier detection supporting method, and outlier detection supporting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an outlier detection supporting program, an outlier detection supporting method, and an outlier detection supporting device capable of increasing the efficiency of the detection or removal of an outlier. <P>SOLUTION: This outlier detection supporting device is provided with a control part 103 for setting a data identifier for identifying each data in a data group being the target of outlier detection, and for visualizing the data group by a plot, and for making a display part 102 display the data identifier corresponding to respective data in the visualized data group, and an input part 101 for making an operator designate the data identifier corresponding to the data of an outlier in the visualized data group. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an outlier detection support program, an outlier detection support method, and an outlier detection support device for supporting detection of an outlier from a data set, and particularly to an efficiency related to outlier detection and deletion. The present invention relates to an outlier detection support program, an outlier detection support method, and an outlier detection support device that can enhance the outlier.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, in the fields of data mining, data analysis, and pattern recognition, it is very important to detect data having abnormal values that adversely affect analysis, that is, outliers. For example, a technique for detecting outliers has been developed mainly in the field of statistics, because the presence of only one outlier may cause a decrease in the accuracy of identification or prediction.
[0003]
FIG. 12 is a diagram showing a conventional data table 10. The data table 10 stores data of attribute 1 (for example, temperature) and attribute 2 (for example, length).
[0004]
In the data table 10, the record number is a number associated with each record. Attribute 1 is, for example, temperature data. Attribute 2 is, for example, length data. That is, in the data table 10, the correspondence between the temperature and the length is expressed.
[0005]
FIG. 13 is a scatter diagram 20 in which data (attribute 1: temperature, attribute 2: length) of each record in the data table 10 shown in FIG. 12 is plotted. In the scatter diagram 20, the horizontal axis corresponds to attribute 1 (temperature), and the vertical axis corresponds to attribute 2 (length). In the scatter diagram 20, for example, a plot in a circle 21 is an outlier.
[0006]
Accordingly, the operator reads the data (attribute 1 and attribute 2) of the plot from the scatter diagram 20 in order to detect and remove the plot (data) within the circle 21 as an outlier. Next, the operator searches the data table 10 shown in FIG. 12 for a record corresponding to the read data (attribute 1 and attribute 2), and then deletes the record (outlier).
[0007]
Thereby, outliers (plots in the circle 21) are removed from the scatter diagram 20. Thereafter, the above operation is repeated to detect and remove outliers.
[0008]
[Non-patent document 1]
Takashi Suenaga, Arata Sato, Akira Sakano, "Visualization of Feature Space Focusing on Cluster Structure-Cluster Discrimination Method", "Transactions of the Institute of Electronics, Information and Communication Engineers D-II, Vol. J85-D-II No. 5," Published by The Institute of Electronics, Information and Communication Engineers, May 2002, pp785-795
[0009]
[Problems to be solved by the invention]
By the way, as described above, in the related art, since the correspondence between each plot and each record of the data table 10 (see FIG. 12) cannot be understood only by looking at the scatter diagram 20 shown in FIG. The record corresponding to the outlier must be searched and deleted from the data table 10 using the data of the plot (outlier) read in the above as a key.
[0010]
As described above, conventionally, there has been a problem that the time and labor required for detecting and removing outliers are extremely long and the efficiency is low.
[0011]
The present invention has been made in view of the above, and an object of the present invention is to provide an outlier detection support program, an outlier detection support method, and an outlier detection support apparatus that can increase the efficiency of detecting and removing outliers. And
[0012]
[Means for Solving the Problems]
In order to achieve the above object, the invention according to claim 1 provides a computer with data identifier setting means for setting a data identifier for identifying each data in a data set to be detected as an outlier, and visualizing the data set. Visualizing means, data identifier displaying means for displaying the data identifier in association with each data in the visualized data set, and out-of-position for specifying a data identifier corresponding to outlier data in the visualized data set. This is an outlier detection support program for functioning as a value designation unit.
[0013]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and a data identifier corresponding to outlier data is specified in the visualized data set. Therefore, it is possible to increase the efficiency of detecting an outlier.
[0014]
According to a second aspect of the present invention, in the outlier detection support program according to the first aspect, the computer functions as enlargement display means for enlarging and displaying a portion where data overlaps in the visualized data set. It is characterized by making it.
[0015]
According to the present invention, a portion where data is overlapped in the visualized data set is enlarged and displayed. Therefore, even when data overlaps, the efficiency of detecting an outlier can be increased. .
