JP2001202383A - Device and method for mining data and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for mining data and a recording medium for extracting time correlation rules except for rules with no dependent relation because of a limitation based on the time of transmission between spots by extracting the pattern of a label value to highly frequently appear from the time sequential data of label values acquired at plural spots of a space while considering time for an influence to propagate between spots. SOLUTION: A calculation means 3 calculates a respective distance between respective spots on the basis of spatial information held in a map information holding means and calculates the time of medium transmission between spots by dividing the distance between the spots with a medium transmission velocity. Then, a high frequency label value set is extracted from the time sequential data of label values held in a label value sequence holding means 8 by an inter- label value limitation deciding means 7 and label value pair generating means 9 to rule extracting means 15 while considering the time of medium transmission between spots and on the basis of the high frequency label value set, the rules of if-then form are generated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data mining apparatus, a data mining method, and a recording medium for detecting a pattern appearing in a database at a high frequency and extracting an association rule. The present invention relates to a data mining apparatus, a data mining method, and a recording medium suitable for extracting a correlation rule from a database that records a time series of label values for one or a plurality of measurement elements.

[0002]

2. Description of the Related Art As one method of data mining, which is a technique for finding knowledge from a large-scale database, there is extraction of an association rule. The association rules are extracted from a database or the like in which each record is configured by a set of items. The association rule is a rule of the form “A → B” indicating that “when A occurs, B often occurs”.

[0003] In the first conventional data mining method,
For example, "Fast Algorithms for Mining Association Rul
es "(Agrawall et al., Proc. of the 20th VLDB C
onference, Santiago, Chile, 1994)
-Apriori described in Japanese Patent No. 287106
Based on the law, the correlation rules between the item sets in the database are extracted from the plurality of records recorded in the database. In the apriori method, the percentage of records that include the item set “A, B,..., X, Y” out of all records is used as the support level, and the rules “A, B,.
Item set “A, B,..., X” in the condition part of “X → Y”
Is defined as a certainty factor, and when the support and the certainty factor are higher than predetermined reference values, respectively, the rules “A, B,... ., X → Y ”is extracted as an association rule.

In the second conventional data mining method, for example, as described in Japanese Patent Application Laid-Open No. 9-34719, a time-series data is symbolized and then rules are extracted to convert the time-series data into an if-then format. Are extracted. In the second conventional data mining method, a user prepares a symbolization dictionary in which correspondence between time-series data and symbol values is stored in advance. Then, the time-series data is collated with the symbolization dictionary, and one or a plurality of time-series data is converted into a set of symbol values to which time is added. Based on the symbol value sequence, an if-then format rule that uses a combination of symbol values as a condition part is extracted.

[0005] Furthermore, in the third conventional data mining method, for example, "Searching for Structure in Multipl
e Streams of Data "(Oates et al., Proc. of the
13th Int'l Conference on Machine Learning, 199
6 years) MSDD (Multi-Stream Depen)
According to the dency detection method, temporal dependency between label value patterns is extracted from a plurality of time-series data composed of categorical data (hereinafter, referred to as “label values”). In the MSDD method, a rule of the form "a pattern y occurs at time t + d with a probability p if a pattern x occurs at time t" is obtained. FIG. 21 shows MS
FIG. 2 is a diagram showing an example of data to be subjected to the DD method, and FIG.
FIG. 2 is a diagram showing an example of a rule extracted from the data shown in FIG. 21 by the MSDD method. The data shown in FIG. 21 is data of three discrete time series.
By processing according to the D method, for example, the rules shown in FIG. 22 are extracted. The rule shown in FIG. 22 is that “a label value A occurs in time series 1 at time t, a label value 7 occurs in time series 3, and a label value W occurs in time series 2 at time t + 1.
Occurs, it is expected that at time t + 3, a label value W occurs in the time series 2 and a label value 3 occurs in the time series 3 ”with high probability.

[0006]

Since the conventional data mining method is configured as described above, the first conventional data mining method does not consider dealing with the temporal change of the attribute value. There is a problem that it is difficult to obtain a temporal correlation rule from the time-series data of the label value. In the second and third conventional data mining methods, the time-series data is obtained at different points in space. When the transmission time at which data at each point affects data at other points cannot be ignored, it is difficult to extract rules taking into account the dependence of such time-series data. There is a problem that even rules without dependencies may be extracted due to time constraints.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and is based on time-series data of label values acquired at a plurality of points in space, taking into account the time required for transmitting influence between points. Data mining apparatus, data mining method, and recording for extracting patterns of label values appearing at a frequency and extracting temporal correlation rules excluding rules that have no dependency due to restrictions due to transmission time between points The purpose is to obtain a medium.

[0008]

A data mining apparatus according to the present invention is based on a correspondence between a range of a medium transmission speed which is a speed at which a medium moves in a space and a representative medium transmission speed representing each range. Medium speed classification means for converting each medium transmission speed constituting a medium transmission speed sequence which is a time series of medium transmission speeds into a representative medium transmission speed to generate a representative medium transmission speed sequence; and spatial information of each point. And the representative medium transmission speed, with the two points of each point pair as the first point and the second point, the time that the medium moves from the first point to the second point, and the first point from the second point. Calculating means for calculating the time required for the medium to move to the point as a medium-to-point medium transmission time; 1 The label value on the temporally front side of the two label value set consisting of Whether the time difference between the current label value and the current time of the rear label value is equal to or greater than the inter-point medium transmission time from the current position of the previous label value to the current position of the rear label value at the current time of the previous label value Label-value constraint determining means for determining whether or not each of the high-frequency label value candidates appears in a label value series which is a time series of label values relating to one or more elements depending on a medium; A label value set verification unit that counts the appearance frequency of each high-frequency label value set candidate in the label value series only when it is determined that the constraint condition is satisfied by the inter-value constraint determination unit; A high-frequency label value set extracting means for setting the high-frequency label value candidates or the high-frequency label value set candidates to a high-frequency label value or a high-frequency label value set, respectively; And the time difference between the present time of the first label value and the present time of the second label value in the label value set combined as the second label value is the first label at the present time of the first label value. A high-frequency label value set candidate that is longer than the inter-point medium transmission time between the point of appearance of the value and the point of appearance of the second label value and less than or equal to the predetermined maximum transmission time, and which further appears in the label value series. And a high-frequency label value set candidate generating means for generating a new high-frequency label value set candidate by combining a plurality of different high-frequency label value sets.

[0009] The data mining apparatus according to the present invention comprises:
Of the rule candidates with the temporally last label value as the conclusion part and the remaining label values as the condition parts among the label values that make up the frequent label value set, the frequency of appearance of the condition part in the label value series and its high A rule extracting means for extracting, as a rule, a rule whose ratio with the frequency of appearance of the frequency label value set is within a predetermined range is provided.

[0010] The data mining apparatus according to the present invention comprises:
Distance calculating means for calculating a distance between the first point and the second point based on spatial information of the point as a point-to-point distance; and a half-line passing through the second point from the first point. Angle difference calculation means for calculating the angle between the direction of the representative medium transmission speed as the angle difference, and the point-to-point distance calculated by the distance calculation means, the representative medium transmission speed, and the angle difference calculated by the angle difference calculation means. Medium transmission time calculating means for calculating the point-to-point medium transmission time on the basis of the calculation means.

[0011] The data mining apparatus according to the present invention comprises:
The distance calculating means calculates the distance between the points from the latitude and longitude of each point as spatial information.

[0012] The data mining apparatus according to the present invention comprises:
The distance calculating means calculates the distance between the points from the latitude, longitude and altitude of each point as spatial information.

[0013] The data mining apparatus according to the present invention comprises:
Whether the medium moves from the first point to the second point according to whether the angle difference calculated by the angle difference calculating means is within a predetermined range, and whether the medium moves from the second point to the first point. Angle difference determining means for determining whether or not the medium moves to the point, and the point-to-point distance calculated by the distance calculating means when the medium is determined to move from the first point to the second point. The ratio with the representative medium transmission speed is calculated as the point-to-point medium transmission time from the first point to the second point, and if it is determined that the medium does not move from the first point to the second point, The point-to-point medium transmission time from the first point to the second point is set to a predetermined large value, and when it is determined that the medium moves from the second point to the first point, the point-to-point distance and the representative medium transmission speed Is measured as the point-to-point medium transmission time from the second point to the first point. If it is determined that the medium does not move from the second point to the first point, the distance-to-medium speed is set to a predetermined large value for the point-to-point medium transmission time from the second point to the first point. And a ratio calculating means in the medium transmission time calculating means.

[0014] The data mining apparatus according to the present invention comprises:
A distance direction speed calculating means for calculating, as a distance direction speed, a component of a representative medium transmission speed in a direction of a straight line connecting the first point and the second point, and a half passing through the second point starting from the first point. When the direction of the straight line and the direction of the distance direction match, the ratio of the distance between the points and the magnitude of the distance direction speed is calculated as the point-to-point medium transmission time from the first point to the second point, and When the direction of the straight line does not match the direction of the distance direction speed, the point-to-point medium transmission time from the first point to the second point is set to a predetermined large value, and the first point is set with the second point as a starting point.
When the direction of the semi-straight line passing through the point and the direction of the distance direction velocity coincide, the ratio of the distance between the points and the magnitude of the distance direction velocity is determined by the medium transmission time between points from the second point to the first point. When the direction of the semi-straight line does not match the direction of the distance direction speed, the distance-to-medium speed ratio calculation that sets the point-to-point medium transmission time from the second point to the first point to a predetermined large value Means in the medium transmission time calculating means.

[0015] The data mining apparatus according to the present invention comprises:
A point-to-point representative speed sequence calculating means for calculating a point-to-point representative speed sequence from two medium transmission speed sequences respectively corresponding to two points of each point pair when the medium transmission speed sequence is obtained for each point; The medium speed classification means uses a point-to-point representative speed sequence instead of the medium transmission speed sequence.

[0016] The data mining apparatus according to the present invention comprises:
The medium transmission speed at each time point constituting the medium transmission speed sequence is converted into an average of the medium transmission speed and the medium transmission speed at one or more time points near the medium transmission speed, and the medium transmission speed sequence is smoothed. And a medium velocity class smoothing means for converting the medium velocity series to the medium velocity series. The medium velocity classification means uses the smoothed medium transmission velocity series converted by the medium velocity series smoothing means instead of the medium transmission velocity series.

[0017] The data mining apparatus according to the present invention comprises:
A label value subsequence extracting means for extracting a subsequence included in a predetermined time width from the label value series as a label value subsequence, and a method in which the label value at the first time point of the label value subsequence is a high frequency label value. A high-frequency label value detecting unit that uses the label value as a first high-frequency label value;
Is extracted as a transmission time constraint satisfaction point when the point-to-point medium transmission time is less than or equal to a predetermined maximum transmission time when the point of appearance of the first high frequency label value at the present time is the origin. And a label value relating to the transmission time constraint satisfying point in the label value subsequence, the label value from the first time point of the label value subsequence to the interpoint medium transmission time corresponding to that point, and A label value that is a frequency label value is set as a second high frequency label value, and a label value pair extraction unit that generates a high frequency label value set candidate with the first high frequency label value and the second high frequency label value is labeled. This is provided in the value pair generation means.

[0018] The data mining apparatus according to the present invention comprises:
The label value pair generating means includes an element designation extracting means for extracting an element corresponding to an element designated in advance from the label value partial series by the label value partial series extracting means, and the label value pair extracting means comprises a label value partial sequence extracting means. Are used instead of the label value subsequences described above.

The data mining device according to the present invention comprises:
The point extracting means detects a point including an appearance point of the first high-frequency label value among the predetermined non-correlation point pairs, and transmits a point of the non-correlation point pair that is not the appearance point of the first high-frequency label value. It is deleted from the time constraint satisfaction point.

[0020] The data mining method according to the present invention comprises:
Based on the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed that represents each range, each of the media transmission speed series, which is a time series of the medium transmission speed, is formed. Converting the medium transmission speed to a representative medium transmission speed to generate a representative medium transmission speed sequence;
Based on the spatial information of the two points and each representative medium transmission speed, the time taken for the medium to move from the first point to the second point and the second point are defined as the two points as a first point and a second point. Calculating the time required for the medium to move from the point to the first point as a point-to-point medium transmission time; and calculating each time in a label value series that is a time series of label values for one or more elements depending on the medium. Counting the frequency of appearance of the frequency label value candidates; setting the high frequency label value candidates whose appearance frequency is equal to or higher than a predetermined lower limit value to the high frequency label value;
Of the label value pair combined as the first label value and the second label value, the time difference between the present time of the first label value and the present time of the second label value is the second label value. Labels which are longer than the point-to-point medium transmission time between the point of appearance of the first label value and the point of appearance of the second label value and less than or equal to the predetermined maximum transmission time, and which appear in the label value series are labeled as high-frequency labels. A value set candidate, and a temporally two label value set composed of a temporally last label value and one of the remaining label values of the label values constituting the frequent label value set candidate The label value on the front side is used as the front label value and the label value on the back side is used as the rear label value. Output time of front label value Determining whether the constraint condition that is the front label value occurrence point point-to-point medium transmission time over the emergence point of the rear label value from,
Counting the occurrence frequency of each high-frequency label value set candidate in the label value series only when it is determined that the constraint condition is satisfied; The method includes a step of setting a high-frequency label value set and a step of generating a new high-frequency label value set candidate by combining a plurality of different high-frequency label value sets.

