JP5871842B2 - Information visualization apparatus, method, and program - Google Patents

Description

  The present invention relates to an information visualization apparatus, method, and program for drawing a data set.

There are several data set recording systems, such as RDB (Relational DataBase), CoachDB, and uTupleSpace, as data bases and file systems that record data such as sensors and make it possible to search a set of recorded data (for example, non-patent literature) 1).
Among them, uTupleSpace and CoachDB are called schemaless databases, and manage a wide variety of data, that is, a set of heterogeneous and multi-attribute data having different types and different numbers of attributes. For example, in uTupleSpace, data in a format in which an attribute name and an attribute value corresponding to the attribute name are freely expressed by one or more “key = value” pairs called a uTuple format are managed. At this time, “key = value” is synonymous with “attribute name = attribute value”. For example, by expressing certain data as “sensor ID = 1, time = 3, temperature = 20”, three attribute names of sensor ID, time, and temperature and corresponding attribute values are expressed.
On the other hand, there is a trie tree as a data structure used for an index structure of a character string set (see Non-Patent Document 2, for example). A trie tree is a data structure having a tree structure.
In the trie tree, a character is assigned to each node constituting the tree structure, and a character string is represented by a sequence of characters appearing on a route from the root node to the leaf node.

T. Nakamura, K. Kashiwagi, Y. Arakawa, and M. Nakamura, "Design and implementation of new uTupleSpace enabling storage and retrieval of large amount of schema-less sensor data," 2nd Intl. Workshop on Enablers for Ubiquitous and Context- Aware Services on Sensor Networks (EUCASS2011), pp.414-420, July 2011. Fredkin E, “Trie memory,” Communications of the ACM, Vol. 3, No. 9, pp.490-500, 1960.

Understanding what data is recorded in a database or file system is useful in the use of the data and maintenance of the system. For example, it is useful when considering how to use the recorded data, designing a search sentence represented by SQL, or determining data to be deleted when the data capacity increases and the storage capacity is under pressure. .
However, especially when recording a wide variety of data, such as uTupleSpace, the type of data increases as the data increases, and it is difficult to know what data is recorded from the data list. It becomes.
In such a case, simply express what kind of data exists, that is, what kind of combination of attribute names is recorded, and what kind of attribute names are included. I can do it.
For example, there is a trie tree as a conventional method.
In a trie tree, a character is assigned to each node constituting the tree structure, and a character string is represented by a sequence of characters appearing on a route from a root node to a leaf node. I can do it.
However, the characters assigned to each node do not make sense alone, and the characters assigned to each node are meaningful in the order of assignment, so combinations are not considered.
For this reason, attribute names that are meaningful character strings cannot be used for an overview of the set of combinations.

  The present invention has been made paying attention to the above circumstances, and an object thereof is to provide an information visualization apparatus, method, and program capable of easily grasping a wide variety of data.

  One aspect of the present invention for achieving the above object is to generate at least an appearance pattern that is a combination of attribute names included in one data from a set of data including one or more attribute names; From the list of appearance patterns, an appearance pattern tree in which each node in the tree structure includes 0 or 1 or more attribute names, and a set of attribute names that appear once on the route from the root node to the leaf node represents the appearance pattern. Constructing means for constructing, wherein the constructing means extracts at least one grouping target attribute name included in common in one or more node groups having a common parent node in the appearance pattern tree, A recombination process is performed in which the grouping target attribute names are grouped into a single node.

That is, according to the present invention, an appearance pattern list that is a list of appearance patterns of attribute names included in each data is generated from a set of data including at least one attribute name recorded in a database or file system such as uTupleSpace. What kind of data exists by constructing an appearance pattern tree in which a set of attribute names that appear once on the path from the root node to the leaf node from the list of appearance patterns represents the appearance pattern Various sets of data can be drawn with a tree structure so that it is possible to easily grasp whether data including such combinations of attribute names is recorded.
Furthermore, according to the present invention, in the appearance pattern tree, at least one grouping target attribute name that is commonly included in one or more nodes having a common parent node is extracted, and the grouping target attribute name is a single group name. By repeatedly executing the recombination process that starts with nodes, even if all of the attribute names held by each node are not the same, they can be bundled if some of them are in common. The total number of names can be reduced. That is, the appearance pattern list can be drawn with a simplified tree structure.

  Furthermore, according to the present invention, attribute names appearing in the appearance pattern list and their appearance frequencies are calculated, and attribute names with a high appearance frequency are preferentially grouped out as target attribute name candidates. The attribute names included can be bundled, and the effect of reducing the total value of the number of attribute names included in each node is greater than when bundled with attribute names having a low appearance frequency. Therefore, the appearance pattern list can be drawn with a simplified tree structure. In addition, among the attribute names in which each occurrence pattern appears once on the route from the root node to the leaf node, the attribute name having a high appearance frequency is preferentially classified, so the closer the root node is to the appearance pattern Since attribute names having a high appearance frequency in the list are included, it is possible to easily grasp which attribute name has a high appearance frequency from the appearance pattern tree.