[0016]
According to a third aspect of the present invention, in the outlier detection support program according to the first or second aspect, the visualization means includes a data as an outlier candidate satisfying a predetermined condition in the visualized data set. Are visualized by color coding.
[0017]
According to the present invention, in the visualized data set, data as an outlier candidate that satisfies a predetermined condition is visualized by color coding, so that the efficiency of outlier detection can be further increased.
[0018]
According to a fourth aspect of the present invention, in the outlier detection support program according to any one of the first to third aspects, the computer is configured to store data corresponding to a data identifier specified by the outlier specifying means. It is characterized by functioning as a removing means for removing as an outlier.
[0019]
According to the present invention, since the data corresponding to the specified data identifier is removed as an outlier, the efficiency of removing the outlier can be increased.
[0020]
Further, the invention according to claim 5 is a computer, comprising: a data identifier setting means for setting a data identifier for identifying each data in a data set to be detected as an outlier; a visualization means for visualizing the data set; Data identifier display means for displaying the data identifier in association with each data in the data set, and removing means for removing outlier data satisfying a predetermined condition in the visualized data set. This is an outlier detection support program.
[0021]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and outlier data satisfying a predetermined condition is removed from the visualized data set. Therefore, the efficiency of detecting and removing outliers can be increased.
[0022]
The invention according to claim 6 is a data identifier setting step of setting a data identifier for identifying each data in a data set to be detected as an outlier, a visualization step of visualizing the data set, A data identifier displaying step of displaying the data identifier in association with each data in the data set, and an outlier specifying step of specifying a data identifier corresponding to the outlier data in the visualized data set. It is characterized by including.
[0023]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and a data identifier corresponding to outlier data is specified in the visualized data set. Therefore, it is possible to increase the efficiency of detecting an outlier.
[0024]
The invention according to claim 7 is a data identifier setting step of setting a data identifier for identifying each data in a data set to be detected as an outlier, a visualization step of visualizing the data set, A data identifier display step of displaying the data identifier in association with each data in the data set, and a removing step of removing outlier data satisfying a predetermined condition in the visualized data set. Features.
[0025]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and outlier data satisfying a predetermined condition is removed from the visualized data set. Therefore, the efficiency of detecting and removing outliers can be increased.
[0026]
The invention according to claim 8 is a data identifier setting means for setting a data identifier for identifying each data in a data set which is an outlier detection target; a visualization means for visualizing the data set; Data identifier display means for displaying the data identifier in association with each data in the data set, and outlier specifying means for specifying a data identifier corresponding to outlier data in the visualized data set, It is characterized by having.
[0027]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and a data identifier corresponding to outlier data is specified in the visualized data set. Therefore, it is possible to increase the efficiency of detecting an outlier.
[0028]
The invention according to claim 9 is a data identifier setting means for setting a data identifier for identifying each data in a data set to be detected as an outlier, a visualization means for visualizing the data set, Data identifier display means for displaying the data identifier in association with each data in the data set, and removing means for removing outlier data satisfying a predetermined condition in the visualized data set. It is characterized by.
[0029]
According to the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and outlier data satisfying a predetermined condition is removed from the visualized data set. Therefore, the efficiency of detecting and removing outliers can be increased.
[0030]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of an outlier detection support program, an outlier detection support method, and an outlier detection support apparatus according to the present invention will be described in detail with reference to the drawings.
[0031]
FIG. 1 is a block diagram showing a configuration of an embodiment according to the present invention. In this figure, an outlier detection support apparatus 100 is an apparatus that supports detection and deletion of an outlier from a data set. In the outlier detection support device 100, the input unit 101 is a keyboard, a mouse, a data input device, or the like. The display unit 102 displays various data scatter diagrams (see FIGS. 4 to 8).
[0032]
The control unit 103 performs various controls (display control of a scatter diagram, control for detecting outliers, and the like). Details of the operation of the control unit 103 will be described later. The storage unit 104 stores various data under the control of the control unit 103.
[0033]
The original database 110 is a database that stores original data before performing outlier detection processing for detecting outliers in file units. The original data table 200 shown in FIG. 2A is obtained by opening one of a plurality of files stored in the original database 110.
[0034]
In the original data table 200, the record number is a number associated with each record. Here, in the original data table 200, when a record is deleted, the record numbers corresponding to the records below the deleted record are incremented by one, and are serialized from the highest record to the lowest record. . Attribute 1 to attribute n are data having n attributes. For example, attribute 1 is blood pressure data, attribute 2 is height data,..., Attribute n is blood sugar level data.