[0021] The recording medium according to the present invention is based on the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed representing each of the ranges.
Medium speed classification means for converting each medium transmission speed constituting a medium transmission speed sequence which is a time series of medium transmission speeds into a representative medium transmission speed to generate a representative medium transmission speed sequence, spatial information of each point, and a representative medium Based on the transmission speed, the time at which the medium travels from the first point to the second point and the medium from the second point to the first point, with the two points of each point pair as a first point and a second point Calculating means for calculating the travel time of each as a medium-to-point medium transmission time, comprising a temporally last label value of label values constituting a high-frequency label value set candidate and one of the remaining label values The label value on the temporally front side of the two label value set to be used is the front label value, and the label value on the temporal side is the rear label value. Label value A label that determines whether or not the time difference from the present time satisfies the constraint condition that the media transfer time from the point of appearance of the front label value to the point of appearance of the rear label value is equal to or longer than the point-to-point medium transmission time at the present time of the preceding label value The inter-value constraint determining means counts the frequency of occurrence of each high-frequency label value candidate in a label value series that is a time series of label values relating to one or more elements depending on the medium, and restricts by the inter-label value constraint determining means. A label value set verification unit that counts the appearance frequency of each high frequency label value set candidate in the label value series only when it is determined that the condition is satisfied, a high frequency label value candidate whose appearance frequency is equal to or more than a predetermined lower limit, or A high-frequency label value set extracting unit that sets the high-frequency label value set candidates as high-frequency label values or high-frequency label value sets, respectively, and the two high-frequency label values as a first label value and a second label value; The time difference between the present time of the first label value and the present time of the second label value in the label value set combined as the first label value is the appearance point of the first label value at the present time of the first label value A label value that is longer than the point-to-point medium transmission time between the second label value and the point of appearance of the second label value and less than or equal to a predetermined maximum transmission time, and that appears in the label value sequence as a high-frequency label value set candidate A pair generating means and a program for causing a computer to function as a high frequency label value set candidate generating means for generating a new high frequency label value set candidate by combining a plurality of different high frequency label value sets. .

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Embodiment 1 FIG. FIG. 1 is a block diagram showing a configuration of a data mining device according to Embodiment 1 of the present invention. In the figure, reference numeral 1 denotes map information holding means for holding spatial information of each point, and 2 denotes a medium transmission speed range which is a speed at which the medium moves in space and a medium transmission speed which represents each range. Correspondence holding means for holding the correspondence with the medium transmission speed. In the case of weather data, for example, air is the medium, the wind speed corresponds to the medium transmission speed, the humidity and the like are the attributes (elements) of the air at each point, and change with the movement of the air.

3 is based on the spatial information of the points and the representative medium transmission speed, and for each representative medium transmission speed, the two points of each pair of points are defined as a first point and a second point.
Time the medium travels from the point to the second point and the second
Calculating means for calculating the time required for the medium to move from the point to the first point as the point-to-point medium transmission time.

In the calculating means 3, an angle difference calculating means 21 calculates an angle formed between a direction of each representative medium transmission speed and a half line passing through the second point from the first point as a starting point, 22 is the first based on the spatial information of the point
23 is a distance calculating means for calculating the distance between the second point and the second point as a point-to-point distance, and 23 is a point-to-point distance calculated by the distance calculating means 22, a representative medium transmission speed, and an angle difference calculating means 21. Is a medium transmission time calculation unit that calculates the point-to-point medium transmission time based on the angle difference calculated by In the medium transmission time calculating means 23, it is determined whether the medium moves from the first point to the second point depending on whether the angle difference calculated by the angle difference calculating means 21 is within a predetermined range. No, and angle difference determination means for determining whether the medium moves from the second point to the first point, and 25 is determined to move the medium from the first point to the second point. In this case, the ratio between the point-to-point distance calculated by the distance calculating means 22 and the representative medium transmission speed is calculated as the point-to-point medium transmission time from the first point to the second point,
When it is determined that the medium does not move from the first point to the second point, the point-to-point medium transmission time from the first point to the second point is set to a predetermined large value, and the second point is transmitted from the second point to the second point. When it is determined that the medium moves to the first point, the ratio between the point-to-point distance and the representative medium transmission speed is calculated as the point-to-point medium transmission time from the second point to the first point, and the second point is calculated. If it is determined that the medium does not move from the second point to the first point, the distance-to-medium speed ratio calculation means sets the point-to-point medium transmission time from the second point to the first point to a predetermined large value. The predetermined large value is a predetermined value that is larger than a value that can be actually taken as the point-to-point medium transmission time.

Reference numeral 4 denotes a medium transmission time holding unit for holding the point-to-point medium transmission time calculated by the calculation unit 3;
Reference numeral 6 denotes a medium transmission speed sequence holding unit that holds a medium transmission speed sequence that is a time series of medium transmission speeds. Reference numeral 6 denotes each of the media transmission speed sequences that form the medium transmission speed sequence based on the correspondence held by the correspondence holding unit 2. This is a medium speed classification unit that converts a medium transmission speed into a representative medium transmission speed and generates a representative medium transmission speed sequence.

Reference numeral 7 denotes a temporally preceding label of a two-label value set composed of the temporally last label value of the label values constituting the high-frequency label value candidate and one of the remaining label values. Using the label value as the front label value and the temporally rear label value as the rear label value, for all front label values, the time difference between the current time of the front label value and the current time of the rear label value is the front label value. Is an inter-label value constraint judging unit for judging whether or not a constraint condition that the inter-point medium transmission time from the appearance point of the front label value to the appearance point of the rear label value at the present time is equal to or more than the time is satisfied.

Reference numeral 8 denotes label value series holding means for holding a label value series which is time-series data of one or a plurality of elements depending on a medium. The label value is so-called categorical data, which is data that takes any one of a predetermined finite number of values, and a label value series is, specifically, at a plurality of points in space. It is a time series of weather data such as measured temperature and humidity. Since the label value depends on the medium, it changes based on the transmission of the medium. In addition, it is considered that the label value at each point is affected not only by the movement of the medium but also by the past label value at that point. In the case of the point A and the point B different from each other, the influence of the label value at the point A at a certain point on the label value at the point B exists after the time when the medium moves from the point A to the point B has elapsed.

Reference numeral 9 denotes a label value set in which two high-frequency label values are combined as a first label value and a second label value, and the first label value and the second label value. Is greater than or equal to the point-to-point medium transmission time between the point of appearance of the first label value and the point of appearance of the second label value at the present time of the first label value, and is equal to or less than the predetermined maximum transmission time. And label value pair generation means that sets a label that appears in the label value series as a high-frequency label value set candidate.

In the label value pair generation means 9, reference numeral 31 denotes label value subsequence extracting means for extracting a subsequence included in a predetermined time width from the label value series as a label value subsequence, and 32 denotes a label value subsequence. Is a high-frequency label value detecting means for setting the label value as a first high-frequency label value when the label value at the first point in time is a high-frequency label value. Point extraction means for extracting, as a transmission time constraint satisfaction point, a point whose medium-to-point medium transmission time is less than or equal to a predetermined maximum transmission time when the point of appearance of the first high-frequency label value with respect to the representative medium transmission speed is set as a starting point And 34 is an inter-point medium corresponding to the point from the first time point of the label value subsequence among the label values related to the transmission time constraint satisfaction points in the label value subsequence. A label value that is a label value after the transmission time and that is a high-frequency label value is defined as a second high-frequency label value, and a high-frequency label value set candidate is determined by the first high-frequency label value and the second high-frequency label value. And label value pair extraction means for counting the appearance frequency of the high frequency label value set candidates.

Numeral 10 counts the frequency of occurrence of each high-frequency label value candidate in the label value series, and only when the label value constraint determining means 7 determines that the constraint condition is satisfied, each high-frequency label value candidate in the label value series is counted. This is a label value set verification unit that counts the appearance frequency of label value set candidates. Note that the appearance frequency of the high-frequency label value set candidate having a length of 2 is counted by the label value pair generation unit 9, but the label value pair verification unit 10
May be counted. Reference numeral 11 denotes a high-frequency label value candidate or a high-frequency label value set candidate, and the appearance frequency of the high-frequency label value candidate or the high-frequency label value set candidate counted by the label value pair generation unit 9 or the label value pair verification unit 10. 12 is a high-frequency label value set candidate management unit, and 12 is a high-frequency label value set or a high-frequency label value set candidate whose appearance frequency is equal to or higher than a predetermined lower limit, respectively. Is a high-frequency label value set extracting means.

Numeral 13 denotes a high-frequency label value set management means for holding the high-frequency label values or the high-frequency label value sets extracted by the high-frequency label value set extraction means 12 together with their appearance frequencies. Label value set generation means (high frequency label value set candidate generation means) for generating new high frequency label value set candidates by combining frequency label value sets, and 15 is held in the high frequency label value set management means 13 This is a rule extracting means for extracting a rule from a high frequency label value set.

Next, the operation will be described. FIG. 2 is a diagram illustrating an example of points A, B, and C, FIG. 3 is a diagram illustrating an example of a medium transmission speed sequence, and FIG. 4 is an example of a label value sequence obtained at points A, B, and C FIG. 5 is a diagram showing
6 is a diagram illustrating an example of a correspondence relationship between each range of the medium transmission speed and the representative medium transmission speed, FIG. 6 is a diagram illustrating an example of a representative medium transmission speed series, and FIG. FIG. 4 is a diagram illustrating an example of an angle between the direction of a medium transmission speed and a direction; The map information holding means 1 holds spatial information such as, for example, latitude and longitude of each point such as three points A, B, and C shown in FIG. With respect to the points A, B, and C in FIG.

The medium transmission speed sequence holding means 5 holds a medium transmission speed sequence, for example, as shown in FIG. The medium transmission speed is represented by its size and direction, and the medium transmission speed sequence is a series of medium transmission speeds sampled at predetermined unit times. In FIG. 3, the unit time, that is, the sampling interval is 600 seconds (= 10 minutes). In the medium transmission speed series shown in FIG.
It is shown that the medium is moving in the direction of 0 degrees, the moving speed is 3 meters per second, and thereafter, the moving direction and the size of the medium are sequentially changed.

The label value series holding means 8 holds, for example, as shown in FIG. 4, a label value series concerning one or a plurality of elements obtained at each point held by the map information holding means 1. In FIG. 4, A, B, and C represent points in FIG. 2, and a label value sequence for one element is shown for each point. In the case of weather data, for example, values of measurement elements such as “temperature” and “humidity” are label values. Temperature and humidity are usually continuous values,
Some boundary values are set, continuous values are divided into several regions, and values such as temperature and humidity are converted into label values. Note that each label value depends on the medium.

That is, in the case of temperature, for example, a label value p is defined as less than -10 degrees and a label value q is defined as -10 degrees or more and less than 0 degrees, and a label value r is defined as 0 degrees or more and less than 10 degrees.
By setting the label value s to a label value s from 20 degrees to less than 20 degrees as a label value t and a label value u to 30 degrees or more, a time series of temperature as a continuous value is represented by a label value p,
It is converted into a label value series of q, r, s, t, u.

In FIG. 4, at three points A, B, and C,
A label value of a predetermined element measured at predetermined sampling intervals starting from a certain time point is shown. In this case,
The label values that this element can take are p, q, r, s, t, u
Is one of The label value series at the point A is p, q, r, p, u, r, t, q, s. In addition,
In the following, when it is necessary to distinguish the points, the label value is represented in a format in which the point names are linked. For example, the label value p that has appeared at the point A is expressed as “A: p”. 3 and 4
Are the same. For example, the direction and magnitude of the medium transmission speed at the first time point are 10 degrees and 3 meters per second, respectively, and the label values of predetermined elements at points A, B, and C are p, u, and q, respectively. .

The correspondence holding means 2 holds the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed representing each of the ranges. The correspondence between each range of the medium transmission speed and the representative medium transmission speed is, for example, as shown in FIGS. It is composed of a correspondence relationship and a correspondence relationship between each range of the direction of the medium transmission speed and the direction of the representative medium transmission speed.

In FIG. 5A, the magnitude of the medium transmission speed, which is a continuous value, is classified into five ranges. In FIG. 5B,
The directions of the medium transmission speed are classified into eight ranges. The classification in FIG. 5B is based on a predetermined boundary (dashed line in FIG. 5C) as shown in FIG. 5C, and the direction of the medium transmission speed is east, northeast, north, northwest, west, It is divided into eight ranges: southwest, south, and southeast. Therefore, according to the correspondence relationships shown in FIGS. 5A and 5B, the medium transmission speed is associated with one of the 33 types of representative medium transmission speeds. However, the representative medium transmission speeds of size 0 having different directions are the same.