Furthermore, according to the present invention, by allowing each node to hold a plurality of attribute names instead of holding one attribute name, the number of nodes in the appearance pattern tree can be reduced, and the appearance pattern list can be displayed. It is possible to draw with a simplified tree structure.
Further, according to the present invention, when there are a plurality of grouping target attribute name candidates having the same number of appearance frequencies in one or more node groups having a common parent node, each node has a grouping target attribute name candidate. , The number of attribute names combined with attribute names other than the target attribute name candidates that are common to all nodes including the target attribute name candidates, that is, the number of attribute names included in the common attribute name group, Calculate the total value of the attribute names included in the node and the number of attribute names excluding the common attribute name group, and group the common attribute name group in the target attribute name candidate that minimizes the calculated total value By using attribute names, the total number of attribute names included in each node can be further reduced, and the appearance pattern list can be drawn with a simplified tree structure.

The block diagram which shows the information visualization apparatus which concerns on this embodiment. The figure which shows the specific example of the collection of the data accumulate | stored in a data storage part. The flowchart which shows operation | movement of an information visualization apparatus. The flowchart which shows the appearance pattern production | generation process in an appearance pattern production | generation part. The figure which shows the specific example of an appearance pattern production | generation process result. The flowchart which shows the tree structure construction process in a tree structure construction part. The figure which shows the specific example of an appearance frequency list. The figure which shows the specific example of an initial appearance pattern tree. The flowchart which shows the detail of the recombination process of an appearance pattern tree. The flowchart which shows the detail of an evaluation process. The figure which shows the 1st progress of the recombination process of an appearance pattern tree. The figure which shows the 2nd progress of the recombination process of an appearance pattern tree. The figure which shows the 3rd progress of the recombination process of an appearance pattern tree. The figure which shows the 4th progress of the recombination process of an appearance pattern tree. The figure which shows the final appearance pattern tree. The figure which shows an example which added the node showing the frequency | count of appearance of each appearance pattern to the appearance pattern tree.

Hereinafter, an information visualization apparatus, method, and program according to an embodiment of the present disclosure will be described in detail with reference to the drawings. Note that, in the following embodiments, the same numbered portions are assumed to perform the same operation, and repeated description is omitted.
The information visualization apparatus according to the present embodiment will be described with reference to the block diagram of FIG.
The information visualization apparatus 100 according to the present embodiment includes a data storage unit 101, an appearance pattern generation unit 102, a tree structure construction unit 103, and a display unit 104.

The data storage unit 101 stores data including one or more attribute names. The attribute name is the name of the attribute, and examples thereof include temperature, humidity, current, voltage, fluid flow rate, substance concentration, brightness, noise, position, and acceleration. The data will be described later with reference to FIG.
The appearance pattern generation unit 102 extracts a combination of attribute names included in each data from a set of data stored in the data storage unit 101, and generates an appearance pattern that is a combination of attribute names for each data. The appearance pattern generation unit 102 then generates an appearance pattern list that is a list of appearance patterns.

The tree structure construction unit 103 creates an appearance frequency list from the appearance pattern list, constructs an initial appearance pattern tree from the appearance pattern list, and constructs an appearance pattern tree from the initial appearance pattern tree and the appearance frequency list. The appearance pattern tree is a tree structure in which a set of attribute names that appear once on a route from a root node to a leaf node indicates an appearance pattern.
The display unit 104 displays the appearance pattern tree.

Next, the tree structure construction unit 103 will be described in detail.
The tree structure construction unit 103 includes a list creation unit 105 and a rearrangement unit 106.
The list creation unit 105 calculates an appearance frequency for each attribute name in the appearance pattern list based on the appearance pattern list, and creates an appearance frequency list.
The rearrangement unit 106 constructs an initial appearance pattern tree based on the appearance pattern list. The initial appearance pattern tree is a tree structure in which a node including all attribute names included in each appearance pattern in the appearance pattern list is a child node of the root node. Further, the recombination unit 106 refers to the appearance frequency list and repeats the recombination process with the child node group of the root node in the initial appearance pattern tree as the processing target node group. Thereby, an appearance pattern tree is constructed.

Next, an example of a set of data stored in the data storage unit 101 will be described with reference to FIG.
In this embodiment, the attribute name 201 and the attribute value 202 are associated with each other as one attribute 203, and a set of one or more attributes 203 is referred to as one data 204. The attribute value 202 is a value corresponding to each attribute name 201.

