JP4711863B2 - Sequence pattern discovery program, sequence pattern discovery method, and sequence pattern discovery device - Google Patents

Description

  The present invention relates to a sequence pattern discovery program, a sequence pattern discovery method, and a sequence pattern discovery device that extract a sequence pattern that frequently appears in the sequence data from sequence data in which fluctuating item values are arranged in order. The present invention relates to a sequence pattern discovery program, a sequence pattern discovery method, and a sequence pattern discovery device that can reduce the processing time required for sequence pattern extraction by efficiently counting the number of occurrences of a sequence pattern.

  In recent years, in a wide range of fields such as sales, distribution, stock trading, etc., a technique called data mining is used to analyze past data accumulated in computers and search for correlations and patterns between item values contained in the data. Supporting decision making is done. For example, in the field of stock trading, a series pattern representing fluctuations in stock prices is predicted from past stock price data, and the predicted pattern is used as an index for determining the timing of stock trading.

  As such a pattern prediction technique, for example, in Patent Document 1, there is a technique for extracting a time series pattern indicating the order in which each event appears from time series data in which the appearing events are arranged in time series as item values. It is shown. In this technology, concepts that classify possible events are stored in a hierarchical structure in advance, and the frequency of occurrence of each concept (event) is counted by referring to time-series data in order from the lowest concept. To do.

  Further, with respect to the time series pattern in which the concepts are combined in time series, the frequency of appearance of each time series pattern is totaled by referring to the time series data while increasing the number of concepts to be combined. When increasing the number of concepts to be combined, new concepts are added one by one to a time-series pattern that has already been aggregated in frequency and whose frequency is equal to or greater than a predetermined threshold. In this way, finally, a time series pattern of concepts (events) occurring at a frequency equal to or higher than a predetermined threshold is extracted.

JP 2005-84919 A

  However, in the above technique, when the frequency of occurrence of each time series pattern is totaled, it takes time for the totaling process to refer to all the time series data for each time series pattern. If the amount of time-series data to be extracted is enormous, if the predetermined threshold is set small, or if the concept hierarchy is set deep, processing may not end within business hours. There is.

  The present invention has been made to solve the above-described problems caused by the prior art, and is a sequence pattern that can shorten the processing time required to extract a sequence pattern by efficiently counting the number of occurrences of the sequence pattern. An object is to provide a discovery program, a sequence pattern discovery method, and a sequence pattern discovery device.

  In order to solve the above-described problems and achieve the object, the sequence pattern finding program according to the invention of claim 1 is a sequence that frequently appears in sequence data from sequence data in which fluctuating item values are arranged in order. A sequence pattern finding program for extracting a pattern, including a sequence data storage procedure for storing sequence data to be sequence pattern extracted in a storage device, and an arbitrary item value and an item value that allow arbitrary item values Defining a sequence pattern as a generalized pattern, sequentially reading out the sequence data stored by the sequence data storage procedure, and counting the number of occurrences of the generalized pattern that matches the sequence data; and the sequence pattern A generalized pattern in which the number of occurrences counted by the counting procedure is equal to or greater than a predetermined threshold is extracted and included in the extracted generalized pattern. By setting a selection item value that selects an allowable value to an arbitrary item value that is generated, sequentially generating a sequence pattern, sequentially reading out the sequence data stored by the sequence data storage procedure, and collating with the sequence pattern, A sequence pattern finding procedure for counting the number of occurrences of a sequence pattern that matches the sequence data, and a sequence for outputting a sequence pattern in which the number of occurrences counted by the sequence pattern finding procedure is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result A pattern output procedure is executed by a computer.

  According to a second aspect of the invention, there is provided the sequence pattern finding program according to the first aspect, wherein the sequence pattern counting procedure includes a generalized pattern number for uniquely identifying the generalized pattern and the appearance of the generalized pattern. A generalized map generation procedure for generating a generalized map which is a map for deriving the generalized pattern number from the sequence data, comprising a generalized pattern table associated with a generalized pattern counter for counting numbers And sequentially reading out the sequence data stored by the sequence data storage procedure, deriving the generalized pattern number from the read sequence data using the generalized mapping, and deriving the derived generalization pattern table Generalized pattern counter counts up the generalized pattern counter corresponding to the generalized pattern number Characterized in that to execute the order, to the computer.