The medium speed classifying means 6 reads out the medium transmission speed series held in the medium transmission speed series holding means 5,
Based on the correspondence held in the correspondence holding means 2, each medium transmission speed constituting the medium transmission speed sequence is converted into a representative medium transmission speed, and the medium transmission speed sequence is represented by a representative medium transmission speed sequence. Convert to transmission speed series. FIG.
Based on the correspondence shown in FIGS. 3A and 3B, for example, FIG.
When the medium transmission speed series shown in FIG. 6 is converted, a representative medium transmission speed series shown in FIG. 6 is obtained.

Next, the calculation means 3 comprises the map information holding means 1
Of the points stored in the first and second points are stored in the correspondence storing unit 2 based on the spatial information of the first and second points and the representative medium transmission speed. For each representative medium transmission speed appearing in the obtained correspondence, the point-to-point medium transmission time from the first point to the second point and the point-to-point medium transmission time from the second point to the first point are calculated.

At this time, the distance calculating means 22 of the calculating means 3
Calculates the distance between points based on the spatial information (latitude, longitude, etc. of each point) of each point held in the map information holding means 1. When the spatial information is composed of latitudes and longitudes, the distance calculation unit 22 calculates a distance along the ground surface between each point when the earth is approximated by a sphere as a point-to-point distance. When the spatial information includes latitude, longitude, and altitude, the distance calculation unit 22 calculates, for example, the distance between line segments connecting the points as the point-to-point distance. When calculating the distance, the weighting may be changed between the altitude direction and the ground surface direction. In that case, for example, instead of the actual altitude, a value obtained by multiplying the actual altitude by a predetermined constant may be used.

Further, the angle difference calculating means 21 determines each of the points appearing in the correspondence held in the correspondence holding means 2 based on the spatial information of each point held in the map information holding means 1 and the representative medium transmission speed. With respect to the representative medium transmission speed, an angle formed by an angle between a half-line passing through the second point starting from the first point and the direction of the representative medium transmission speed with the two points as a first point and a second point. calculate. Note that the angle difference is measured counterclockwise from the direction of the representative medium transmission speed toward the half-line, and takes a value of 0 degree or more and less than 360 degrees. For example, FIGS. 7A and 7B show the angle difference θ when the point P is the first point and the point Q is the second point.

The medium transmission time calculation means 23 calculates the distance between the points calculated by the distance calculation means 22, the representative medium transmission speed, and the angle difference calculated by the angle difference calculation means 21 for each point pair. Calculate the point-to-point medium transfer time.

At this time, the angle difference determination means 24 of the medium transmission time calculation means 23 determines whether the angle difference calculated by the angle difference calculation means 21 is within a predetermined range,
Generating point-to-point medium movement information that is information about whether the medium moves from the first point to the second point, and whether the medium moves from the second point to the first point; It is supplied to the distance-to-medium speed ratio calculating means 25.

If the angle between the half-line passing from one point P to the other point Q and the direction of the representative medium transmission speed is smaller than a predetermined angle, the point P moves from one point P to the other point Q. It is determined that the medium moves. That is, assuming that a predetermined angle (that is, a threshold value of the angle) is θth degree,
The angle difference is in the range of 0 degree to θth degree or (360−θt
h) If it is in the range of degrees to 360 degrees, it is determined that the medium moves from the first point to the second point. On the other hand, when the angle difference is in the range of (180−θth) degrees to (180 + θth) degrees, it is determined that the medium moves from the second point to the first point.

The distance-to-medium speed ratio calculating means 25 refers to the point-to-point medium movement information from the angle difference judging means 24, and
When it is determined that the medium moves from the first point to the second point, the distance between the points calculated by the distance calculating means 22 is divided by the magnitude of the representative medium transmission speed, and the distance from the first point to the second point is calculated. Calculate the point-to-point medium transmission time to the point. on the other hand,
When it is determined that the medium does not move from the first point to the second point, the point-to-point medium transmission time from the first point to the second point is set to a predetermined large value. Similarly, when it is determined that the medium moves from the second point to the first point, the distance-to-medium speed ratio calculating means 25 divides the distance between the points by the magnitude of the representative medium transmission speed. Then, the point-to-point medium transmission time from the second point to the first point is calculated. On the other hand, if it is determined that the medium does not move from the second point to the first point, the point-to-point medium transmission time from the second point to the first point is set to a predetermined large value.

The point-to-point medium transmission time by the distance-to-medium speed ratio calculation means 25 is supplied to the medium transmission time holding means 4,
Will be retained.

Here, for example, for points A, B, and C in FIG. 2, the distance calculation means 22 calculates that the distance between AB, the distance between AC and the distance between BC are 1073 meters, respectively.
Calculated as 2520 meters and 2280 meters,
The operation of each section of the calculation means 3 when the angle threshold θth in the angle difference determination means 24 is 30 degrees will be specifically described. In addition, from the first time point in FIG.
As in the period up to the third time point, the direction of the representative medium transmission speed at this time is 0 degrees, that is, the direction of east, and its magnitude is 4 meters per second.

First, the processing for points A and B will be described. Assuming that the point A is a first point and the point B is a second point, the angle difference calculation means 21 measures the angle between the direction of the representative medium transmission speed and the half-line AB in a counterclockwise direction, and calculates 270 as the angle difference. Output degrees. Next, the angle difference determination means 24 determines whether or not the angle difference is within a predetermined range, and outputs the point-to-point medium movement information as a result of the determination. If the angle difference is 270 degrees, 330 degrees to 3 also in the range of 0 degrees to 30 degrees
It is determined that there is no movement of the medium from point A to point B because it is not in the range of 60 degrees, and there is no movement of the medium from point B to point A because the angle difference is not in the range of 150 to 210 degrees. Is determined. In this case, since the medium does not move between the points A and B, the distance-to-medium velocity ratio calculating means 25 calculates the medium-to-point medium transmission time from the point A to the point B and the medium between the points B to the point A. The transmission time is set to a predetermined large value.

Next, the processing for the points B and C will be described. Assuming that point B is a first point and point C is a second point, the angle difference calculated by the angle difference calculating means 21 is 0.
Degree. When the angle difference is 0 degree, the angle difference is in the range of 0 degree to 30 degrees, so that the angle difference determination unit 24 determines that the medium has moved from the point B to the point C, and the angle difference is 150 degrees to Since it is not within the range of 210 degrees, it is determined that the medium has not moved from the point C to the point B. Then, the distance between the points BC (=
2280 meters) is divided by the representative medium transfer speed (= 4 meters per second), and the point-to-point medium transfer time from point B to point C is calculated to be 570 seconds. The point-to-point medium transmission time from point C to point B is set to a predetermined large value.

Finally, the processing for points A and C will be described. Assuming that the point A is the first point and the point C is the second point, the angle difference calculated by the angle difference calculating means 21 in the same manner is about 334.7 degrees. Angle difference is about 33
In the case of 4.7 degrees, since the angle is in the range of 330 degrees to 360 degrees, the angle difference determination means 24 determines that the medium has moved from the point A to the point C, and the angle difference is 150 degrees to 210 degrees. It is determined that there is no movement of the medium from point C to point A because it is not within the range of degrees. Then, the distance-to-medium speed ratio calculating means 25 calculates the distance between the points AC (= 2520).
Meters) is divided by the medium transmission speed (= 4 meters per second), and the point-to-point medium transmission time from point A to point C is 63
It is calculated as 0 seconds. The point-to-point medium transmission time from point C to point A is set to a predetermined large value.

As described above, when the direction of the representative medium transmission speed is 0 degrees (the direction of east) and the size is 4 meters per second, from the point A to the point B, from the point B to the point A
, The point-to-point medium transmission time from point C to point B, and from point C to point A are all predetermined large values,
The point-to-point medium transmission time from point A to point C and the point-to-point medium transmission time from point B to point C are each 630.
Seconds and 570 seconds. Note that the point-to-point medium transmission time between the same points is 0 second.

When the direction of the representative medium transmission speed is 270 degrees (in the direction of true south) and the size is 4 meters per second, from point B to point A, from point A to point C,
The point-to-point medium transmission time from point C to point A, point B to point C, and point C to point B are all predetermined large values, and the point-to-point medium transmission time from point A to point B is approximately 268.3 seconds. The point-to-point medium transmission time is similarly calculated for other representative medium transmission speeds.

Next, the inter-point value transmission time is considered from the label value series held by the label value series holding means 8 by the label value constraint determining means 7 and the label value pair generating means 9 to the rule extracting means 15. Then, a high-frequency label value set is extracted, and a rule in the if-then format is extracted based on the high-frequency label value set.

Note that a label value set is a pattern composed of a plurality of label values each given a time difference with respect to a certain point in time, and a frequent label value set is a time series of label values. A label value set whose frequency of occurrence in data is equal to or higher than a predetermined lower limit value. A label value whose frequency of occurrence in the time-series data of the label value is equal to or higher than a predetermined lower limit value is referred to as a high-frequency label value. For the sake of simplicity, the “label value” may be treated as a label value set composed of one label value.

In the following, a label value set is represented by a point, a label value, and an elapsed time from the reference time. For example, the label value is expressed in a format such as “A: p (0), B: x (+2)”. This represents a "pattern in which the label value x appears at the point B two units after the appearance of the label value p at the point A". Here, it is assumed that label value sets that differ only in reference time are the same. For example, the label value set “A: p (0), B: x (+2)” and the label value set “A: p (−2), B: x (0)” are considered to be the same.

Further, the number of label values forming a label value set is called the length of the label value set, and a set having a high-frequency label value set of length k as an element is represented by L (k).
Further, a label value set that is a candidate for a high-frequency label value set is referred to as a high-frequency label value set candidate, and a set including a high-frequency label value set candidate having a length k as an element is represented by C (k).

First, all the label values included in the label value series held in the label value series holding means 8 are used by the high frequency label value set candidate management means 11 as elements of the set C (1) of the high frequency label value candidates. The label value set verification means 10 searches for each high-frequency label value candidate in the label value series, counts the frequency of occurrence, and the high-frequency label value set extraction means 12 calculates the high-frequency label value candidates from C (1) based on the frequency of occurrence. A set L (1) of frequent label values is extracted and held by the frequent label value set management unit 13. Then, the label value pair generation means 9 takes into account the medium transmission time between each point, and
A set C (2) of frequent label value set candidates is generated from (1), and held by the frequent label value set candidate management unit 11. Then, L (2) is extracted from C (2) based on the appearance frequency by the frequent label value set extracting means 12 and held by the frequent label value set managing means 13. Hereinafter, L
Until (k) becomes empty, label value set generation means 14 generates C (k + 1) from L (k), and high frequency label value set extraction means 12 extracts L (k + 1) from C (k + 1). You. Then, a rule in the if-then format is extracted from the high-frequency label value set by the rule extracting unit 15.

Hereinafter, the processing of each section will be described in detail. First, the frequent label value set candidate management unit 11 converts all the label values included in the label value series held in the label value series holding unit 8 into a set C (1) of frequent label value candidates.
As an element. The label value set verifying means 10 searches each high frequency label value candidate held in the high frequency label value set candidate management means 11 by a label value series, and every time the high frequency label value candidate is detected by the label value series. , Count the number of times. And high frequency label value set candidate management means 11
Updates the appearance frequency of the high-frequency label value candidate accordingly.

Next, the frequent label value set extracting means 12
Based on the high-frequency label value candidates held by the high-frequency label value set candidate management unit 11 and the frequency of appearance, high frequency label value candidates whose appearance frequency is equal to or higher than a predetermined lower limit are extracted as high frequency label values, It is supplied to the high-frequency label value set management means 13 together with the appearance frequency. The frequent label value set management unit 13 holds the frequent label value and its appearance frequency.

Then, the label value pair generating means 9 combines the two high frequency label values in consideration of the representative medium transmission speed and the point-to-point medium transmission time, and sets the high frequency label value group which is an element of C (2). A candidate is generated, and the appearance frequency of each high-frequency label value set candidate in the label value series held in the label value series holding means 8 is counted. Then, the frequent label value set candidate and its appearance frequency are held by the frequent label value set candidate management unit 11.

The operation of each part of the label value pair generation means 9 will now be described. First, the label value subsequence extracting means 31
Extracts a part included in a predetermined time width from the label value series held in the label value series holding means 8 as a label value partial series. This predetermined time width is the same as the maximum transmission time.

Next, the high-frequency label value detecting means 32 outputs the label value subsequence extracted by the label value subsequence extracting means 31 and the set L of the high frequency label values held in the high frequency label value Based on (1), L among the label values at the beginning of the label value
An element that matches the element of (1) is detected as a first high-frequency label value.