In this embodiment, data is described in a data format composed of one or more attributes, that is, a set of “attribute name = attribute value”. For example, in the example of FIG. 2, eight pieces of data 204 from data (a) to data (ku) are stored. As an example, the data (d) is described in a format such as “subject = project_A, date = 602, switch = on”. That is, the data (d) includes three attribute names “subject”, “date”, and “switch”, and attribute values corresponding to these attribute names are expressed.
In this embodiment, the data is described in a data format composed of one or more “attribute name = attribute value” pairs. However, the present invention is not limited to this, and a data format composed of only attribute names may be used. In addition, the pair of “attribute name = attribute value” is separated by a comma, but the present invention is not limited to this, and a space or another symbol may be used. For example, a format of “attribute value 1 / attribute name 2 / attribute name 3 /...” May be used. Also, “attribute name 1, attribute value type 1 (int, float, string, etc.), attribute value 1, attribute name 2, attribute value type 2 (int, float, string, etc.), attribute value 3),... A format that includes information other than the attribute name and attribute value, such that a plurality of three terms are arranged in one set.

  Note that the present invention is not limited to the above-described example, and data acquired via the Web and the Internet may be used, for example, as long as it can be expressed only by the attribute name = attribute value format or the attribute name. In the example of FIG. 2, data including 2 to 6 attribute names is given as an example, but any number of attribute names may be included as long as it is 1 or more. Further, although the case where the number of data stored in the data storage unit 101 is eight has been described as an example, the present invention is not limited thereto, and more data may be stored.

Next, the operation of the information visualization apparatus 100 will be described with reference to the flowchart of FIG.
In step S <b> 301, the appearance pattern generation unit 102 generates an appearance pattern list from the set of data stored in the data storage unit 101. Details of the appearance pattern generation processing in the appearance pattern generation unit 102 will be described later with reference to FIG.
In step S302, the tree structure construction unit 103 constructs an appearance pattern tree from the appearance pattern list. Details of the tree structure building process in the tree structure building unit 103 will be described later with reference to FIGS. 6, 9, and 10.
In step S303, the display unit 104 draws an appearance pattern tree. Above, operation | movement of the information visualization apparatus 100 is complete | finished.

Next, the appearance pattern generation processing in the appearance pattern generation unit 102 shown in step S301 will be described with reference to the flowchart of FIG.
In step S401, a set of data at the time of operation in step S301 is received from the set of data stored in the data storage unit 101.
In step S402, one of the data received in step S401 is selected, and an attribute name included in this data is extracted.
In step S403, an appearance pattern is generated from one or more attribute names extracted in step S402.
In step S404, the appearance pattern generated in step S403 is compared with the appearance patterns included in the appearance pattern list to determine whether or not the same appearance pattern already exists in the appearance pattern list. If the same appearance pattern exists in the appearance pattern list, the process proceeds to step S406. If the same appearance pattern does not exist in the appearance pattern list, the process proceeds to step S405. When comparing the appearance patterns, the appearance order of the attribute names in the data is not limited, and the same appearance pattern may be used as long as the combination is the same attribute name. For example, “subject, date, switch, id” and “date, id, subject, switch” have the same appearance pattern. In the case of data to be processed first, the appearance pattern to be compared does not exist in the appearance pattern list, and is added to the appearance pattern list.
The appearance pattern list may be generated by receiving data one by one from the set of data stored in the data storage unit 101.
If the data contains the same attribute name, it can be handled by combining them into one.

In step S405, a new appearance pattern is added to the appearance pattern list.
In step S406, it is determined whether or not an appearance pattern has been generated for all data. If the appearance pattern is generated for all data, the appearance pattern extraction process is terminated. If no appearance pattern is generated for all data, that is, if there is data for which no appearance pattern is generated, step S402 is performed. Return to and repeat the same process.
Thus, the appearance pattern generation process ends.

Next, a specific example of the appearance pattern generation process will be described with reference to FIG.
FIG. 5 shows an example in which the appearance pattern list 500 is generated from the data set shown in FIG.
As a specific process, first, a set of data at the time of performing an appearance pattern generation process, that is, a set of data shown in FIG. 2 is received. Subsequently, data (A) is selected from the data set of FIG. 2, and attribute names “subject, date, positionx, positiony, id, temp” are extracted from the data (A), and combinations of these attribute names are represented as appearance patterns. Generate as After that, it is determined whether there is the same appearance pattern in the appearance pattern list, but since it is the data to be processed first, it is stored in the appearance pattern list as the appearance pattern 501 (A).
Subsequently, since no appearance pattern has been generated for all data, the attribute name “temp, subject, date, positionx, positiony, id” is similarly extracted from the next data (A) to generate an appearance pattern. . At this time, the appearance pattern of the data (A) is a combination of the same attribute names as the appearance pattern 501 (A) that is the appearance pattern of the data (A), and thus is not added to the appearance pattern list.
Subsequently, since the appearance pattern has not been generated for all the data, the attribute name “subject, id, positionx, positiony, magnesium” is extracted from the data (c) in FIG. 2 to generate the appearance pattern. The appearance pattern “subject, id, positionx, positiony, magnesium” is added to the appearance pattern list as the appearance pattern 501 (B) because the same appearance pattern does not exist in the appearance pattern list. What is necessary is just to perform the above process with respect to all the data. As a result, as shown in FIG. 5, the appearance pattern 501 (C) from data (D), the appearance pattern 501 (D) from data (E), the appearance pattern 501 (E) from data (F), and the data (G) , An appearance pattern 501 (F) and an appearance pattern 501 (G) are generated from the data (K) and added to the appearance pattern list.