  According to a third aspect of the present invention, there is provided the sequence pattern finding program according to the first or second aspect, wherein the number of occurrences counted by the sequence pattern counting procedure is greater than or equal to a predetermined threshold value. A sequence pattern is generated by setting a selected item value to one of the arbitrary item values included in the extracted generalized pattern, and the sequence data storage procedure By comparing the sequence data stored in step (2), the number of occurrences of the sequence pattern that matches the sequence data is counted, the sequence pattern in which the counted occurrence number is equal to or greater than a predetermined threshold is extracted, and the extracted sequence The series pattern is further refined by setting the selection item value to one of the arbitrary item values included in the pattern, and the detailed series pattern is displayed. And the sequence data stored by the sequence data storage procedure, the number of occurrences of the sequence pattern that matches the sequence data is counted, and selection items are selected for all arbitrary item values included in the sequence pattern. Until the value is set, detailing of the sequence pattern and counting of the number of appearances of the sequence pattern are repeated.

  According to a fourth aspect of the invention, there is provided the sequence pattern finding program according to the first or second aspect of the invention, wherein the sequence pattern counting procedure includes a sequence pattern including only a selection item value and an item value for which an allowable value is selected. A detailed pattern table defined as a detailed pattern, in which a detailed pattern number for uniquely identifying the detailed pattern is associated with a detailed pattern counter for counting the number of occurrences of the detailed pattern; A refined map generation procedure for generating a refined map that is a map for deriving the refined pattern number from the sequence data, and sequentially reading the sequence data stored by the sequence data storage procedure, and the read sequence Deriving the refined pattern number from the data using the refined map, and deriving the refined pattern table And Details pattern counter counting procedure for counting up a detailed pattern counter corresponding to the detailed pattern number, characterized by causing a computer to execute the.

  According to a fifth aspect of the invention, there is provided the sequence pattern finding program according to any one of the first to fourth aspects, wherein the sequence data storage procedure discretizes a continuously changing value, and calculates a value for each discretized interval. It is characterized in that series data arranged in order as item values is stored as a reference for pattern extraction.

  The sequence pattern finding method according to the invention of claim 6 is a sequence pattern finding method for extracting a sequence pattern frequently appearing in the sequence data from sequence data in which fluctuating item values are arranged in order. A sequence data storage step for storing sequence data to be sequence pattern extracted in a storage device, and defining a sequence pattern including only an arbitrary item value and an item value that allow an arbitrary item value as a generalized pattern, The sequence data stored in the sequence data storage step is sequentially read out, the sequence pattern counting step for counting the number of appearances of generalized patterns that match the sequence data, and the number of occurrences counted by the sequence pattern counting step is a predetermined number. A selection item that extracts a generalized pattern that is equal to or greater than a threshold and selects an allowable value as an arbitrary item value included in the extracted generalized pattern A sequence pattern is generated by sequentially generating sequence patterns, sequentially reading out the sequence data stored in the sequence data storage step, and comparing the sequence data with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data A pattern finding step; and a sequence pattern output step of outputting a sequence pattern whose number of occurrences counted in the sequence pattern finding step is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result.

  Further, the sequence pattern finding device according to the invention of claim 7 is a sequence pattern finding device for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which the fluctuating item values are arranged in order. Series data storage means for storing series data to be sequence pattern extracted in a storage device, and a series pattern including only an arbitrary item value and an item value that allow an arbitrary item value is defined as a generalized pattern, The sequence data stored in the sequence data storage unit is sequentially read out, the sequence pattern counting unit that counts the number of appearances of the generalized pattern that matches the sequence data, and the number of occurrences counted by the sequence pattern counting unit is a predetermined number. A selection item that extracts a generalized pattern that is equal to or greater than a threshold and selects an allowable value as an arbitrary item value included in the extracted generalized pattern A sequence pattern is generated by sequentially generating sequence patterns, sequentially reading out the sequence data stored by the sequence data storage means, and comparing the sequence data with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data Pattern finding means, and sequence pattern output means for outputting a sequence pattern whose number of occurrences counted by the sequence pattern finding means is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result are provided.