The point extracting means 33 determines the first high-frequency label value detected by the high-frequency label value detecting means 32 and the point-to-point medium transmission time held by the medium transmission time holding means 4. The point at which the point-to-point medium transmission time is less than or equal to a predetermined maximum transmission time when the appearance point of the first high-frequency label value is the starting point at the time when the first high-frequency label value is output is defined as the transmission time constraint satisfaction point. Extract,
Label value pair extracting means 3 together with the medium transmission time between points
4

The label value pair extraction means 34 outputs the label value subsequence extracted by the label value subsequence extraction means 31,
The first high-frequency label value detected by the high-frequency label value detection means 32, the transmission time constraint satisfaction point from the point extraction means 33 and the medium transmission time between the points, and the high-frequency label value set management means 13 are retained. Based on the set L (1) of the frequent label values, among the label values relating to the transmission time constraint satisfaction points in the label value subsequence, the distance between the points corresponding to the point is determined based on the first time point of the label value subsequence. The label value after the medium transmission time, and L
The label value, which is an element of (1), is detected as a second high-frequency label value, and the length is calculated based on the first high-frequency label value, the second high-frequency label value, and both points and the time difference between the two. Is generated, and the appearance frequency is counted.

The set C (2) of the frequent label value set candidates of length 2 generated in this way and the appearance frequency of each frequent label value candidate are held by the frequent label value set candidate management means 11. .

Then, the frequent label value set extracting means 12
Based on the frequent label value set candidates held in the frequent label value set candidate management means 11 and their appearance frequencies, the frequent label value set candidates whose appearance frequency is equal to or higher than a predetermined lower limit are defined as the frequent label value sets. It is extracted and supplied to the high-frequency label value set management means 13 together with its appearance frequency.

Next, the label value set generating means 14 generates C (k + 1) from the set L (k) (k> 1) of the frequent label value sets held in the frequent label value set managing means 13. Then, it is supplied to the high-frequency label value set candidate management unit 11.

At this time, the label value set generating means 14 selects the next one of the set L (k) (k> 1) of the frequent label value sets of length k held in the frequent label value set managing means 13. Two high-frequency label value sets satisfying the two conditions are extracted, and the two high-frequency label value sets are combined to obtain a length (k + 1).
Is generated. First condition: The temporally last label values in the label value set are the same.
However, the label values being the same here are the same including the point. Second condition: the label value sets of length (k-2) excluding the temporally last label value and one other label value in the label value set are the same.

It is assumed that two sets of high-frequency label values satisfying the above conditions can be expressed as follows, using the present time of the last label value in time as a reference point. "A (ta), b (tb), ..., c (tc), p
(Tp), x (0) "" a (ta), b (tb), ..., c (tc), q
(Tq), x (0) "where ta, tb,. . . , Tc, tp, and tq are all numerical values of 0 or less.

Then, a high-frequency label value set candidate of length (k + 1) is generated from these high-frequency label value sets as follows. "A (ta), b (tb), ..., c (tc), p
(Tp), q (tq), x (0) "

In this way, the set L of the frequent label value sets
From (k) (k> 1), a set C of frequent label value set candidates
(K + 1) is generated, supplied to the frequent label value set candidate management unit 11, and held.

Then, the set C of the frequent label value set candidates
Each high-frequency label value set candidate in (k + 1) is searched in the label value series, and when a high-frequency label value set candidate is detected, the above-described constraint condition determination is executed by the label value constraint determination unit 7. Only when the constraint condition is satisfied, the appearance frequency is counted. In response, the frequent label value set candidate management unit 11 updates the held appearance frequency.

At this time, the label value constraint determining means 7 is composed of the temporally last label value of the label values constituting the frequent label value set candidate and one of the remaining label values. For all front label values, the label value on the temporally front side of the two label value pairs (label value pair of length 2) is defined as the front label value, and the label value on the temporal rear side is defined as the rear label value. It is said that the time difference between the present time of the value output and the present time of the rear label value is equal to or longer than the point-to-point medium transmission time from the appearance point of the front label value to the appearance point of the rear label value at the present time of the preceding label value. It is determined whether or not the constraint condition is satisfied, and the determination result is used as a label value set verification unit 1
Supply 0.

Here, the operation of the inter-label-value constraint determining means 7 in the case where the time point at which the last label value appears in time is set to 0 and the high-frequency label value set is indicated as follows will be described. "A (ta), b (tb), ..., c (tc), x
(0) "where ta, tb,. . . , Tc are all numerical values of 0 or less.

First, the two-label value set “a (ta), x
(0) ”, the time difference between these label values (=
| Ta |) is calculated based on the representative medium transmission speed at the time when the label value a (ta) is output.
It is determined whether it is longer than the point-to-point medium transmission time from the appearance point of a) to the appearance point of the label value x (0), and similarly, the other two label value sets “b (tb), x (0) ) ", ...
, "C (tc), x (0)", and if the determination results for all two label value pairs are true, the label value constraint determination means 7 satisfies the constraint condition. Is supplied to the label value set verification means 10.

Until L (k) becomes empty, the inter-label-value constraint determining means 7 and the label-value-set verifying means 10 to the label-value-set generating means 14 operate in the same manner, and the longer high-frequency label-value Are sequentially extracted and accumulated and held in the high-frequency label value set management unit 13 in order.

After the process of extracting the high-frequency label value sets has been completed in this way, a rule in the if-then format is generated by the rule extracting means 15 from the high-frequency label value sets of each length. At this time, the rule extracting unit 15 determines the high-frequency label value set held by the high-frequency label value set managing unit 13 and the appearance frequency thereof, and the high-frequency label value series With the temporally last label value in the label value set as the conclusion part and the remaining label values as the condition part, the ratio of the ratio of the occurrence of the condition part to the ratio of the appearance of the entire high-frequency label value is a predetermined lower limit. The pair of the condition part and the conclusion part described above is extracted as a rule in the if-then format.

For example, the frequent label value set “a (ta),
b (tb),. . . , C (tc), x (tx) ”and rule candidates“ a (ta), b (tb),.
c) → x (tx) ”is generated. Where ta, t
b,. . . , Tc are numerical values equal to or less than tx, and “a
(Ta), b (tb),. . . , C (tc) ”is the condition part, and“ x (tx) ”is the conclusion part.

Then, the rule extracting means 15 calculates the frequency of occurrence of the label value set corresponding to the condition part in the label value series, and, based on the calculated frequency, generates a label value set corresponding to the condition part to the conclusion part. The rate at which the corresponding label value appears is calculated, and a rule candidate whose ratio between the rate at which the label value set corresponding to the condition part appears and the rate at which the high-frequency label value set appears is equal to or greater than a predetermined lower limit is defined as a rule. I do. At this time, the condition part whose start point is any one of the number of times equal to the number of times obtained by dividing the time width spanned by the high-frequency label value set corresponding to the condition part and the conclusion part by the unit time, of the last of the label value series Are ignored, and are not included in the appearance frequency. This is for rule candidates,
In order to accurately determine the rate at which the label value corresponding to the conclusion part appears when the label value pair corresponding to the condition part occurs, the label corresponding to the condition part where the appearance of the label value corresponding to the conclusion part is unknown This is because it is necessary to exclude the value set.

When calculating the appearance frequency of the label value set corresponding to the condition part, if the label value set is a high-frequency label value set, the appearance frequency has already been calculated and the high-frequency label value , So you can use it. However, the condition part whose start point is any one of the number of times equal to the number of times obtained by dividing the time width set by the high-frequency label value pair corresponding to the condition part and the conclusion part by the unit time at the end of the label value series Is not included in the appearance frequency.

In this way, rules are extracted from the frequent label value set.

Hereinafter, a process for extracting a high-frequency label value set and a rule from the label value sequence shown in FIG. 4 based on the representative medium speed sequence shown in FIG. 6 will be described.

First, generation of L (1) will be described.
The frequent label value set candidate management unit 11 holds each label value in the label value series shown in FIG. 4 as an element of a set C (1) of frequent label value candidates. Note that the appearance frequency of each high-frequency label value candidate is initialized to zero. The label value set verification means 10 searches the label value series for the frequent label value candidates, counts the number of times each of the frequent label value candidates is detected, and accordingly, the frequent label value set candidate management means 11 Updates the appearance frequency that is held. Therefore, in the case of C (1) relating to the point A in the label value series in FIG. 4, “A: p”, “A: q”, “A: r”, “A:
The appearance frequencies of "s", "A: t" and "A: u" are 2, 2, 2, 1, 1 and 1, respectively.

The high-frequency label value set extracting means 12 extracts high-frequency label value candidates whose appearance frequency is equal to or higher than a predetermined lower limit as elements of the high-frequency label value L (1). When the predetermined lower limit is 2, “A: p”, “A: q”, and “A: r” whose appearance frequency is 2 among the elements of C (1) relating to point A are L (1 ), And are stored in the high-frequency label value set management unit 13 together with the appearance frequency. Similarly, “B: q”, “B: t”, “C:
“p” and “C: r” are held as elements of L (1).

Next, generation of L (2) will be described. FIG. 8 is a diagram showing an example of the label value subsequence extracted by the label value subsequence extracting means 31. The label value subsequence extracting means 31 of the label value pair generating means 9 extracts a part included in a predetermined time width from the label value series held in the label value series holding means 8 as a label value subsequence.
For example, the predetermined time width is 2 unit times (in this case, 1200
Second), the first label value subsequence for the label value sequence of FIG. 4 is as shown in FIG. However,
It is assumed that both the label values at the start time and the end time of two unit times having a predetermined time width are included in the same label value partial sequence.

Then, the label value subsequence extracting means 31
While moving the start time point by one unit time, the corresponding label value subsequences are sequentially extracted, and after extracting the label value subsequence whose start time point is the last time point of the label value series, the operation ends. .

Based on the label value subsequence and the set L (1) of frequent label values, the frequent label value detecting means 32 calculates the element of L (1) among the label values at the first time point of the label value subsequence. Is detected as the first high-frequency label value. In the case of the label value subsequence shown in FIG. 8, "A: p" and "C: q" are extracted as the first high-frequency label values.

Then, the point extracting means 33 uses the point of appearance of the first high-frequency label value in the point-to-point medium transmission time relating to the representative medium transmission speed at the time when the first high-frequency label value is output as the starting point. A point whose value is shorter than the maximum transmission time is extracted as a transmission time constraint satisfaction point, and is supplied to the label value pair extraction means 34 together with the medium transmission time between the points. In this case, the maximum transmission time is 1200 seconds.

In the case of the above-mentioned first high-frequency label value A: p, the present time is the first time point, and as shown in FIG. 6, the direction of the representative medium transmission speed at that time is 0 degree. Yes, its magnitude is 4 meters per second, and the point where the label value was sampled is A. Also, as described above, in this case, the point-to-point medium transmission time from point A to point B and the point-to-point medium transmission time from point A to point C are each a predetermined large value and 630 seconds, as described above. is there. Therefore, of the points B and C, only the point C has the maximum transmission time of 1200 seconds or less when the point A is the starting point, and the point C is extracted as the transmission time constraint satisfaction point. . In addition, since the point A itself is also a transmission time constraint satisfaction point because the inter-point medium transmission time is 0, points A and C are extracted as transmission time constraint satisfaction points for the first high-frequency label value A: p. Is done.

Then, the label value pair extracting means 34 generates the label value subsequence, the first high frequency label value, the transmission time constraint satisfaction point regarding the first high frequency label value and the medium transmission time between the points, and the high frequency label. Based on the value set L (1), a high-frequency label value set candidate having a length of 2 is generated. At this time, among the label values related to the transmission time constraint satisfaction point in the label value subsequence, the label value is the label value after the inter-point medium transmission time corresponding to the point with respect to the first time point of the label value subsequence, and L The label value, which is an element of (1), is defined as a second high-frequency label value, and the first high-frequency label value and the second high-frequency label value, and both points and the time difference between the two, the length of which is 2 A frequency label value set candidate is generated, and the appearance frequency of the high frequency label value set candidate is counted.

For example, the first high-frequency label value A: transmission time constraint satisfaction point C relating to p (medium transmission time between points = 63)
0 second), the label value after 630 seconds from the first point in the label value subsequence is only “C: r”. Since the label value “C: r” is an element of L (1), it is combined with the first high-frequency label value A: p to obtain “A: p (0),
C: r (+2) ”is generated, the appearance frequency of the high-frequency label value set candidate is counted, and the high-frequency label value candidate stored in the high-frequency label value candidate management unit 11 is counted. The appearance frequency of the value set candidate is updated.
Similarly, for example, the transfer time constraint satisfaction point A (the point-to-point medium transfer time = 0) for the first high-frequency label value A: p
Second), the frequent label value set candidate “A: p (0),
A: q (+1) "and the frequent label value set candidate" A: p
(0), A: r (+2) "is generated, the appearance frequency of this high-frequency label value set candidate is counted, and this high-frequency label value set candidate Is updated.

The high-frequency label value set candidate management means 11
If a high-frequency label value set candidate that is not held is detected, an area for storing the high-frequency label value set candidate and the appearance frequency is newly secured and held.