Here, although the result of generating the appearance patterns in the order of the notation from the data (a) to the data (ku) in FIG. 2 is shown, the present invention is not limited to this and may be extracted in the reverse order. Further, the appearance patterns may be generated while sorting the appearance patterns included in the list of appearance patterns being generated according to the number of attribute names. Thereby, the appearance patterns can be compared efficiently. For example, the appearance patterns that may have the same number of attribute names as the appearance patterns with 5 attribute names are only the appearance patterns with 5 attribute names, and the appearance pattern list is sorted by the number of attribute names. Thus, it is possible to easily find an appearance pattern having five attribute names from the appearance pattern list.
In addition, in the appearance pattern list being created, the appearance patterns may be generated while sorting the attribute names constituting the appearance patterns in the order of the dictionary. Thereby, since the display order of the attribute names constituting the appearance pattern is uniquely determined, it is possible to easily compare the appearance patterns.
In the present embodiment, the appearance pattern is such that the attribute name is separated from the attribute name by a comma. However, the present invention is not limited to this, and a space, another symbol, or the like may be used.

Next, the tree structure building process in the tree structure building unit 103 shown in step S302 will be described with reference to the flowchart of FIG.
In step S601, the list creation unit 105 extracts attribute names from the appearance pattern list, calculates the appearance frequency of each attribute name in the appearance pattern list, and creates an appearance frequency list.
In step S602, the rearrangement unit 106 constructs an initial appearance pattern tree based on the appearance pattern list.

In step S603, the rearrangement unit 106 refers to the appearance frequency list and repeatedly executes the recombination process of the initial appearance pattern tree to construct the appearance pattern tree.
Next, a specific example of the appearance frequency list created in step S602 will be described with reference to FIG.
The appearance frequency list 700 is a table in which attribute names 701 and their appearance frequencies 702 are associated with each other. The appearance frequency 702 may be calculated by counting the number of identical attribute names that appear in the appearance pattern list for each attribute name. The appearance frequency list 700 illustrated in FIG. 7 calculates the appearance frequency of the attribute name with reference to the appearance pattern list in FIG. 5. For example, the attribute name 701 “subject” is 5 in the appearance pattern list in FIG. 5. Since there are three appearance frequencies 702 “5” and three attribute names 701 “date”, they are stored in association with the appearance frequency 702 “3”.

Next, a specific example of the initial appearance pattern tree constructed in step S602 will be described with reference to FIG.
The initial appearance pattern tree 800 shown in FIG. 8 refers to the appearance pattern list of FIG. 5, the parent node is the root node 801, and the node 802 to the node 808 including all the attribute names included in each appearance pattern in the appearance pattern list. As its child nodes. Appearance patterns (A) to (G) in FIG. 5 correspond to child nodes 802 to 808, respectively. Here, the root node 801 is an empty node, but the root node 801 may be indicated by a symbol or the like.

Next, details of the recombination process in step S603 will be described with reference to the flowchart of FIG. The following processing from step S901 to step S907 is referred to as recombination processing 900.
In the first process, the child node group of the root node in the initial appearance pattern tree becomes the process target node group. The processing target node group is a child node group having a common parent node, and is one or more nodes to be subjected to recombination processing.
In step S901, it is determined whether or not a grouping operation can be performed on the processing target node group. If the grouping operation is possible, the process proceeds to step S902. If the grouping operation is not possible, that is, if the grouping operation is not possible, the tree includes the attribute names included in each node in the processing target node group and the relationship between the nodes. Returns structure information.
Also, whether or not the grouping operation is possible is determined by determining whether or not there is a common attribute name among the nodes included in the processing target node group. If there is a common attribute name, the grouping operation is performed. If there is no common attribute name, it may be determined that the grouping operation is impossible.
In step S902, with reference to the appearance frequency list, among the attribute names included in the processing target node group, the attribute name having the highest appearance frequency is bundled and extracted as a target attribute name candidate. If there are a plurality of attribute names having the highest appearance frequency, each attribute name is a candidate for a target attribute name.
In step S903, evaluation processing is performed for each bundle target attribute name candidate. The evaluation process will be described later with reference to FIG.