[0035]
Returning to FIG. 1, the processed database 120 is a database that stores, on a file-by-file basis, processed data on which the above-described outlier detection has been performed and outliers have been excluded from the original data.
[0036]
(Operation example 1)
Next, an operation example 1 of the embodiment will be described with reference to FIGS. FIG. 3 is a flowchart illustrating an operation example 1 of the embodiment. In step SA1 shown in the figure, the control unit 103 opens an outlier detection target file from the original database 110 and develops the initial original data table 200 shown in FIG.
[0037]
In step SA2, the control unit 103 creates the identification number table 210 in association with each record of the initial original data table 200 shown in FIG. The identification number table 210 is a table for storing identification numbers for identifying each data in the original data table 200.
[0038]
In the identification number table 210 corresponding to the initial original data table 200, the same identification number as the record number is assigned. Here, in the identification number table 210, when the record 200a of the original data table 200 is deleted as shown in FIG. 2B, the record 210a corresponding to the record 200a is deleted, and the identification number 2 is a missing number. , But the other identification numbers 1, 3 to m are unchanged. Thus, the identification number table 210 'shown in FIG. 2C is created.
[0039]
On the other hand, in the original data table 200, when the record 200a is deleted as shown in FIG. 2B, the record corresponding to the record 200a is deleted, and the other identification numbers 3 to m are one. The identification numbers are incremented one by one and set to identification numbers 2 to m-1. As a result, the original data table 200 'shown in FIG. 2C is created.
[0040]
In step SA3 and step SA4, a process of compressing high-dimensional (in this case, n-dimensional in this case, attributes 1 to n) into low-dimensional (in this case, two-dimensional) data is performed by a well-known cluster discriminating method. Is done. The cluster discrimination method is a data visualization method that performs clustering in a high-dimensional space, regards the data belonging to the obtained cluster as an independent category, and obtains a mapping to a low-dimensional space using discriminant analysis.
[0041]
That is, it is difficult to plot high-dimensional data on a scatter diagram and visualize the data. Therefore, by compressing high-dimensional data into low-dimensional (two-dimensional) data, two-dimensional data can be plotted on a scatter diagram to visualize the data.
[0042]
In step SA3, the control unit 103 specifies the number of clusters. In step SA4, the control unit 103 converts the high-dimensional (n-dimensional) original data stored in the original data table 200 shown in FIG. Compress to data. Thereby, the attributes 1 to n are compressed into a one-dimensional attribute and a two-dimensional attribute.
[0043]
If only two-dimensional (attribute 1 and attribute 2) original data has already been stored in the original data table 200, the processing of step SA3 and step SA4 is skipped.
[0044]
In step SA5, the control unit 103 scatters each record of the original data table 200 (however, data compressed from n dimensions to two dimensions (one-dimensional attribute and two-dimensional attribute) as plots shown in FIG. 4). 300 is displayed on the display unit 102. In the scatter diagram 300, the horizontal axis corresponds to one dimension (attribute), and the vertical axis corresponds to two dimensions (attribute).
[0045]
In step SA6, the control unit 103 displays each identification number of the identification number table 210 shown in FIG. 2A near each plot (each record) of the scatter diagram 300 shown in FIG. Let 300 be the scatter plot 310 shown in FIG. For example, “14” in the scatter diagram 310 shown in FIG. 5 corresponds to the identification number (= 14) in the identification number table 210 shown in FIG.
[0046]
In step SA7, the control unit 103 determines whether or not the plots overlap in the scatter diagram 310 shown in FIG. In this case, since a plurality of plots existing in the frame 311 shown in FIG. 6 overlap, the control unit 103 sets the determination result of step SA7 to “Yes”. If the result of the determination in step SA7 is “No”, the control section 103 makes a determination in step SA8.
[0047]
In step SA12, the control unit 103 enlarges the frame 311 (overlapping portion) shown in FIG. 6 and adds the enlarged scatter diagram 320 shown in FIG. 7 to the display unit 102 in parallel with the scatter diagram 310 (see FIG. 6). To be displayed. This makes the identification number and the plot easier to see.
[0048]
In step SA8, the control unit 103 determines whether or not the operator has instructed that an outlier exists in the scatter diagram 310 (see FIG. 6) and the enlarged scatter diagram 320 (see FIG. 7).
[0049]
In this case, the operator recognizes a plot (data) corresponding to “14” (circle frame 312) in the scatter diagram 310 shown in FIG. Give an indication that you are.