  According to the first, sixth, and seventh aspects of the present invention, the sequence data that is the target of sequence pattern extraction is stored, and an arbitrary item value that allows arbitrary item values and a sequence pattern that includes only item values are used as generalized patterns. Define and store the sequence data stored in sequence, count the number of occurrences of generalized patterns that match the sequence data, extract generalized patterns whose count is equal to or greater than a predetermined threshold, and extract the general By setting a selection item value that selects an allowable value to an arbitrary item value included in the conversion pattern and sequentially generating a series pattern, and further sequentially reading stored series data and collating with the series pattern, Count the number of occurrences of the sequence pattern that matches the sequence data, and output the sequence pattern whose counted occurrence number is equal to or greater than a predetermined threshold as the sequence pattern of the extraction result First, the number of occurrences of the sequence pattern is counted under rough conditions, and after narrowing down with a predetermined threshold, the number of occurrences of the sequence pattern can be counted under more detailed conditions. Is efficiently counted, so that the processing time required to extract the sequence pattern can be shortened.

  According to the invention of claim 2, the generalized pattern table in which the generalized pattern number for uniquely identifying the generalized pattern is associated with the generalized pattern counter for counting the number of appearances of the generalized pattern is provided. In addition, it generates a generalized map that is a map for deriving a generalized pattern number from the sequence data, reads out the stored sequence data sequentially, and generalizes it using the generalized map from the read sequence data Since the pattern number is derived and the generalized pattern counter corresponding to the derived generalized pattern number in the generalized pattern table is configured to be counted up, the series data is generated each time a series pattern group that may occur frequently is generated. Each time a generalized map is generated, instead of reading all By reading all, the number of occurrences of all the generalized pattern expressed by the same mapping is counted collectively. As a result, the generalized pattern finding that appears frequently only by scanning the series data a predetermined number of times without being affected by the explosion of the number of candidate patterns is completed, so that the processing time is not affected by the threshold value.

  According to the invention of claim 3, a generalized pattern in which the counted number of appearances is equal to or greater than a predetermined threshold is extracted, and a selection item value is set as one of the arbitrary item values included in the extracted generalized pattern The sequence pattern is sequentially generated, and the sequence pattern thus generated and the stored sequence data are collated to count the number of occurrences of the sequence pattern that matches the sequence data. A series pattern that is equal to or greater than the threshold value is extracted, and the series pattern is further refined by setting a selection item value to one of the arbitrary item values included in the extracted series pattern. By matching, the number of occurrences of the series pattern that matches the series data is counted, and the selection item value is set for all arbitrary item values included in the series pattern. Since the detail of the sequence pattern and the counting of the number of occurrences of the sequence pattern are repeated, the number of occurrences of the generalized pattern is first counted, and then the generalized pattern is narrowed down using a predetermined threshold, and then narrowed down. In addition, one of the optional item values included in the generalized pattern is replaced with the selected item value to generate the sequence pattern sequentially, and further, the series pattern is detailed step by step by replacing the arbitrary item value with the selected item value one by one Each time, it is possible to eliminate a series pattern that does not need to be counted by using a predetermined threshold, efficiently counting the number of occurrences of the series pattern, and reducing the processing time required to extract the series pattern. There is an effect that can be done.

  According to the invention of claim 4, the detailed pattern table in which the detailed pattern number for uniquely identifying the detailed pattern is associated with the detailed pattern counter for counting the number of appearance of the detailed pattern is provided. In addition, a detailed mapping, which is a map for deriving the detailed pattern number from the sequence data, is generated, the stored sequence data is read sequentially, and the generalized map number is first applied by applying the generalized map. The detailed pattern table corresponding to the generalized pattern number is obtained by referring to the generalized pattern table. Subsequently, the detailed mapping is applied to the series data to obtain a detailed pattern number, and by referring to the detailed pattern table, a detailed pattern counter corresponding to the detailed pattern number is obtained and counted. Because each time a series pattern group that may occur frequently is generated, it is not necessary to read out all the series data and count the number of occurrences of the series pattern, but to generate a generalized map and a detailed map each time. By reading all the series data, the number of appearances of all the detailed patterns represented by the same mapping set is counted in a lump. As a result, the generalized pattern finding that appears frequently only by scanning the series data a predetermined number of times without being affected by the explosion of the number of candidate patterns is completed, so that the processing time is not affected by the threshold value.