In step S904, the grouping target attribute name is determined from the evaluation processing result. When there are a plurality of grouping target attribute name candidates, the grouping target attribute name may be determined from the result of the evaluation process in step S903. In the present embodiment, the common attribute name group when the evaluation process is performed on the grouping target attribute name candidate having the smallest evaluation value as a result of the evaluation process is defined as the grouping target attribute name. The common attribute name group will be described later.
In step S905, the recombination process 900 is executed with a node group that does not include the grouping target attribute name in the processing target node group as a new processing target node group. That is, the entire process shown in FIG. 9 is recursively called to execute the recombination process.
In step S906, for the node group including the grouping target attribute name in the processing target node group, a grouping operation for grouping the grouping target attribute name is performed. Specifically, one child node is generated for each node including the grouping target attribute name in the processing target node group, and attribute names other than the grouping target attribute name are transferred to the respective child nodes. At this time, if no attribute name other than the grouping target attribute name is included, no child node includes the attribute name. A symbol φ indicates that no node includes any attribute name. Further, a node group including only the grouping target attribute name is merged into one node, and each generated child node is a child node of the merged node. In the present embodiment, the symbol φ indicates that no node includes any attribute name, but other symbols may be used instead of the symbol “φ”, or an empty node may be used. . In addition, a common parent node may be generated for the node including the grouping target attribute name in the processing target node group, and the grouping target attribute name may be transferred to the generated parent node.
In step S907, the child node group of the node including only the grouping target attribute name in the processing target node group, that is, the child node group having the node including only the grouping target attribute name fused in step S906 as a parent node is newly processed. A recombination process 900 is executed as a target node group. That is, the entire process shown in FIG. 9 is recursively called to execute the recombination process.
When the process of step S907 is completed, the tree structure information recombined by the recombination process is returned. The recombination process is recursively executed by the above procedure, and an appearance pattern tree is constructed.
In the present embodiment, the recombination process is recursively executed as means for repeatedly executing the recombination process. However, each time a recombination process is executed, a group of nodes that can be grouped and operated, that is, a child having a common parent node. Search for a node group that satisfies the condition that it is a node group and the attribute name common to the nodes in the child node group is included, and perform recombination processing until there is no node group that can be grouped. It is also good.

Next, the evaluation process in step S903 will be described with reference to the flowchart of FIG.
In step S1001, for the grouping target attribute name candidate X that is the evaluation target, a group other than the grouping target attribute name candidate X that exists in common for all nodes including the grouping target attribute name candidate X in the processing target node group. The attribute name Y is extracted. Note that the attribute name Y other than the grouping target attribute name candidate X that exists in common does not have to exist, and a plurality of attribute names Y may exist.
In step S1002, the total number of the common attribute name group including the bundled target attribute name candidate X and the attribute name Y and the number of attribute names other than the common attribute name group included in each node in the processing target node group Is calculated as the evaluation value of the target attribute name candidate X. When the process of S1002 ends, the evaluation value is returned. The evaluation process ends here.

In this embodiment, an example is shown in which recombination processing is performed from a node group that does not include the grouping target attribute name in the processing target node group. However, the recombination processing is performed first from the node group that includes the grouping target attribute name in the processing target node group. Processing may be performed.
Next, a specific example of the appearance pattern tree recombination processing shown in FIG. 9 and the evaluation processing shown in FIG. 10 will be described with reference to the processing progress from FIG. 11 to FIG.
FIG. 11 is an example of extracting the bundled target attribute name candidates from the initial appearance pattern tree. Since the child node group of the root node in the initial appearance pattern tree is the first processing target node group, it is determined whether the processing target node group can be bundled. Here, since attribute names such as “subject” and “date” are included in common among nodes included in the processing target node group, it is determined that the grouping operation is possible.
Subsequently, among the attribute names common among the nodes included in the processing target node group, the attribute name having the highest appearance frequency is extracted. Referring to the appearance frequency list in FIG. 7, the attribute name “subject” has the maximum appearance frequency of 5 times, so the attribute name “subject” is extracted as a target attribute name candidate.

Subsequently, an evaluation process is performed on “subject” which is a grouping target attribute name candidate.
First, regarding the grouping target attribute name candidate “subject” to be evaluated, the grouping target attribute name candidate “subject” that exists in common in all nodes including the grouping target attribute name candidate “subject” in the processing target node group. ”Is extracted. Here, there is no attribute name other than the grouping target attribute name candidate “subject” that is common to all nodes including the grouping target attribute name candidate “subject” in the processing target node group.
Subsequently, when the evaluation value of the grouping target attribute name candidate “subject” is calculated, it is common to all the nodes including the grouping target attribute name candidate “subject” and the grouping target attribute name candidate “subject” in the processing target node group. Thus, the number of attribute names including attribute names other than the grouping target attribute name candidate “subject”, that is, the number of attribute names included in the common attribute name group is 1 + 0 = 1. In addition, the number of attribute names other than the common attribute name group included in each node in the processing target node group is 5, 4, 2, 3, 3, 3, 2 from the child nodes 802 to 808, respectively. Therefore, the evaluation value of the grouping target attribute name candidate “subject” is calculated as 1 + 22 = 23.
Here, the grouping target attribute name candidate is only “subject”, and therefore the grouping target attribute name group “subject” when the evaluation process is performed on the grouping target attribute name candidate “subject” having the smallest evaluation value. It is determined as a grouping target attribute name.

  As a result, the processing target node group in the initial appearance pattern tree shown in FIG. 11 includes a node group N-1 including the grouping target attribute name “subject”, and a node group N− including no grouping target attribute name candidate “subject”. It can be classified as 2.