[0050]
Thereby, the control unit 103 sets the determination result of step SA8 to “Yes”. In step SA9, the operator uses the input unit 101 to specify the identification number “14” corresponding to the outlier.
[0051]
In step SA10, the control unit 103 deletes the record corresponding to the specified identification number “14” from the original data table 200 and the identification number table 210. Here, the identification number “14” is a missing number. Thereafter, steps SA3 to SA10 are repeated until the determination result of step SA8 becomes "No".
[0052]
Then, when the result of the determination in step SA8 is “No”, in step SA11, the control unit 103 stores the original data table 200 from which the record corresponding to the outlier has been deleted as processed data in the processed database 120.
[0053]
(Operation example 2)
Now, in one embodiment, outlier candidates may be displayed in different colors in a scatter diagram to assist the operator in detecting outliers. Hereinafter, this case will be described as Operation Example 2. FIG. 9 is a flowchart illustrating Operation Example 2 of the embodiment.
[0054]
In steps SB1 to SB4 shown in the same drawing, the same processing as the above-described steps SA1 to SA4 (see FIG. 3) is executed.
[0055]
In step SB5, the control unit 103 shows each record (however, data compressed from n dimensions to two dimensions (one-dimensional attribute and two-dimensional attribute)) of the original data table 200 as plots in FIG. A scatter diagram 300 is displayed on the display unit 102.
[0056]
In step SB6, the control unit 103 displays each identification number of the identification number table 210 shown in FIG. 2A near each plot (each record) of the scatter diagram 300 shown in FIG. Let 300 be the scatter plot 310 shown in FIG.
[0057]
In step SB7, the control unit 103 displays a plot (identification number) in which the number of clusters is 1 in red as an outlier candidate. In this case, it is assumed that the identification number “14” and the corresponding plot are displayed in red.
[0058]
In step SB8, the control unit 103 determines whether or not there is a portion where the plots overlap in the scatter diagram 310 illustrated in FIG. 5 and the identification number is difficult to see. In this case, the determination result is “Yes”. . If the result of the determination in step SB8 is “No”, the control section 103 makes a determination in step SB9.
[0059]
In step SB13, the control unit 103 enlarges the frame 311 (overlapping portion) shown in FIG. 6 and adds the enlarged scatter diagram 320 shown in FIG. 7 to the display unit 102 in parallel with the scatter diagram 310 (see FIG. 6). To be displayed.
[0060]
In step SB9, the control unit 103 determines whether or not the operator has instructed that a plot to be removed exists in the scatter diagram 310 (see FIG. 6) and the enlarged scatter diagram 320 (see FIG. 7).
[0061]
In this case, the operator recognizes a plot (data) corresponding to “14” (circled frame 312) displayed in red as an outlier in the scatter diagram 310 shown in FIG. Give an indication that the plot you want exists.
[0062]
Thereby, the control unit 103 sets the determination result of step SB9 to “Yes”. In step SB10, the operator uses the input unit 101 to specify the identification number “14” of the removal plot corresponding to the outlier.
[0063]
In step SB11, the control unit 103 deletes the record corresponding to the specified identification number “14” from the original data table 200 and the identification number table 210. Here, the identification number “14” is a missing number. Thereafter, Step SB3 to Step SB11 are repeated until the determination result of Step SB9 becomes “No”.
[0064]
Then, when the result of the determination in step SB9 is “No”, in step SB12, the control unit 103 stores the original data table 200 from which the record corresponding to the outlier has been deleted in the processed database 120 as processed data.
[0065]
(Operation example 3)
By the way, in the operation example 2 of the embodiment, a case has been described where outlier candidates are displayed in different colors on a scatter diagram, and outliers are detected and removed in accordance with an operator's instruction. Alternatively, the color-coded plot may be automatically removed as an outlier. Hereinafter, this case will be described as Operation Example 3. FIG. 10 is a flowchart illustrating an operation example 3 of the embodiment.
[0066]
In steps SC1 to SC4 shown in the same drawing, the same processing as the above-described steps SA1 to SA4 (see FIG. 3) is executed.
[0067]
In step SC5, the control unit 103 shows each record of the original data table 200 (however, data compressed from n dimensions to two dimensions (one-dimensional attribute and two-dimensional attribute)) as respective plots in FIG. A scatter diagram 300 is displayed on the display unit 102.
[0068]
In step SC6, the control unit 103 displays each identification number of the identification number table 210 shown in FIG. 2A near each plot (each record) of the scatter diagram 300 shown in FIG. Let 300 be the scatter plot 310 shown in FIG.