For example, the next 11 high-frequency label value set candidates having the label value relating to the point A as the first high-frequency label value are finally obtained. "A: p (0), A: q (+1)" (frequency of appearance = 1) "A: p (0), A: r (+2)" (frequency of appearance = 2) "A: q (0), “A: p (+2)” (appearance frequency = 1) “A: q (0), A: r (+1)” (appearance frequency = 1) “A: r (0), A: p (+1)” ( Appearance frequency = 1) “A: r (0), A: q (+2)” (appearance frequency = 1) “A: p (0), C: r (+2)” (appearance frequency = 2) “A: q (0), C: p (+2) ”(frequency of appearance = 1)“ A: r (0), C: p (+2) ”(frequency of appearance = 1)“ A: r (0), B: q (+1) "(appearance frequency = 1)" A: r (0), B: t (+2) "(appearance frequency = 1)

For example, a high-frequency label value set candidate is not generated from the label value A: r at the sixth time point and the label value C: q at the eighth time point in FIG. This is because the point-to-point medium transmission time from the point A to the point C regarding the representative medium transmission speed at the sixth time point is set to a predetermined large value. On the other hand, the point-to-point medium transmission time from the point A to the point B regarding the representative medium transmission speed at the sixth time point is about 268.3 seconds, which is shorter than the maximum transmission time.
The high-frequency label value set candidates “A: r (0), B: q (+1)”, “A:” corresponding to the label value A: r at the sixth time point
r (0), B: t (+2) "is generated.

Next, the frequent label value set extracting means 12
Of the generated C (2), a high-frequency label value set candidate whose appearance frequency is equal to or higher than a predetermined lower limit value is extracted as a high-frequency label value set of length 2 (element of L (2)) and It is supplied to the label value set management means 13. Assuming that the predetermined lower limit is 2, as in the case of extracting L (1), for example, a high-frequency label value set candidate “A: p (0 ), A: r (+2) ",
“A: p (0), C: r (+2)” is extracted as an element of L (2), and is stored in the high-frequency label value set management unit 13 together with the appearance frequency. Similarly, “B: q (0),
B: t (+1) "," C: r (0), C: p (+
1), “B: q (0), C: r (+1)”, “B: q
(0), C: p (+2) "and" B: t (0), C:
“p (+1)” is extracted as an element of L (2), and is stored in the high-frequency label value set management unit 13.

Next, generation of L (3) will be described. The label value set generation means 14 removes the temporally last label value and the temporally last label value and one other label value from L (k) (k> 1). Two high-frequency label value sets having the same length (k-2) label value set are extracted and combined to generate a high-frequency label value set candidate of length (k + 1). In the case of L (2) described above, the first
The two frequent label value pairs that satisfy the condition
“A: p (0), C: r (+2)” and “B: q (0),
C: r (+1) ", and since these two high-frequency label value sets are all empty except for two label values from these high-frequency label value sets, the second condition Fulfill.
Therefore, by combining these frequent label value sets, the frequent label value set candidate “A: p (0), B: q
(+1), C: r (+2) "is generated and held in the high-frequency label value set candidate management unit 11. Length 3 in the same way
Of the high-frequency label value set candidates “B: q (0), C: r (+
1), C: p (+2) "," B: t (0), C: r
(0), C: p (+1) "," B: q (0), B: t
(+1), C: p (+2) "is generated as an element of C (3).

The label value set verification means 10 searches the element of C (3) in the label value series and detects it, and if it detects the high frequency label value set candidate which is the element of C (3), the label value constraint determination means 7 Only when it is determined that the above-mentioned constraint condition is satisfied, the appearance frequency of the detected element of C (3) is counted, and the high-frequency label value set candidate management unit 11 accordingly stores the appearance frequency To update.

Here, as an example, a high-frequency label value set candidate “B: q (0), B: t (+1), C: p (+
The determination of “2)” by the label value constraint determination unit 7 will be described.

The frequent label value set candidate “B: q (0),
B: t (+1), C: p (+2) "first appears in FIG. 4 from the second time point to the fourth time point. At this time, the two label value set “B: q (0), C: p
(+2) ", the label value B: q (0) is present at the second time point, and the point-to-point medium transmission time from point B to point C relating to the representative medium transmission speed at that time is as described above. 570 seconds. Therefore, the time difference (= 1200 seconds) between the label value B: q (0) output time and the label value C: p (+2) output time is changed from the point B to the point C relating to the representative medium transmission speed at the second time point. Since the medium transmission time between points is equal to or longer than (570 seconds), this two-label value set satisfies the constraint condition. Similarly, a two-label value set “B: t (+1), C: p (+2)”
Also satisfies the constraint condition. Therefore, the high-frequency label value set candidate “B: q (0),
B: t (+1), C: p (+2) "are determined to satisfy the above-described constraint conditions.

In FIG. 4, the next most frequent label value set candidate “B: q (0), B: t (+1), C: p (+2)” is calculated from the seventh time to the ninth time. Appear. At this time, the two label value set “B: q (0), C: p
(+2) ", the label value B: q (0) is output at the seventh time, and the point-to-point medium transmission time from point B to point C relating to the representative medium transmission speed at that time is as described above. This is a predetermined large value. Therefore, the label value B: q (0) and the label value C: p (+2)
The time difference (= 1200 seconds) from the point B to the point C relating to the representative medium transmission speed at the seventh point in time.
Since this is not longer than the point-to-point medium transmission time (= predetermined large value), the two label value set does not satisfy the constraint. Therefore, the high-frequency label value set candidate “B: q (0), B: t (+1), C: p
(+2) "is determined not to satisfy the above-described constraint condition.

In this way, the appearance frequency of the element of C (3) in the label value series is finally calculated. High frequency label value set candidate “B: q (0), B: t (+1), C: p
(+2) ", the appearance frequency is 1. Then, the frequent label value set extraction unit 12 extracts the frequent label value set candidates whose appearance frequency is equal to or more than the predetermined lower limit value from C (3) to the length 3.
As a high-frequency label value set (element of L (3))
It is supplied to the high-frequency label value set management means 13.

Assuming that the predetermined lower limit is 2, the frequent label value set candidate “B: q (0), B: t (+1), C:
Since the appearance frequency of “p (+2)” is 1, it is not set as a high-frequency label value set of length 3. Further, the other high-frequency label value set candidates “A: p (0), B: q (+1), C: r (+
2) "," B: q (0), C: r (+1), C: p (+
2) "," B: t (0), C: r (0), C: p (+
1) are 2, 2 and 1, respectively, so that the high-frequency label value set candidates “A: p (0), B: q (+
1), C: r (+2) "and the frequent label value set candidate" B: q (0), C: r (+1), C: p (+2) "as the frequent label value set of length 3 It is extracted and stored in the high-frequency label value set management unit 13.

Next, the label value set generating means 14 sets C (4)
Of the element of L (3), "A:
p (0), B: q (+1), C: r (+2) "and" B: q (0), C: r (+1), C: p (+2) ", the temporally last label value Since (C: p and C: r) are not the same, the element of C (4) cannot be generated. Therefore, C (4) becomes empty and L (4) becomes empty, so that the generation of the high-frequency label value set ends up to L (3). In the above description, the lower limit of the frequency of appearance at the time of generating a high-frequency label value set is uniformly set to 2, but the lower limit may be changed according to the length of the high-frequency label value set.

Then, the rule extracting means 15 outputs, for example, the frequent label value set “A: p (0), B: q (+1), C: r
(+2) ”and rule candidates“ A: p (0), B: q (+
1) → C: r (+2) "is generated. The label value set “A: p (0), B: q” corresponding to the condition part of this rule candidate
(+1) "is not a high-frequency label value set, so it is necessary to calculate a new appearance frequency. In this case, the time width spanned by the high-frequency label value sets corresponding to the condition part and the conclusion part is two units starting at the label value “A: p” and ending at the label value “C: r”. Because of the time, the appearance of a label value set whose start point is one of the last two time points of the label value series (that is, the eighth time point and the ninth time point in FIG. 4). Exclude and calculate the appearance frequency. Therefore, the label value set “A: p (0), B:
The appearance frequency of “q (+1)” is calculated as 2.

Next, the rule extracting means 15 obtains the ratio of the ratio of the occurrence of the condition part to the ratio of the entire high-frequency label value. Rule candidate “A: p (0), B: q (+1) →
In the case of "C: r (+2)", the appearance frequency of the label value set corresponding to the condition part is 2, and the appearance frequency of the high-frequency label value set corresponding to the condition part and the conclusion part is also 2 as described above. is there. Therefore, the ratio is 100%.

Then, the rule extracting means 15 determines a rule candidate whose ratio is equal to or more than a predetermined lower limit value if-th.
Extract and output as rules in en format. For example, if the predetermined lower limit is 80%, the rule candidate “A: p
Since the ratio of “(0), B: q (+1) → C: r (+2)” is 100%, it is output as an if-then rule. Similarly, the rule “B: q (0),
C: r (+1) → C: p (+2) ”is also extracted.

As described above, according to the first embodiment, the influence is transmitted between the points based on the medium transmission speed that changes with time from the time-series data of the label values acquired at a plurality of points in the space. Since the pattern of label values that appear at high frequency is extracted in consideration of time, temporal correlation rules are excluded except for rules that are determined to have no dependencies due to restrictions on medium transmission time between points. Can be obtained. In other words, there is an effect that a meaningless pattern having no dependency is not generated due to the restriction due to the medium transmission time between points.
Further, since a meaningless pattern is not generated, an effect is obtained that the processing time for rule extraction can be reduced.

According to the first embodiment, when calculating the distance between the points from the latitude and longitude of the point, if the time required for transmitting the medium in the altitude direction can be neglected, the appropriate point-to-point medium Transmission time can be obtained, and when calculating the distance between points from the latitude, longitude, and altitude of a point, an appropriate point-to-point medium transmission time can be obtained even if the time required for transmission of the medium in the altitude direction cannot be ignored. The effect that it can be obtained is obtained.

Furthermore, according to the first embodiment, the point-to-point medium transmission time is calculated based on the distance between the two points and the angle between the half-line connecting the two points and the direction of the representative medium transmission speed. Therefore, it is possible to obtain an effect that an appropriate point-to-point medium transmission time can be calculated in consideration of the movement of the medium.

Further, according to the first embodiment, it is determined whether or not the angle formed by the half line connecting the two points and the direction of the representative medium transmission speed is within a predetermined range, and the medium between the points is determined. Since the presence / absence of the movement of the point is determined, it is possible to easily and appropriately calculate the medium transmission time between points.

Further, according to the first embodiment, for each label value subsequence, after extracting the first high-frequency label value from the label value at the first time, the second high-frequency label value satisfying the constraint on the transmission time of the medium is obtained. Extract the frequent label values of
Combining the first and second frequent label values to length 2
Since the high-frequency label value set candidate is generated, it is possible to easily generate a high-frequency label value set candidate having a length of 2 that satisfies the restriction on the medium transmission time.

Embodiment 2 FIG. 9 is a block diagram showing a configuration of a data mining device according to Embodiment 2 of the present invention. In the figure, 41 is the representative medium transmission speed,
Distance direction speed calculating means for calculating a component in a direction of a straight line connecting the first point and the second point as a distance direction speed.
The other components in FIG. 9 are the same as those according to the first embodiment, and a description thereof will not be repeated. However, the distance-to-medium speed ratio calculating means 25 uses the distance direction speed calculated by the distance direction speed calculating means 41 instead of the representative medium transmission speed.

Next, the operation will be described. FIG. 10 is a diagram showing the correspondence between the representative medium transmission speed and the distance direction speed. The distance direction speed calculating means 41 is the angle difference calculating means 21.
Based on the angle difference calculated by the above and the correspondence held in the correspondence holding means 2, the component of the direction of the straight line connecting the first point and the second point of each representative medium transmission speed appearing in the correspondence is calculated. Calculate as distance speed.

For example, in FIG. 10, the distance direction velocity is calculated by projecting the vector of the representative medium transmission velocity onto a straight line AB passing through the first and second points A and B. That is, the distance direction velocity is calculated by multiplying the cosine of the angle difference by the magnitude of the representative medium transmission velocity. At this time, when the distance direction speed is positive, the direction of the distance direction speed coincides with the direction of a half-line passing through the second point from the first point. On the other hand, when the distance direction speed is negative, the direction of the distance direction speed coincides with the direction of a half-line passing through the first point from the second point.

The distance to medium speed ratio calculating means 25
Based on the point-to-point distance calculated by the distance calculating means 22 and the distance direction speed calculated by the distance direction speed calculating means 41, the direction of the half-line passing through the second point from the first point and the direction of the distance direction speed If the two coincide, the ratio between the distance between the points and the magnitude of the distance direction velocity is calculated as the point-to-point medium transmission time from the first point to the second point. The point-to-point medium transmission time from the first point to the second point is a predetermined large value. Further, the distance-to-medium speed ratio calculating means 25 calculates the distance between the points and the distance direction speed when the direction of the distance direction speed coincides with the half line passing through the first point starting from the second point. The ratio of the magnitudes is calculated as the point-to-point medium transmission time from the second point to the first point. If the two do not match, the point-to-point medium transmission time from the second point to the first point is calculated. Set to a predetermined large value.

For example, in the case of the points A and B and the representative medium transmission speed shown in FIG. 10, the direction of the distance direction speed coincides with the direction of the half line passing through the point B starting from the point A.
The point-to-point medium transmission time from point A to point B is point AB
It is calculated by dividing the distance by the magnitude of the speed in the distance direction. On the other hand, the point-to-point medium transmission time from point B to point A is set to a predetermined large value.

The other operations are the same as those according to the first embodiment, and the description is omitted.