  Further, according to the invention of claim 5, since the continuously changing values are discretized, the series data in which the discretized intervals are arranged in order as item values is stored as a pattern extraction reference. For example, even for item values that change continuously, such as stock prices and product sales, the values are sampled by discretizing them at predetermined intervals, and the sampled values are used as series data, and frequently change There is an effect that the appearing series pattern can be extracted.

  Exemplary embodiments of a sequence pattern discovery program, a sequence pattern discovery method, and a sequence pattern discovery apparatus according to the present invention will be described below in detail with reference to the accompanying drawings. In this embodiment, a case where a stock price series pattern is found from stock price time series data will be described.

  First, the concept of the sequence pattern finding apparatus according to the first embodiment will be described. FIG. 1 is a diagram illustrating an example of sequence data from which a sequence pattern is extracted. The figure shows time-series data in which the closing prices of stock prices for a week for a certain security are arranged for each day. The “numerical data” shown in the figure is the closing price of the daily stock price, and the “increase / decrease from the previous day” is the difference between the closing price of the stock price and the previous day. The “discretized data” is a numerical value obtained by discretizing the difference in the closing price of the stock price from the previous day in five stages. When the difference is −100 or more, “0 (significantly low)” is changed from −99. If it is between -20, "1 (small reduction)", if it is between -19 and +19, "2 (no change)", if it is between 20 and 99, "3" When (small width and height) is 100 or more, “4 (significant height)” is set. The series data whose item value is this “discretized data” is expressed as, for example, <3 1 3 1 4 0>.

  In the first embodiment, a case will be described in which a series pattern is extracted from a series database in which series data having “discretized data” described above as an item value is stored for each security.

  First, the sequence pattern to be extracted is defined. For example, the sequence pattern <3 {0, 1, 2, 3, 4} 3 1 {0, 1, 2, 3, 4} 0> is set to the second item and the fifth item value. {0, 1, 2, 3, 4} represents an arbitrary item value that allows an arbitrary item value, and the second item value of the sequence data when this sequence pattern is compared with the sequence data And the fifth item may be any of “0”, “1”, “2”, “3”, “4”. Such a series pattern including only a predetermined number of arbitrary item values and a predetermined number of item values is referred to as a generalized pattern.

Here, when the length of the sequence pattern to be extracted is 1 and the number of arbitrary item values included in the sequence pattern to be extracted is k, the generalized pattern T _M (l, k) described above is expressed by the following formula (1). ).

In the above formula (1), Σ is an item value that can be set to the series data to be extracted (here, item values that can be set to “discretized data”, that is, “0”, “1”, “2”, “3”, “4”), and Π _M indicates an arbitrary item value.

  For example, the sequence pattern <3 {0, 1} 3 1 {3, 4} 0> is set to the second item and the fifth item value {0, 1} or {3,4} Represents a selection item value for selecting an allowable item value including {2}. When this series pattern is compared with the series data, the second item value of the series data is “0”, “1”. It may be any one, indicating that the fifth item value of the series data may be either “3” or “4”. Such series data including only a predetermined number of selection item values and a predetermined number of item values is referred to as a refinement pattern.

The detailed pattern T _m (l, k) is expressed by the following mathematical formula (2).

In the above equation (2), the [pi _m, shows a selection value. The role of the selection item value is to limit the range of fluctuation in more detail than the arbitrary item value. For example, by setting {0, 1}, it is possible to specify a pattern in which the closing price of the stock price fluctuated to a low price regardless of the size. By specifying {2}, the closing price of the stock price did not change. A pattern can be specified, and by specifying {3, 4}, a pattern in which the closing price of the stock price has changed to a high value can be specified regardless of the size.

  A sequence pattern that is sequentially generated is defined by replacing one or more arbitrary item values among the arbitrary item values included in the generalized pattern with selected item values. For example, a detailed sequence pattern <3 {0, 1} 3 1 {0, 1, 2, 3, 4} 0>, and such a predetermined number of arbitrary item 巷 is a selected item value and a predetermined value. A sequence pattern including a number of item values is simply called a sequence pattern.