[0069]
In step SC7, the control unit 103 displays a plot (identification number) in which the number of clusters is 1 in red as a candidate for an outlier. In this case, it is assumed that the identification number “14” and the corresponding plot are displayed in red.
[0070]
In step SC8, the control unit 103 determines whether or not the plots overlap in the scatter diagram 310 illustrated in FIG. 5 and there is a part where the identification number is difficult to see, and in this case, the determination result is “Yes”. . If the result of the determination in step SC8 is “No”, the control section 103 makes a determination in step SC9.
[0071]
In step SC12, the control unit 103 enlarges the frame 311 (overlapping portion) shown in FIG. 6 and adds the enlarged scatter diagram 320 shown in FIG. 7 to the display unit 102 in parallel with the scatter diagram 310 (see FIG. 6). To be displayed.
[0072]
In step SC9, the control unit 103 determines whether or not there is a red plot as an outlier in the scatter diagram 310 (see FIG. 6) and the enlarged scatter diagram 320 (see FIG. 7). In this case, since there is a plot (data) corresponding to “14” (circled frame 312) displayed in red in the scatter diagram 310 illustrated in FIG. 8, the control unit 103 determines the determination result of step SC9. "Yes".
[0073]
In step SC10, the control unit 103 deletes the record corresponding to the specified identification number “14” from the original data table 200 and the identification number table 210. Here, the identification number “14” is a missing number. Thereafter, steps SC3 to SC10 are repeated until the determination result of step SC9 becomes “No”.
[0074]
Then, when the result of the determination in step SC9 is “No”, in step SC12, the control unit 103 stores the original data table 200 from which the record corresponding to the outlier has been deleted as processed data in the processed database 120.
[0075]
As described above, according to one embodiment (Operation Example 1), as shown in FIG. 6, each data (plot) in the data set visualized by the scatter diagram 310 is associated with each data (plot). Since the identification number for identification is displayed and the identification number corresponding to the outlier data is specified by the input unit 101 in the visualized data set, the efficiency of the detection of the outlier can be increased.
[0076]
Further, according to the embodiment (Operation Example 1), since the data corresponding to the specified identification number is removed as an outlier, the efficiency of removing the outlier can be increased.
[0077]
According to one embodiment (operation example 1), as shown in FIGS. 6 and 7, a portion (frame 311) where data (plot) overlaps in the visualized data set is enlarged and displayed. Therefore, even when data (plots) overlap, the efficiency of detecting outliers can be increased.
[0078]
According to one embodiment (Operation Example 2), as described with reference to FIG. 9, in the visualized data set, data (plot, identification number) as an outlier candidate satisfying a predetermined condition ) Is visualized by color coding, so that the efficiency of detecting outliers can be further increased.
[0079]
Further, according to one embodiment (Operation Example 3), as described with reference to FIG. 10, an identification number for identifying each data is displayed in association with each data in the visualized data set. Since outlier data that satisfies a predetermined condition (the number of clusters is 1) is removed from the visualized data set, the efficiency of detection and removal of outliers can be increased.
[0080]
An embodiment according to the present invention has been described in detail with reference to the drawings. However, a specific configuration example is not limited to the embodiment, and a design change within a range not departing from the gist of the present invention. The present invention is also included in the present invention.
[0081]
For example, in one embodiment described above, a program for realizing the function of the outlier detection support apparatus 100 shown in FIG. 1 is recorded on the computer-readable recording medium 500 shown in FIG. Each function may be realized by reading and executing the program recorded on the medium 500 into the computer 400 shown in FIG.
[0082]
The computer 400 shown in the figure includes a CPU (Central Processing Unit) 410 for executing the above program, an input device 420 such as a keyboard and a mouse, a ROM (Read Only Memory) 430 for storing various data, an arithmetic parameter and the like. (Random access memory) 440 for storing a program, a reading device 450 for reading a program from the recording medium 500, and an output device 460 such as a display or a printer.
[0083]
The CPU 410 realizes the above-described functions by reading the program recorded on the recording medium 500 via the reading device 450 and executing the program. Note that the recording medium 500 includes an optical disk, a flexible disk, a hard disk, and the like.
[0084]
【The invention's effect】
As described above, according to the first, sixth, and eighth aspects of the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and the visualized data is displayed. Since the data identifier corresponding to the outlier data is specified in the set, there is an effect that the efficiency of detecting the outlier can be increased.