Subsequently, a recombination process is performed in which a node group that does not include the grouping target attribute name “subject”, that is, the node group N-2 is a new processing target node group. Here, since there are “object” and “status” as attribute names common to a plurality of nodes in the processing target node group, it is determined that the grouping operation is possible. Subsequently, among the attribute names common among the nodes included in the processing target node group, the attribute name having the highest appearance frequency is extracted. Referring to the appearance frequency list of FIG. 7, since the attribute names “object” and “status” have the maximum appearance frequency of 2 times, the attribute names “object” and “status” are bundled and extracted as target attribute name candidates. Furthermore, evaluation processing is performed on the attribute name candidates that are to be bundled. First, with regard to the grouping target attribute name “object” to be evaluated, the grouping target attribute name candidate “object” that is common to all nodes including the grouping target attribute name candidate “object” in the processing target node group. Extract attribute names other than. At this time, the attribute name “status” is extracted. Accordingly, the attribute name candidate “object” and the attribute names “status” other than the grouping target attribute name candidate that exist in common in all nodes including the grouping target attribute name candidate “object” in the processing target node group. The number of common attribute name groups including “1” is 1 + 1 = 2. In addition, since the number of attribute names other than the common attribute name group included in each node in the processing target node group is only “gate_num” in the child node 807, the evaluation value of the grouping target attribute name candidate “object” is 2 + 1 = 3 is calculated.
Similarly, when calculating the evaluation value of the grouping target attribute name candidate “status”, the grouping target attribute that is common to all nodes including the grouping target attribute name candidate “status” in the processing target node group. “Object” exists as an attribute name other than the name candidate “status”. At this time, since the number of attribute names other than the common attribute name group included in each node in the processing target node group is only “gate_num” in the child node 807, the evaluation value of the grouping target attribute name candidate “status” is 2 + 1 = 3 is calculated.

Subsequently, the common attribute name group when the evaluation target attribute name candidate with the smallest evaluation value is evaluated is used as the target attribute name. In this case, either of the target attribute name candidates is selected. Even the evaluation value is equally minimum. In the present embodiment, when there are a plurality of grouping target attribute name candidates having the smallest evaluation value, the grouping target attribute name candidate that has been subjected to the evaluation process is selected. That is, the grouping target attribute name candidate “object” is selected, and the common attribute name group “object, status” is determined as the grouping target attribute name.
In this embodiment, when there are a plurality of bundle target attribute name candidates having the smallest evaluation value, the bundle target attribute name candidate that has been evaluated first is selected. However, it is also possible to select a grouping target attribute name candidate having a large number of common attribute name groups. In addition, when there are a plurality of grouping target attribute name candidates, an arbitrary grouping target attribute name candidate may be selected.

Next, the grouping operation of the grouping target attribute name “object, status” is performed. Specifically, first, one child node is generated for each node including the grouping target attribute name “object, status” in the processing target node group, that is, the nodes 807 and 808 in FIG. The attribute name of is transferred to each child node. As a result, the node 807 and the node 808 are divided into a node 1201-1 and its child node 1202, and only a node 1201-2 and its child node 1203 that include only the grouping target attribute name “object, status” as shown in FIG. The Note that since the node 1203 is a node that does not include any attribute name, it is indicated by the symbol φ.
Further, the node 1201-1 and the node 1201-2 including only the grouping target attribute name “object, status” in the processing target node group are merged into one node 1301, and the node 1202 and the node 1203 are child nodes of the node 1301. And The result is the tree structure shown in FIG. Subsequently, in the processing target node group N-2, a new node group N-2-c including the node 1202 and the node 1203, that is, a child node group of the node 1301 including only the grouping target attribute name “object, status” is newly added. The recombination process (step S900) is performed as a processing target node group, but it is determined that the grouping operation is not possible because there is no common attribute name among the nodes included in the node group N-2-c. , The recombination process (step S900) for the node group N-2-c ends. As a result, the recombination process (step S900) of the processing target node group N-2 is also terminated.

  Subsequently, a grouping operation is performed on the processing target node group in the initial appearance pattern tree. Since the grouping target attribute name is determined as “subject”, one child node is generated for each of the nodes including the grouping target attribute name in the processing target node group, that is, the nodes 802 to 806 in FIG. However, attribute names other than the target attribute name are transferred to their child nodes. Further, nodes including only the grouping target attribute name “subject” in the processing target node group are merged into one node 1401, and the nodes 1402 to 1406 are child nodes of the node 1401. The result is shown in FIG.