  The sequence pattern discovered by the sequence pattern discovery apparatus according to the first embodiment is expressed by the following formula (3).

It should be noted that the length l of the sequence pattern to be extracted, the number k of arbitrary item values included in the sequence pattern to be extracted, the item value Σ, the arbitrary item value Π _M , the selection item value that can be set in the sequence data to be extracted Assume that the set Π _m and the predetermined threshold value are set in advance by the user.

  In the first embodiment, the case where the values set in the item values of the series data to be extracted are “0” to “4” has been described, but the present invention is not limited to this. For example, when the item value set of the sequence data to be extracted is represented by Σ, the generalized pattern, the detailed pattern, and the sequence pattern discovered by the sequence pattern finding device are expressed by the following formulas (4), (5), and (6). ).

  Next, the configuration of the sequence pattern finding apparatus according to the first embodiment will be described. FIG. 2 is a functional block diagram illustrating the configuration of the sequence pattern finding apparatus according to the first embodiment. As shown in the figure, the sequence pattern finding device 100 includes a sequence data storage unit 110, a template generation unit 120, a mapping generation unit 130, a sequence pattern counting unit 140, a sequence pattern discovery unit 150, a sequence pattern, and the like. And an output unit 160.

  The sequence data storage unit 110 is a storage unit that stores sequence data from which a sequence pattern is to be extracted. In the first embodiment, the series data storage unit 110 stores the “discretized data” shown in FIG. 1 for each security.

Here, the values of “discretized data” 6 days ago, 5 days ago, 4 days ago, 3 days ago, 2 days ago, and 1 day ago of the series data D _E stored by the series data storage unit 110 are respectively expressed as d ₅ , d ₄ , d ₃ , d ₂ , d ₁ , and d ₀ , and the series data D _E is expressed by the following formula.

D _E = <d ₅ d ₄ d ₃ d ₂ d ₁ d ₀ >

  For example, the series data of “discretized data” shown in FIG. 1 is expressed by the following equation.

D _E = <3 1 3 1 4 0>

  In addition, although the example which discretizes the amount of difference between the day before the closing price of the stock price and the current day has been described here, a more detailed change in the difference width is shown by setting the level of discretization to five or more steps. The sequence pattern may be found from the sequence data, or the sequence pattern may be found from the sequence data that shows a coarse change in the difference width by setting the discretization step to five or less steps.

  In this way, the series data storage unit 110 discretizes continuously changing fluctuation elements, and stores the series data in which values indicating the fluctuation ranges for the discrete units are arranged as series pattern extraction targets, It is possible to extract a series pattern that frequently appears in a continuously changing variable element such as a stock price or a sales amount of a product.

  The template generation unit 120 generates a template for defining a sequence pattern based on the length l of the sequence pattern specified by the user and the number k of arbitrary item values or selection item values included in the extracted sequence pattern. A processing unit to be generated. Here, the template is a definition representing an item structure of a series pattern including a predetermined number of arbitrary item values or selection item values. The arbitrary item value or selection item value and a specific item value (hereinafter simply referred to as a template item). Item values) are combined in a predetermined order. Hereinafter, a template generation method by the template generation unit 120 will be described.

When the length of the sequence pattern to be extracted designated by the user is l and the number of arbitrary item values or selection item values included in the extracted sequence pattern is k, the template generation unit 120 has k arbitrary item values. Alternatively, _l C _k pattern templates X are generated by combining selected item values and l−k item values in order.

  For example, the above-mentioned “discretized data” series data

D _E = <d ₅ d ₄ d ₃ d ₂ d ₁ d ₀ >

Is a pattern pattern extraction target, the length l of the sequence pattern to be extracted is set to “6”, and the number k of arbitrary item values or selection item values included in the sequence pattern to be extracted is set to “2”. X is generated in ₆ C ₂ (= 15) ways.

A template X _E that is one of the 15 templates X generated here is expressed by, for example, the following expression.

X _E = <x ₅ * x ₃ x ₂ * x ₀ >

  In the above formula, “*” represents an arbitrary item value or a selected item value.