[0085]
According to the second aspect of the present invention, a portion where data overlaps in the visualized data set is enlarged and displayed, so that even when there is data overlap, detection of outliers is required. The effect that efficiency can be raised is produced.
[0086]
According to the third aspect of the invention, in the visualized data set, data as an outlier candidate that satisfies a predetermined condition is visualized by color coding. There is an effect that it can be further increased.
[0087]
According to the fourth aspect of the present invention, since the data corresponding to the specified data identifier is removed as an outlier, the effect of removing outliers can be improved.
[0088]
According to the fifth, seventh, and ninth aspects of the present invention, a data identifier for identifying each data is displayed in association with each data in the visualized data set, and a predetermined identifier is displayed in the visualized data set. Since the outlier data satisfying the condition (1) is removed, the effect of detecting and removing outliers can be improved.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an embodiment according to the present invention.
FIG. 2 is a diagram showing various tables in the same embodiment.
FIG. 3 is a flowchart illustrating an operation example 1 of the same embodiment.
FIG. 4 is a scatter diagram 300 in the same embodiment.
FIG. 5 is a scatter diagram 310 in which identification numbers corresponding to the respective plots shown in FIG. 4 are assigned.
6 is a diagram showing a frame 311 in the scatter diagram 310 shown in FIG.
FIG. 7 is an enlarged scatter diagram 320 corresponding to the frame 311 shown in FIG.
FIG. 8 is a diagram showing a round frame 312 in the scatter diagram 310 shown in FIG. 5;
FIG. 9 is a flowchart illustrating an operation example 2 of the same embodiment.
FIG. 10 is a flowchart illustrating an operation example 3 of the same embodiment.
FIG. 11 is a block diagram showing a configuration of a modification of the same embodiment.
FIG. 12 is a diagram showing a conventional data table 10.
FIG. 13 is a scatter diagram 20 in which each record of the data table 10 shown in FIG. 12 is plotted.
[Explanation of symbols]
100 Outlier detection support device
101 Input unit
102 Display
103 control unit
104 storage unit
110 Original Database
120 Processed database

Claims (9)

Computer
Data identifier setting means for setting a data identifier for identifying each data in a data set which is an outlier detection target,
Visualization means for visualizing the data set,
Data identifier display means for displaying the data identifier in association with each data in the visualized data set,
Outlier specifying means for specifying a data identifier corresponding to outlier data in the visualized data set,
Outlier detection support program to function as a function. 2. The outlier detection support program according to claim 1, wherein the computer is caused to function as enlargement display means for enlarging and displaying a portion where data overlaps in the visualized data set. 3. The outlier detection support program according to claim 1, wherein the visualization unit visualizes data as an outlier candidate that satisfies a predetermined condition by color coding in the visualized data set. 4. The outlier according to claim 1, wherein the computer functions as an outlier that removes data corresponding to the data identifier specified by the outlier specifying unit as an outlier. Value detection support program. Computer
Data identifier setting means for setting a data identifier for identifying each data in a data set which is an outlier detection target,
Visualization means for visualizing the data set,
Data identifier display means for displaying the data identifier in association with each data in the visualized data set,
Removing means for removing outlier data satisfying a predetermined condition in the visualized data set;
Outlier detection support program to function as a function. A data identifier setting step of setting a data identifier for identifying each data in a data set that is an outlier detection target,
A visualization step of visualizing the data set;
A data identifier display step of displaying the data identifier in association with each data in the visualized data set,
An outlier specifying step for specifying a data identifier corresponding to outlier data in the visualized data set,
An outlier detection support method, comprising: A data identifier setting step of setting a data identifier for identifying each data in a data set that is an outlier detection target,
A visualization step of visualizing the data set;
A data identifier display step of displaying the data identifier in association with each data in the visualized data set,
A removing step of removing outlier data that satisfies a predetermined condition in the visualized data set;
An outlier detection support method, comprising: Data identifier setting means for setting a data identifier for identifying each data in a data set that is an outlier detection target,
Visualization means for visualizing the data set,
Data identifier display means for displaying the data identifier corresponding to each data in the visualized data set,
Outlier specifying means for specifying a data identifier corresponding to outlier data in the visualized data set,
An outlier detection support device comprising: Data identifier setting means for setting a data identifier for identifying each data in a data set that is an outlier detection target,
Visualization means for visualizing the data set,
Data identifier display means for displaying the data identifier corresponding to each data in the visualized data set,
Removing means for removing outlier data that satisfies a predetermined condition in the visualized data set;
An outlier detection support device comprising:

Images