Subsequently, for the processing target node group in the initial appearance pattern tree, a child node group of the node 1401 including only the grouping target attribute name “subject”, that is, a node group N-1-c including the nodes 1402 to 1406 is newly added. Recombination processing is performed as a group of nodes to be processed.
At this time, the grouping operation is possible, and the attribute names “date”, “id”, “positionx”, and “positiony” with the maximum appearance frequency are each evaluated as the target attribute name candidates.
For example, when the evaluation value of the grouping target attribute name candidate “positionx” is calculated, it exists in common for all nodes including the grouping target attribute name candidate “positionx” in the processing target node group N-1-c. There is “positiony” as an attribute name other than the grouping target attribute name candidate “positionx”. Therefore, the number of common attribute name groups is “positionx, position”, and the number of attribute names other than the common attribute name group included in each node in the processing target node group is “date” in order from the node 1402 to the node 1406. , Id, temp ”, two“ id, magnesium ”, two“ date, switch ”, one“ post_num ”, and three“ date, id, humidity ”.

  Therefore, the total number is 11, and the evaluation value of the grouping target attribute name candidate “positionx” is 2 + 11 = 13. Similarly, when evaluation values are calculated for the grouping target attribute name candidates “date”, “id”, and “positiony”, the evaluation values are 15, 15, and 13, respectively. As a result, among the grouping target attribute names with the smallest evaluation value, the grouping target attribute name candidate “positionx” that has been evaluated first is selected, and the common attribute name group “positionx, positiony” is grouped. Determine as the target attribute name.

Similarly, the appearance pattern tree can be constructed by recursively executing the recombination process until the initial appearance pattern tree cannot be bundled and operated. FIG. 15 shows a final appearance pattern tree generated by the tree structure construction unit 103 in this embodiment.
In this way, by performing recombination processing, it is possible to construct a tree structure that further simplifies the appearance pattern list, and by drawing this, there is data including any attribute name in the data set Can be easily grasped.
Furthermore, the effect of reducing the total value of the number of attribute names included in each node in the appearance pattern tree is increased by performing the recombination process in which attribute names with high appearance frequency are preferentially bundled and used as target attribute name candidates. Therefore, the appearance pattern list can be drawn with a simplified tree structure. Furthermore, since the attribute name with higher appearance frequency is included as it is closer to the root node, it is possible to easily grasp what kind of attribute name is included in the appearance pattern list.

In this embodiment, as shown in a node 1501 in FIG. 15, attribute names with different appearance frequencies may be included in one node. However, the present invention is not limited to this, and nodes including attribute names with different appearance frequencies may be included. It may be divided so that there are no nodes including attribute names having different appearance frequencies.
For example, in the example of the node 1501, since the node 1501 includes an attribute name “date” having an appearance frequency of 3 times and an attribute name “temp” having an appearance frequency of 1 time, The child node may be divided into nodes including the attribute name “temp”. Thereby, the difference in appearance frequency between attribute names can be read in more detail from the appearance pattern tree.
If there is one or more attribute names common to all appearance patterns in the appearance pattern list, the tree structure may be drawn with the root node omitted.

Alternatively, the number of appearances of each appearance pattern in the data set may be counted, and a node representing the number of appearances of the data of each appearance pattern in the data set may be added to the appearance pattern tree for drawing. The number of appearances is a number indicating how many times the data of each appearance pattern has appeared, that is, how many have been stored in the set of data stored in the data storage unit.
FIG. 16 shows an example in which a node representing the number of appearances of each appearance pattern is added to the appearance pattern tree.
For example, a node 1601 is added to the end of the leftmost node, and “2” is described in the node 1601. Thereby, it can be seen that the appearance pattern “subject, positionx, positiony, id, date, temp” appears twice in the data set. Also, drawing may be performed by omitting an empty node (φ node) that does not include any attribute name.

Next, a display example on the display unit 104 will be described.
For example, the display unit 104 draws the appearance pattern tree illustrated in FIG. 15 on the display device on a display or the like. In the present embodiment, the appearance pattern tree is drawn with a two-dimensional tree structure, but may be drawn with a three-dimensional tree structure using a technique such as Cone Tree (for example, George G. Robertson, See Jock D. Mackinlay, Stuart K. Card, "Cone Trees: Animated 3D Visualizations of Hierarchical Information," Proceedings of ACM Conference on Human Factors in Computing Systems (CHI'91), ACM Press, 1991, pp.189-194. ). In addition, the parent-child relationship between nodes in a tree structure may be drawn using a data jewel box or Venn diagram that represents the inclusion relationship of a two-dimensional polygon or closed curve (for example, Takayuki Ito, Yasumasa Tominaga, Yuko Ikehata) , “Data Jewel Box: Graphics Showcase of Large-Scale Hierarchical Data,” Information Processing Society of Japan, Graphics & CAD Study Group, 2001-CG-104, pp.65-70, 2001.).
Although not shown, the image may be drawn on a paper medium or the like by a printer connected to the information visualization apparatus 100 shown in FIG. Furthermore, the drawing image of the appearance pattern tree and the structure information of the appearance pattern tree, that is, the connection information between the nodes and the attribute name information included in each node are sent to the client computer connected to the information visualization apparatus 100 via the network. Also good. Note that such a display process may use an existing drawing library, and a detailed description thereof is omitted here.