As described above, according to the second embodiment, the component of each representative medium transmission speed in the linear direction connecting two points is calculated as the distance direction speed, and is used instead of the representative medium transmission speed. Since the point-to-point medium transmission time is calculated in this way, the effect that the point-to-point medium transmission time can be more appropriately calculated is obtained.

Embodiment 3 FIG. 11 is a block diagram showing a configuration of a data mining device according to Embodiment 3 of the present invention. In the figure, when a medium transmission speed is obtained for each point, a point-to-point representative speed is calculated from two medium transmission speeds corresponding to two points of each point pair, and a medium transmission speed series is indicated by a point 51. This is a point-to-point representative speed sequence calculation unit that converts the data into an inter-point representative speed sequence. The other components in FIG. 11 are the same as those according to the first embodiment, and a description thereof will not be repeated. However, the medium transmission speed is set for each point, and the medium speed classification means 6 uses a point-to-point representative speed sequence instead of the medium transmission speed sequence.

Next, the operation will be described. FIG. 12 is a diagram showing the correspondence between the medium transmission speed at two points and the representative speed between points. The point-to-point representative speed sequence calculating means 51 reads the medium transmission speed sequence for each point held in the medium transmission speed sequence holding means 5 and stores a point based on the two medium transmission speeds corresponding to each point of each point pair. The medium representative speed sequence is calculated to convert the medium transmission speed sequence into a point-to-point representative speed sequence.

For example, when the medium transmission speeds at the point A and the point B are as shown in FIG. 12A, based on the medium transmission speed corresponding to the point A and the medium transmission speed corresponding to the point B, for example, as shown in FIG. As shown in (b), the point-to-point representative speed for the points A and B is calculated by multiplying the sum of the vectors representing the two medium transmission speeds by 0.5.

The medium speed classifying means 6 uses the point-to-point representative speed sequence instead of the medium transmission speed sequence, and generates a representative medium transmission speed sequence based on the correspondence held in the correspondence holding means 2.

The other operations are the same as those in the first embodiment, and the description thereof will not be repeated.

As described above, according to the third embodiment, when the medium transmission speed is given to each point, for each point pair, the medium transmission speed corresponding to each of the two points of the point pair is obtained. Since the point-to-point representative speed is calculated based on the speed and the point-to-point representative speed is used instead of the medium transmission speed, an effect that the point-to-point medium transmission time can be more appropriately calculated can be obtained.

Embodiment 4 FIG. 13 is a block diagram showing a configuration of a data mining device according to Embodiment 4 of the present invention. In the figure, reference numeral 61 denotes a medium transmission speed sequence which is converted at each point in the medium transmission speed sequence into an average of the medium transmission speed and the medium transmission speed at one or more time points in the vicinity thereof to smooth the medium transmission speed sequence. Medium speed series smoothing means for converting into a converted medium transmission speed series. The other components in FIG. 13 are the same as those according to the first embodiment, and a description thereof will not be repeated. However, the medium speed classification means 6 uses the smoothed medium transmission speed sequence converted by the medium speed sequence smoothing means 61 instead of the medium transmission speed sequence.

Next, the operation will be described. FIG. 14 is a diagram showing an example of a smoothed medium transmission speed sequence, and FIG. 15 is a diagram showing an example of a representative medium transmission speed sequence generated from the smoothed medium transmission speed sequence of FIG. The medium speed sequence smoothing means 61 compares the medium transmission speed at each time constituting the medium transmission speed sequence held by the medium transmission speed sequence holding means 5 with one or more of the medium transmission speed and a time near the medium transmission speed. The medium transmission speed sequence is converted into an average of the medium transmission speeds to convert it to a smoothed medium transmission speed sequence, and the smoothed medium transmission speed sequence is supplied to the medium speed classification means 6.
When calculating the average of the angles of the medium transmission speed, first, 360 degrees are added to the angles as necessary, and the value obtained by subtracting the minimum angle from the maximum angle is 180.
Less than degrees. If this is not possible, no angle smoothing is performed. After adding 360 degrees,
If the value obtained by subtracting the minimum angle from the maximum angle is less than 180 degrees, the average is calculated by dividing the sum of those angles by the number of the medium transmission speed. For example, 355
Consider the case of calculating the average of degrees and five degrees. Since the difference between these angles is 350 degrees, 360 degrees is added to 5 degrees to obtain 365 degrees. Then, the difference between the angle of 355 ° and the angle is 10 °, which is less than 180 °. By calculating the average of 365 ° and 355 °, 360 °, that is, 0 ° is obtained as the average of 355 ° and 5 °. .

For example, when calculating the average of the medium transmission speed at each time and the medium transmission speed at the next time based on the medium transmission speed sequence shown in FIG. 3, a smoothed medium transmission speed sequence is generated. The converted medium transmission speed sequence is as shown in FIG. For example, since the directions of the medium transmission speed at the first time point and the second time point in FIG. 3 are 10 and 15, the direction of the speed at the first time point in the smoothed medium transmission speed series is 10 and 15, respectively. And 12.5 which is the average of 15. In FIG. 14, the medium transmission speed at the tenth time point in FIG. 3 is the same as that at the ninth time point.

The medium speed classifying means 6 uses the smoothed medium transmission speed series converted by the medium speed series smoothing means 61 instead of the medium transmission speed series, and uses the correspondence held in the correspondence holding means 2. , A representative medium transmission speed sequence is generated. For example, when the smoothed medium transmission speed sequence shown in FIG. 14 is supplied and a representative medium transmission speed sequence is generated based on the correspondence shown in FIGS. 5 (a) and 5 (b), the medium speed classifying means 6 shown in FIG. A representative medium transmission speed sequence shown is generated.

As described above, according to the fourth embodiment, the medium transmission speed at each time of forming the medium transmission speed sequence is calculated by using the medium transmission speed and the medium transmission speed at one or more nearby times. The average of the medium transmission speed series is converted to a smoothed medium transmission speed series, and the smoothed medium transmission speed series is used instead of the medium transmission speed series. The constraint between the label values is determined based on the smoothed medium transmission speed incorporating the above, and an effect is obtained that a more appropriate determination can be executed.

Embodiment 5 FIG. FIG. 16 is a block diagram showing a configuration of a data mining device according to Embodiment 5 of the present invention. In the figure, reference numeral 71 denotes an element designation extracting means for extracting a label value partial series extracted by the label value partial series extracting means 31 corresponding to an element designated in advance. The other components in FIG. 16 are the same as those according to the first embodiment, and a description thereof will not be repeated. However, the label value pair extraction means 34
Instead of the label value subsequence extracted by the label value subsequence extraction means 31, the series extracted by the element designation extraction means 71 is used.

Next, the operation will be described. FIG. 17 is a diagram showing another example of the label value series obtained at points A, B, and C. FIG. 18 is a diagram showing an example of the label value subsequence extracted by the label value subsequence extracting means 31. FIG. 19 is a diagram showing an example of a label value subsequence extracted by the element designation extraction unit 71.

In FIG. 17, points A, B, and C represent respective points, and a and b represent elements of the label values held respectively. The label value subsequence extracting means 31 of the label value pair generating means 9 extracts, for example, a part included in a predetermined time width from the label value series shown in FIG. 17 as a label value subsequence. For example, the predetermined time width is 2 unit times (in this case, 12
00 second), the first label value subsequence for the label value sequence in FIG. 17 is as shown in FIG.
At this time, it is assumed that both the label values at the start time and the end time of the two unit times having a predetermined time width are included in the same label value subsequence.

Then, the element designation extracting means 71 extracts a part corresponding to the element designated in advance from the label value subsequence extracted by the label value subsequence extracting means 31, and supplies the extracted part to the label value pair extracting means. The label value pair extraction unit 34 detects the label value subsequence from the element designation extraction unit 71 and the first value detected by the high frequency label value detection unit 32.
High-frequency label value, the transmission time constraint satisfaction point extracted by the point extraction means 33 and the point-to-point medium transmission time,
In addition, based on the set L (1) of frequent label values held in the frequent label value set management unit 13, the element designation extraction unit 71 out of the label values related to the transmission time constraint satisfaction point.
From the first time point of the label value subsequence from the reference point, the label value that is the label value after the point-to-point medium transmission time corresponding to the point and is an element of L (1) is replaced by the second high-frequency label value Generating a first high-frequency label value and a second high-frequency label value, and a high-frequency label value set candidate having a length of 2 based on both points and a time difference between the two; 11, the appearance frequency of the high-frequency label value set candidate is updated.

For example, when the label value subsequence extracting means 31 extracts the label value subsequence shown in FIG. 18, if the element designation extracting means 71 has been designated in advance to extract the part relating to the element b, Figure 1 of the subsequence
As shown in FIG. 9, only the part related to the element b is extracted. The label value subsequence shown in FIG.
, The generated frequent label value set candidate is
The label value for element b is always one that is later in time.

Therefore, the element of L (2) is similarly
The label value for comes later. Further, since the label value set generating means 14 generates the element of L (3) by combining the elements of L (2), even in the high-frequency label value set which is the element of L (3), the last in time. All label values are for element b. Similarly, in all high-frequency label value sets, the temporally last label value
Will be about. Further, rules are extracted by the rule extracting unit 15 from these high-frequency label value sets, so that the label value of the conclusion part of the rule is related to the element b.

Note that the other operations are the same as those in the first embodiment, and a description thereof will be omitted.

As described above, according to the fifth embodiment, the second high-frequency label value is detected from the label value subsequence in which only the portion relating to the specified element is extracted, and the length of the second high-frequency label value is determined.
Since the frequent label value set candidate is generated, only the frequent label value set in which the label value of the specified element is the last label value in time can be extracted, and the conclusion part is specified. This has the effect of extracting only the rule that is the label value for the element that has been set. In addition, since the format of the high-frequency label value set is restricted as described above, the effect that the processing time is consequently reduced is obtained.

Embodiment 6 FIG. FIG. 20 is a block diagram showing a configuration of a data mining device according to Embodiment 6 of the present invention. In the figure, reference numeral 81 denotes a non-correlated point pair specifying means for specifying one or a plurality of non-correlated point pairs composed of two points in response to a user operation. The other components in FIG. 20 are the same as those according to the first embodiment, and a description thereof will not be repeated.

Next, the operation will be described. When one or a plurality of uncorrelated point pairs composed of two points are specified by the user, the uncorrelated point pair specifying means 81 supplies the uncorrelated point pairs to the point extracting means 33. An uncorrelated pair of points is a pair of points that are known in advance not to transmit an influence between those points.

The point extracting means 33 detects a point including the appearance point of the first high-frequency label value among the uncorrelated point pairs, and detects the point which is not the appearance point of the first high-frequency label value among the uncorrelated point pairs. The points are deleted from the transmission time constraint satisfaction points, and the remaining transmission time constraint satisfaction points and the inter-point medium transmission time are supplied to the label value pair extraction means 34.

Note that the other operations are the same as those in the first embodiment, and a description thereof will be omitted.

As described above, according to the sixth embodiment, when there is a pair of uncorrelated points whose influence is known not to be transmitted in advance, the transmission time to one point of the pair of uncorrelated points is determined. Since the other point of the uncorrelated point pair is not extracted as the constraint satisfaction point, there is an effect that an insignificant high-frequency label value set regarding the uncorrelated point pair whose influence is not transmitted is not generated. In addition, the format of the high-frequency label value set is restricted as described above, and the effect that the processing time is shortened is obtained.

In the first to sixth embodiments,
A computer-readable recording medium recording a program for causing a computer to execute the above-described processing,
It can also be realized by a computer. In this case, a RAM or a hard disk device connected to the CPU via a bus in the computer functions as each of the above-described holding units, the CPU functions as other units, and various types of data are exchanged via the bus.

[0145]

As described above, according to the present invention, based on the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed representing each range, Medium speed classification means for converting each medium transmission speed constituting a medium transmission speed sequence which is a time series of medium transmission speeds into a representative medium transmission speed to generate a representative medium transmission speed sequence, and spatial information and representatives of each point Based on the medium transmission speed, the time at which the medium moves from the first point to the second point and the first point from the second point, with the two points of each point pair as a first point and a second point. Calculating means for calculating the time required for the medium to move to the medium as the point-to-point medium transmission time, and one of the temporally last label value and the remaining label value among the label values constituting the high frequency label value set candidate
The label value on the front side in time of the two label value set composed of the first and second label values is regarded as the front label value, and the label value on the rear side in time is regarded as the rear label value. Whether the time difference between the current label value and the current time of the rear label value is equal to or greater than the point-to-point medium transmission time from the current position of the previous label value to the current position of the rear label value at the current time of the previous label value Label-value constraint determining means for determining whether or not each of the high-frequency label value candidates appears in a label value series which is a time series of label values relating to one or more elements depending on a medium; A label value set verifying means for counting the frequency of occurrence of each high-frequency label value set candidate in the label value series only when the inter-value constraint determining means determines that the constraint condition is satisfied; A high-frequency label value set extracting means for setting the high-frequency label value candidates or the high-frequency label value set candidates as the high-frequency label values or the high-frequency label value sets, respectively, and the two high-frequency label values as the first label values And the second
Of the label value pairs combined as the label value of
The time difference between the present time of the label value and the present time of the second label value is the time difference between the appearance point of the first label value and the appearance point of the second label value at the present time of the first label value. A label value pair generating means that is longer than the point-to-point medium transmission time and shorter than a predetermined maximum transmission time, and that appears in the label value sequence as a high-frequency label value set candidate; and a plurality of different high-frequency label value sets. And a high-frequency label value set candidate generating means for generating a new high-frequency label value set candidate by combining the rules. There is an effect that temporal association rules can be extracted except for. That is, there is an effect that a meaningless pattern having no dependency is not generated due to the restriction due to the medium transmission time between points. Further, since a meaningless pattern is not generated, there is an effect that processing time for rule extraction can be reduced.