According to the information visualization apparatus according to the present embodiment described above, an appearance pattern tree in which an appearance pattern indicating a combination of attribute names is generated from various data sets and a list of appearance patterns is drawn in a tree structure. By constructing, it is possible to overlook what kind of attribute name combination is recorded, and it is possible to easily grasp a wide variety of data sets.
Further, by making it possible to hold a plurality of attribute names in one node, the number of nodes in the appearance pattern tree can be reduced, and a simplified tree structure can be obtained.

Furthermore, in one or more node groups having a common parent node, attribute names that appear in common in one or more nodes are bundled into a single node, so that a plurality of attribute names held by each child node can be obtained. Even if they are not all the same, they can be bundled if they are partly in common, so that the appearance pattern list can be drawn with a simplified tree structure.
In addition, by performing the recombination process that preferentially bundles attribute names with high appearance frequency into target attribute names, the total number of attribute names included in each node in the appearance pattern tree can be reduced, and the appearance pattern The list can be drawn with a simplified tree structure. Furthermore, since the attribute name with higher appearance frequency is included as it is closer to the root node, it is possible to easily grasp what kind of attribute name is included in the appearance pattern list.
Furthermore, when there are a plurality of grouping target attribute name candidates that can be grouped in one or more node groups having a common parent node, the grouping target attribute name candidate having the smallest evaluation value is selected, It is possible to reduce the total value of the number of attribute names included in each node in the appearance pattern tree by setting the common attribute name group at the time of performing the evaluation process for the grouping target attribute name as the grouping target attribute name. The appearance pattern list can be drawn with a simplified tree structure.
The information visualization apparatus according to the present embodiment can be realized by a computer and a program, and can be recorded on a recording medium or provided through a network.

  In short, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  DESCRIPTION OF SYMBOLS 100 ... Information visualization apparatus, 101 ... Data storage part, 102 ... Appearance pattern generation part, 103 ... Tree structure construction part, 104 ... Display part, 105 ... List creation part, 106 ... rearrangement unit, 201,701 ... attribute name, 202 ... attribute value, 203 ... attribute, 204 ... data, 500 ... appearance pattern list, 501 ... appearance pattern, 700 ... appearance frequency list, 702 ... appearance frequency, 800 ... initial appearance pattern tree, 801 ... root node, 802 to 808, 1201-1, 1201-2, 1202, 1203, 1301, 1401 1406, 1501, 1601... Nodes.

Claims (9)

Generating means for generating at least an appearance pattern that is a combination of attribute names included in one data from a set of data including one or more attribute names;
From the list of appearance patterns, each node in the tree structure includes 0 or 1 or more attribute names, and an appearance pattern tree in which a set of attribute names that appear once on a route from a root node to a leaf node represents an appearance pattern A construction means for constructing
The construction means extracts at least one grouping target attribute name that is commonly included in one or more node groups having a common parent node in the appearance pattern tree, and sets the grouping target attribute name as a single group name. An information visualization device characterized by performing recombination processing in a node.   The information visualization apparatus according to claim 1, wherein the construction unit repeatedly executes the recombination process. The construction means creates an appearance frequency list in which attribute names included in the appearance patterns are associated with the appearance frequencies of the attribute names in all appearance patterns;
Based on the appearance frequency list, recombination means for constructing the appearance pattern tree by executing the recombination process as the bundled target attribute name in preference to the attribute name having a high appearance frequency among the attribute names; The information visualization apparatus according to claim 1, further comprising:   In the node group including the grouping target attribute name, the construction unit extracts an attribute name included in the grouping group along with the grouping target attribute name, and includes the grouping target attribute name and the grouping target attribute name. The information according to any one of claims 1 to 3, wherein a group of common attributes combined with attribute names included together is grouped into a single node as a new grouping target attribute name. Visualization device.   The recombination means, when there are a plurality of grouping target attribute names having the same appearance frequency, for each grouping target attribute name, the number of attribute names included in the common attribute name group and the attribute name included in each node An evaluation value that is the total value of the number of attribute names excluding the common attribute name group is calculated, and the common attribute name group in the grouping target attribute name having the smallest evaluation value is grouped as a single target attribute name. The information visualization apparatus according to claim 4, wherein the information visualization apparatus is bundled into one node.   The information visualization according to any one of claims 1 to 5, wherein the data is in a data format including at least a set of one or more attribute names and a value corresponding to the attribute name. apparatus.   The information visualization apparatus according to claim 1, wherein the data is u Tuple format data. An information visualization method executed by an information visualization apparatus including a generation unit and a construction unit,
The generating means generates at least an appearance pattern that is a combination of attribute names included in one data from a set of data including one or more attribute names;
The construction means includes, from the list of appearance patterns, a set of attribute names each including one or more of the attribute names in the tree structure and appearing once on a route from a root node to a leaf node. Is created, and at least one grouping target attribute name that is commonly included in one or more node groups having a common parent node is extracted from the patterning tree, and the grouping target attribute name is extracted. An information visualization method characterized by performing a recombination process that bundles a single node.   The program for making a computer run as each means of the information visualization apparatus of any one of Claims 1-7.

Images