According to the present invention, of the rule candidates in which the temporally last label value among the label values constituting the frequent label value set is the conclusion part and the remaining label values are the condition parts,
A rule extracting means for extracting, as a rule, a rule in which the ratio between the appearance frequency of the condition part in the label value series and the appearance frequency of the high-frequency label value set is within a predetermined range is provided, so that transmission between points is performed. There is an effect that a temporal correlation rule can be extracted except for a rule determined to have no dependency due to time constraints.

According to the present invention, the distance calculating means for calculating the distance between the first point and the second point as the point-to-point distance based on the spatial information of the point, and the first point as the starting point Angle difference calculating means for calculating, as an angle difference, an angle between a half line passing through the second point and the direction of the representative medium transmission speed, and a point-to-point distance, representative medium transmission speed, and angle difference calculated by the distance calculating means The medium transmission time calculation means for calculating the point-to-point medium transmission time based on the angle difference calculated by the calculation means is configured to have in the calculation means, so that an appropriate point-to-point medium transmission time taking into account the movement of the medium is calculated. This has the effect of being able to calculate.

According to the present invention, the distance calculating means
Since the distance between points is calculated from the latitude and longitude of each point as spatial information, if the time required for transmitting the medium in the altitude direction can be ignored, an appropriate point-to-point medium transmission time is obtained. There is an effect that can be.

According to the present invention, the distance calculating means
Since the distance between points is calculated from the latitude, longitude and altitude of each point as spatial information, it is appropriate to transmit the medium between points even if the time required to transmit the medium in the height direction cannot be ignored. Is obtained.

According to the present invention, whether or not the medium moves from the first point to the second point according to whether or not the angle difference calculated by the angle difference calculation means is within a predetermined range. And an angle difference determining means for determining whether or not the medium moves from the second point to the first point, and calculating a distance when it is determined that the medium moves from the first point to the second point Calculating the ratio of the point-to-point distance calculated by the means to the representative medium transmission speed as the point-to-point medium transmission time from the first point to the second point, and moving the medium from the first point to the second point. If it is determined that the medium is not to be transmitted, the point-to-point medium transmission time from the first point to the second point is set to a predetermined large value, and if it is determined that the medium moves from the second point to the first point, Is the ratio of the distance between the points and the representative medium transmission speed from the second point to the first point Is calculated as the point-to-point medium transmission time, and when it is determined that the medium does not move from the second point to the first point, the point-to-point medium transmission time from the second point to the first point is calculated as a predetermined time. Since the medium transmission time calculating means is provided with the distance-to-medium speed ratio calculating means having a large value, there is an effect that the point-to-point medium transmitting time can be easily and appropriately calculated.

According to the present invention, the representative medium transmission speed is
A distance direction speed calculating means for calculating a component in a direction of a straight line connecting the first point and the second point as a distance direction speed;
If the direction of the speed in the distance direction matches the direction of the half-line passing through the second point starting from the point of the second point, the ratio of the distance between the points and the magnitude of the distance direction speed is determined from the first point to the second point. Is calculated as the point-to-point medium transmission time, and when the direction of the half-line and the distance direction velocity do not match, the point-to-point medium transmission time from the first point to the second point is set to a predetermined large value. When the direction of the semi-straight line passing through the first point and the direction of the distance direction from the second point coincide, the ratio of the distance between the points and the magnitude of the distance direction speed is calculated from the second point to the second point. 1
Is calculated as the point-to-point medium transmission time to the point, and when the direction of the half-line and the distance direction velocity do not match, the point-to-point medium transmission time from the second point to the first point is set to a predetermined large value. Since the distance-to-medium speed ratio calculating means having the value is provided in the medium transmission time calculating means, there is an effect that the point-to-point medium transmission time can be more appropriately calculated.

According to the present invention, when a medium transmission speed sequence is obtained for each point, an inter-point representative speed sequence is calculated from two medium transmission speed sequences respectively corresponding to two points of each point pair. A point-to-point representative speed series calculating means,
Since the medium speed classification means is configured to use the point-to-point representative speed sequence instead of the medium transmission speed sequence, there is an effect that the point-to-point medium transmission time can be more appropriately calculated.

According to the present invention, the medium transmission speed at each point in the medium transmission speed sequence is converted into an average of the medium transmission speed and the medium transmission speed at one or more time points near the medium transmission speed. Medium speed series smoothing means for converting the speed series into a smoothed medium transmission speed series, wherein the medium speed classification means converts the smoothed medium transmission speed series converted by the medium speed series smoothing means instead of the medium transmission speed series. Since it is configured to use, the constraint condition between the label values is determined based on the smoothed medium transmission speed incorporating the properties of the plurality of medium transmission speeds, and more appropriate determination can be performed. There is.

According to the present invention, a label value subsequence extracting means for extracting a subsequence included in a predetermined time width from a label value sequence as a label value subsequence, and a label value at the first time of the label value subsequence Is a high-frequency label value, the label value is a first high-frequency label value, a high-frequency label value detecting means, and the first high-frequency label value at the time when the first high-frequency label value is output Point extraction means for extracting, as a transmission time constraint satisfaction point, a point where the point-to-point medium transmission time is less than or equal to a predetermined maximum transmission time when the appearance point of the symbol is a starting point, and a label related to the transmission time constraint satisfaction point in the label value subsequence Of the values, the label value that is the label value from the initial point of the label value subsequence to the point-to-point medium transmission time corresponding to that point and that is the frequent label value is the second high frequency value. Since the label value pair generating means has label value pair and label value pair extracting means for generating a high frequency label value set candidate with the first high frequency label value and the second high frequency label value, the medium There is an effect that a high-frequency label value set candidate having a length of 2 that satisfies the constraint on the transmission time can be easily generated.

According to the present invention, the label value pair generating means has an element designation extracting means for extracting, from the label value subsequence by the label value subsequence extracting means, the one corresponding to the element designated in advance. Since the extracting means is configured to use the series extracted by the element specification extracting means instead of the label value subsequence by the label value subsequence extracting means, the label value of the specified element is temporally the last label value. It is possible to extract only the high-frequency label-value group that satisfies the following condition, and it is possible to extract only the rule whose label is the label value of the element whose conclusion is specified. In addition, since the format of the high-frequency label value set is restricted as described above, there is an effect that the processing time is consequently reduced.

According to the present invention, the point extracting means detects a point including an appearance point of the first high-frequency label value among the predetermined uncorrelated point pairs, and detects the first high-frequency label among the uncorrelated point pairs. Since the point which is not the point where the label value appears is deleted from the point of satisfaction of the transmission time constraint, there is an effect that an insignificant high-frequency label value set regarding the uncorrelated point pair whose influence is not transmitted is not generated. In addition, the format of the high-frequency label value set is restricted as described above, and the processing time is shortened.

According to the present invention, based on the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed representing each of the ranges, the medium transmission speed is time-series. Converting each medium transmission speed constituting a certain medium transmission speed sequence into a representative medium transmission speed to generate a representative medium transmission speed sequence; and 2 based on spatial information of two points and each representative medium transmission speed. Two points from the first point to the second point as the first point and the second point.
Time for the medium to move to the point and the first time from the second point
Calculating the time required for the medium to move to a point as a medium-to-point medium transmission time; and each high-frequency label value candidate in a label value series that is a time series of label values for one or more elements depending on the medium. Counting the frequency of appearance, and the step of setting a high frequency label value candidate whose frequency of appearance is equal to or higher than a predetermined lower limit value as a high frequency label value,
Among the label value sets in which two high-frequency label values are combined as a first label value and a second label value, the time difference between the present time of the first label value and the present time of the second label value is the first. Is greater than or equal to the point-to-point medium transmission time between the point of appearance of the first label value and the point of appearance of the second label value at the present time, and less than or equal to a predetermined maximum transmission time, and the label value Making the one appearing in the series a frequent label value set candidate, and comprising a temporally last label value and one of the remaining label values among the label values constituting the frequent label value set candidate The label value on the temporally front side of the two label value set to be used is the front label value, and the label value on the temporal side is the rear label value. Time relative to the present time of the label value And determining whether or not the condition is equal to or longer than the point-to-point medium transmission time from the present time of the front label value to the appearance point of the rear label value at the present time of the front label value, and the constraint condition Counting the frequency of appearance of each high-frequency label value set candidate in the label value series only when it is determined that the above is satisfied. A value set and a step of generating a new high-frequency label value set candidate by combining a plurality of different high-frequency label value sets. There is an effect that a temporal correlation rule can be extracted except for a rule determined to be unrelated. That is, there is an effect that a meaningless pattern having no dependency is not generated due to the restriction due to the medium transmission time between points. Further, since a meaningless pattern is not generated, there is an effect that processing time for rule extraction can be reduced.

According to the present invention, based on the correspondence between the range of the medium transmission speed, which is the speed at which the medium moves in the space, and the representative medium transmission speed representing each of the ranges, the medium transmission speed is represented in time series. Medium speed classification means for converting each medium transmission speed constituting a certain medium transmission speed sequence into a representative medium transmission speed sequence to generate a representative medium transmission speed sequence, based on spatial information at each point and the representative medium transmission speed, The time at which the medium moves from the first point to the second point and the time at which the medium moves from the second point to the first point are defined as the two points of the point pair as the first point and the second point, respectively. Calculating means for calculating the inter-medium transmission time, a two-label value set composed of a temporally last label value of label values constituting one of the high-frequency label value set candidates and one of the remaining label values In time ago For all front label values, the time difference between the current time of the front label value and the current time of the rear label value is the front label for all front label values. The label value constraint determining means for determining whether or not the constraint that the medium transmission time from the point of appearance of the front label value at the present point of time to the point of appearance of the rear label value is longer than the point-to-point medium transmission time is satisfied. Count the appearance frequency of each high-frequency label value candidate in a label value series that is a time series of label values for one or more dependent elements, and when it is determined by the label value constraint determination unit that the constraint condition is satisfied. Label value set verification means for counting the appearance frequency of each high-frequency label value set candidate in the label value series only, a high-frequency label value candidate or a high-frequency label whose appearance frequency is equal to or higher than a predetermined lower limit. Label value set extracting means, each of which sets the value set candidate as a high-frequency label value or a high-frequency label value set, and a label value set in which two high-frequency label values are combined as a first label value and a second label value The time difference between the present time of the first label value and the present time of the second label value is the appearance point of the first label value and the appearance of the second label value at the present time of the first label value A label value pair generation unit that is longer than the point-to-point medium transmission time between the points and is shorter than a predetermined maximum transmission time, and furthermore, a label value series that appears in a label value series as a high-frequency label value set candidate; and Since a program for causing a computer to function as a high-frequency label value set candidate generating means for generating a new high-frequency label value set candidate by combining different high-frequency label value sets is recorded on a recording medium, When the program is executed by a computer, there is an effect that a temporal correlation rule can be extracted except for a rule that is determined to have no dependency due to a restriction due to a medium transmission time between points. That is, there is an effect that a meaningless pattern having no dependency is not generated due to the restriction due to the medium transmission time between points. Further, since a meaningless pattern is not generated, there is an effect that processing time for rule extraction can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a data mining device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing an example of points A, B, and C.

FIG. 3 is a diagram illustrating an example of a medium transmission speed sequence.

FIG. 4 is a diagram showing an example of a label value sequence obtained at points A, B, and C.

FIG. 5 is a diagram illustrating an example of a correspondence relationship between each range of a medium transmission speed and a representative medium transmission speed.

FIG. 6 is a diagram illustrating an example of a representative medium transmission speed sequence.

FIG. 7 is a diagram illustrating an example of an angle between a half line at two points and a direction of a representative medium transmission speed.

FIG. 8 is a diagram showing an example of a label value subsequence extracted by a label value subsequence extracting means.

FIG. 9 is a block diagram showing a configuration of a data mining device according to a second embodiment of the present invention.

FIG. 10 is a diagram showing a correspondence relationship between a representative medium transmission speed and a distance direction speed.

FIG. 11 is a block diagram showing a configuration of a data mining device according to Embodiment 3 of the present invention.

FIG. 12 is a diagram illustrating a correspondence between a medium transmission speed at two points and a representative speed between points.

FIG. 13 is a block diagram showing a configuration of a data mining device according to Embodiment 4 of the present invention.

FIG. 14 is a diagram illustrating an example of a smoothing medium transmission speed sequence.

15 is a diagram showing an example of a representative medium transmission speed sequence generated from the smoothed medium transmission speed sequence of FIG.

FIG. 16 is a block diagram showing a configuration of a data mining device according to Embodiment 5 of the present invention.

FIG. 17 is a diagram showing another example of a label value sequence obtained at points A, B, and C.

FIG. 18 is a diagram showing an example of a label value subsequence extracted by a label value subsequence extracting means.

FIG. 19 is a diagram illustrating an example of a label value subsequence extracted by an element designation extraction unit.

FIG. 20 is a block diagram showing a configuration of a data mining device according to Embodiment 6 of the present invention.

FIG. 21 is a diagram showing an example of data to be subjected to the MSDD method.

FIG. 22 is a diagram showing an example of a rule extracted from the data shown in FIG. 21 by the MSDD method.

[Explanation of symbols]

3 calculation means, 6 medium speed classification means, 7 label value constraint determination means, 9 label value pair generation means, 10 label value set verification means, 12 high frequency label value set extraction means, 14
Label value set generating means (high-frequency label value set candidate generating means), 15 rule extracting means, 21 angle difference calculating means,
22 distance calculation means, 23 medium transmission time calculation means, 2
4 angle difference determination means, 25 distance to medium speed ratio calculation means, 31 label value partial sequence extraction means, 32 high frequency label value detection means, 33 point extraction means, 34 label value pair extraction means, 41 distance direction velocity calculation means, 51 point-to-point representative speed sequence calculating means, 61 medium speed sequence smoothing means, 71 element designation extracting means.

Claims (14)

[Claims] 1. A time series of the medium transmission speed based on a correspondence relationship between a range of a medium transmission speed, which is a speed at which a medium moves in a space, and a representative medium transmission speed representing each range of the medium transmission speed. Medium speed classifying means for converting each medium transmission speed constituting the medium transmission speed sequence to be the representative medium transmission speed to generate a representative medium transmission speed sequence; spatial information of each point and the representative medium transmission speed The time at which the medium moves from the first point to the second point with the two points of each point pair as a first point and a second point, and from the second point to the first point Calculating means for calculating the time required for the medium to travel as a medium-to-point medium transmission time; and a temporally last label value and one of the remaining label values among the label values constituting the high-frequency label value set candidate. Composed of The label value on the temporally front side of the two label value set to be used is the front label value, and the label value on the temporal side is the rear label value. Constraint that the time difference from the present time of the rear label value is equal to or longer than the point-to-point medium transmission time from the appearance point of the front label value to the appearance point of the rear label value at the present time of the front label value. A label value constraint determining means for determining whether or not the above-mentioned condition is satisfied; and determining the appearance frequency of each high-frequency label value candidate in a label value series which is a time series of label values relating to one or more elements depending on the medium. A label value set that counts and counts the appearance frequency of each high-frequency label value set candidate in the label value series only when it is determined that the constraint condition is satisfied by the label value constraint determination unit. Verification means, the high frequency label value candidate or high frequency label value set candidate where the appearance frequency is equal to or more than a predetermined lower limit value, and the high frequency label value set high frequency label value set extraction means, The two frequent label values are a first label value and a second label value
The time difference between the present point of time of the first label value and the present point of time of the second label value is the first label value at the present point of time of the first label value. The frequency between the point-to-point medium transmission time between the point of appearance of the label value and the point of appearance of the second label value is longer than or equal to the predetermined maximum transmission time, and appears in the label value series at a high frequency. Data mining comprising: label value pair generation means as label value set candidates; and high frequency label value set candidate generation means for generating a new high frequency label value set candidate by combining a plurality of different high frequency label value sets. apparatus. 2. The rule part in a label value series among rule candidates in which a temporally last label value among label values constituting a high-frequency label value set is a conclusion part and the remaining label values are condition parts. 2. The data mining apparatus according to claim 1, further comprising a rule extracting unit that extracts, as a rule, a rule having a ratio between the frequency of occurrence of the high frequency label value set and the frequency of occurrence of the high frequency label value set. 3. A distance calculating means for calculating a distance between a first point and a second point as a point-to-point distance based on spatial information of the point, and starting from the first point. Angle difference calculating means for calculating, as an angle difference, an angle formed between a half line passing through the second point and the direction of the representative medium transmission speed, the distance between the points calculated by the distance calculating means, the representative medium transmission 2. The data mining apparatus according to claim 1, further comprising a medium transmission time calculation unit that calculates a point-to-point medium transmission time based on a speed and the angle difference calculated by the angle difference calculation unit. 4. The data according to claim 3, wherein the spatial information of each point includes a latitude and a longitude of the point, and the distance calculating means calculates a distance between the points from the latitude and the longitude. Mining equipment. 5. The spatial information of each point is the latitude of the point,
4. The data mining apparatus according to claim 3, wherein the data mining apparatus has a longitude and an altitude, and the distance calculating means calculates a distance between the points based on the latitude, the longitude and the altitude. 6. The medium transmission time calculating means moves the medium from the first point to the second point according to whether or not the angle difference calculated by the angle difference calculating means is within a predetermined range. Angle difference determining means for determining whether the medium moves from the second point to the first point, and determining that the medium moves from the first point to the second point. In this case, the ratio between the point-to-point distance calculated by the distance calculating means and the representative medium transmission speed is calculated as the point-to-point medium transmission time from the first point to the second point. If it is determined that the medium does not move from the point to the second point, the second point is moved from the first point to the second point.
The point-to-point medium transmission time to the point is a predetermined large value,
When it is determined that the medium moves from the second point to the first point, the ratio between the point-to-point distance and the representative medium transmission speed is set to a point from the second point to the first point. It is calculated as the intermediary medium transmission time, and when it is determined that the medium does not move from the second point to the first point, the interpoint medium transmission time from the second point to the first point is calculated. 4. A data mining apparatus according to claim 3, further comprising a distance-to-medium speed ratio calculating means for setting a predetermined large value. 7. The medium transmission time calculation means includes: a distance direction speed calculation means for calculating a component of a representative medium transmission speed in a direction of a straight line connecting the first point and the second point as a distance direction speed; If the direction of the distance direction speed coincides with the half line passing through the second point starting from the first point, the ratio of the distance between the points and the magnitude of the distance direction speed is set to the first point. Calculated as a point-to-point medium transmission time from a point to the second point, and if the direction of the half-line and the distance direction velocity do not match, the first point to the second point The point-to-point medium transmission time is a predetermined large value, and when the direction of the distance direction velocity coincides with the half-line passing through the first point with the second point as a starting point, the point-to-point distance and The ratio to the magnitude of the distance direction velocity is calculated at the second point. From the second point to the first point when the direction of the half-line is not coincident with the direction of the distance speed. 4. A data mining apparatus according to claim 3, further comprising a distance-to-medium speed ratio calculating means for setting the inter-medium transmission time to a predetermined large value. 8. A point-to-point representative calculating a point-to-point representative speed sequence from two medium transmission speed sequences respectively corresponding to two points of each point pair when a medium transmission speed sequence is obtained for each point. 2. The data mining apparatus according to claim 1, further comprising speed sequence calculating means, wherein the medium speed classifying means uses the point-to-point representative speed sequence instead of the medium transmission speed sequence. 9. The medium transmission speed sequence at each time point constituting the medium transmission speed sequence is converted into an average of the medium transmission speed and the medium transmission speed at one or more time points near the medium transmission speed to smooth the medium transmission speed sequence. Medium speed series smoothing means for converting to a medium transmission speed series, wherein the medium speed classification means replaces the medium transmission speed series with the smoothed medium transmission speed series converted by the medium speed series smoothing means. The data mining device according to claim 1, wherein the data mining device is used. 10. A label value pair generating means, comprising: a label value subsequence extracting means for extracting a subsequence included in a predetermined time width from the label value series as a label value subsequence; A high-frequency label value detecting means for setting the label value as a first high-frequency label value when the label value at the time is a high-frequency label value; and Point extraction means for extracting, as a transmission time constraint satisfaction point, a point whose medium transmission time between points is less than or equal to a predetermined maximum transmission time when the appearance point of the high-frequency label value of 1 is a starting point; Among the label values related to the transmission time constraint satisfying point, the label value is a label value after the point-to-point medium transmission time corresponding to the point from the first time point of the label value subsequence, and Label value pair extracting means for setting a label value being a label value as a second high-frequency label value and generating a high-frequency label value set candidate with the first high-frequency label value and the second high-frequency label value The data mining device according to claim 1, wherein: 11. The label value pair generation means includes element designation extraction means for extracting a value corresponding to a predetermined element from the label value subsequence by the label value subsequence extraction means. 11. The data mining apparatus according to claim 10, wherein a sequence extracted by the element designation extracting unit is used instead of the label value subsequence by the label value subsequence extracting unit. 12. A point extracting means for detecting a point including an appearance point of a first high-frequency label value among predetermined non-correlation point pairs, and detecting the first high-frequency label among the non-correlation point pairs. 11. The data mining apparatus according to claim 10, wherein a point that is not a value appearance point is deleted from a transmission time constraint satisfaction point. 13. A time series of the medium transmission speed based on a correspondence between a range of a medium transmission speed that is a speed at which a medium moves in a space and a representative medium transmission speed representing each range of the medium transmission speed. Converting each medium transmission speed constituting the medium transmission speed sequence into a representative medium transmission speed to generate a representative medium transmission speed sequence, based on spatial information of two points and each representative medium transmission speed The time at which the medium moves from the first point to the second point, and the medium moves from the second point to the first point, using the two points as a first point and a second point. Calculating each time as a point-to-point medium transmission time; and the appearance of each high frequency label value candidate in a label value series that is a time series of label values for one or more elements depending on the medium. A step of counting the degrees, the steps of frequency of occurrence and high frequency label value high frequency label value candidate is equal to or higher than a predetermined lower limit value, two of the high frequency label value first label value and the second
The time difference between the present point of time of the first label value and the present point of time of the second label value is the first label value at the present point of time of the first label value. The frequency between the point-to-point medium transmission time between the point of appearance of the label value and the point of appearance of the second label value is longer than or equal to the predetermined maximum transmission time, and appears in the label value series at a high frequency. A label value set candidate; and a temporally last label value and one of the remaining label values among the label values constituting the frequent label value set candidates.
The label value on the temporally front side of the two label value set composed of two labels is the front label value, and the label value on the temporal side is the rear label value. The time difference between the outgoing current time and the outgoing current time of the rear label value is equal to or longer than the point-to-point medium transmission time from the appearance point of the front label value at the present time of the front label value to the appearance point of the rear label value. Determining whether or not the constraint condition is satisfied; and counting the appearance frequency of each high-frequency label value set candidate in the label value series only when it is determined that the constraint condition is satisfied; A step of setting a high-frequency label value set candidate whose frequency is equal to or higher than a predetermined lower limit value as a high-frequency label value set; and combining a plurality of different high-frequency label value sets with a new high-frequency label value set candidate. Data mining method which includes a step of forming. 14. The computer according to claim 1, wherein said medium transmission speed is determined based on a correspondence relationship between a range of a medium transmission speed, which is a speed at which a medium moves in a space, and a representative medium transmission speed representing each range of said medium transmission speed. Medium speed classifying means for converting each medium transmission speed constituting the medium transmission speed sequence which is a time series into the representative medium transmission speed to generate a representative medium transmission speed sequence, spatial information of each point and the representative medium transmission Based on the speed, two points of each point pair are defined as a first point and a second point, and the time required for the medium to move from the first point to the second point and the first point from the second point to the first point. Calculating means for calculating the time required for the medium to move to the point as the medium-to-point medium transmission time, of the temporally last label value of the label values constituting the high-frequency label value set candidate and the remaining label value A label value on the temporally front side of the two label value set constituted by one of the following label values and a label value on the temporally rear side as the rear label value, and for all the front label values, the front label The time difference between the present time of the value output and the present time of the rear label value is equal to or longer than the point-to-point medium transmission time from the appearance point of the front label value to the appearance point of the rear label value at the present time of the front label value. Label value constraint determination means for determining whether or not the constraint condition is satisfied. Each high-frequency label value candidate in a label value series which is a time series of label values for one or more elements depending on the medium. Is counted, and the frequency of appearance of each high-frequency label value set candidate in the label value series is counted only when it is determined that the constraint condition is satisfied by the label value constraint determining unit. Label value set verification means, high frequency label value set extraction means for setting the high frequency label value candidate or the high frequency label value set candidate whose appearance frequency is equal to or higher than a predetermined lower limit value to be a high frequency label value or a high frequency label value set, respectively The two frequent label values into a first label value and a second
The time difference between the present point of time of the first label value and the present point of time of the second label value is the first label value at the present point of time of the first label value. The frequency between the point-to-point medium transmission time between the point of appearance of the label value and the point of appearance of the second label value is longer than or equal to the predetermined maximum transmission time, and appears in the label value series at a high frequency. Label value pair generating means as label value set candidates; a program for functioning as a high frequency label value set candidate generating means for generating a new high frequency label value set candidate by combining a plurality of different high frequency label value sets. A computer-readable recording medium that has been recorded.