  Here, the generation of the template when the length l of the sequence pattern to be extracted is “6” and the number k of arbitrary item values or selection item values included in the sequence pattern to be extracted is “2” has been described. However, the present invention is not limited to this, and as described above, the template is appropriately generated according to the values set in l and k.

  The mapping generation unit 130 generates a mapping (hereinafter referred to as a generalized mapping) for deriving a generalized pattern number from series data for each template based on the template generated by the template generation unit 120. It is. Here, the generalized pattern number is an identification number that uniquely identifies the generalized pattern defined by the pattern template. Hereinafter, a method for generating a generalized map by the map generation unit 130 will be described.

The map generation unit 130 first acquires the template X generated by the template generation unit 120, and generates a generalized map f _X (D) based on the acquired template X.

For example, a generalized map f _XE generated based on the template X _E shown below is defined by the following expression.

In the above formula, (3), (2), (1), (0) are serial numbers assigned in order from right to left ignoring the symbol “*” representing an arbitrary item value or selection item value. , D ⁽³⁾ , d ⁽²⁾ , d ⁽¹⁾ , d ⁽⁰⁾ are x ⁽³⁾ , x ⁽²⁾ , x ⁽¹⁾ , x ⁽⁰⁾ defined in the template X _E , respectively. The item values of the series data D _{E at} the same position, that is, d ₅ , d ₃ , d ₂ , d ₀ are shown. In the above formula, 5 ^j (j = 3, 2, 1, 0) multiplied by the item values d ₅ , d ₃ , d ₂ , and d ₀ is “5”, which is the series data to be extracted. Is determined by the number of item values Σ that can be set to 5 (here, “0”, “1”, “2”, “3”, “4”), and the index j is a specific value in the template X _E The serial number ((3), (2), (1), (0)) given to each item value is determined.

With this generalized map f _XE (D _E ), for example, the sequence data is

D _E = <3 1 3 1 4 0>

If it was

f _XE (D _E ) = 3 × 5 ³ + 3 × 5 ² + 1 × 5 ¹ + 0 × 5 ⁰
= 455

Thus, the generalized pattern number is “455”.

  Here, the length l of the sequence pattern to be extracted is “6”, the number k of arbitrary item values or selection item values included in the sequence pattern to be extracted is “2”, and items that can be set in the sequence data to be extracted The generation of the generalized map when the number of values Σ is “5” has been described. However, the present invention is not limited to this, and as described above, the generalized map is represented by l, k, and Σ. It is generated appropriately according to the set value.

The sequence pattern counting unit 140 reads the sequence data stored in the sequence data storage unit 110 and uses the generalized mapping generated by the mapping generation unit 130 to calculate the number of occurrences of the generalized pattern that matches the read sequence data. A processing unit for counting. Specifically, the series pattern counting unit 140 includes a pattern table in which generalized pattern numbers and pattern counters are associated with the template X generated by the template generation unit 120. Here, the generalized pattern number corresponds to the generalized pattern number derived from the above-described generalized map f _X (D), and is a serial number starting from 0. The pattern counter is a counter for counting the number of appearances of generalized patterns that match the sequence data stored in the sequence data storage unit 110 for each generalized pattern number.

  FIG. 3 is a diagram illustrating an example of a pattern table used by the pattern counting unit 140 according to the first embodiment. For example, as shown in FIG.

X _E = <x ₅ * x ₃ x ₂ * x ₀ >

In this pattern table, 5 ⁴ (= 625) pattern numbers, that is, pattern numbers “0” to “624” are associated with the pattern counter. The number of pattern numbers and pattern counters is the same as the number of generalized patterns generated from the template XE.

The sequence pattern counting unit 140 initializes the pattern counter of the pattern table (sets “0”) and then sequentially reads the sequence data D stored in the sequence data storage unit 110 and generates the general data generated by the mapping generation unit 130. The generalized pattern number is derived using the generalized map f _X (D), and “1” is added to the pattern counter of the pattern table corresponding to the derived generalized pattern number.

  For example, the series data mentioned above

D _E = <3 1 3 1 4 0>

Is read out,

f _XE (D _E ) = 455

Thus, the series pattern counting unit 140 adds “1” to the pattern counter corresponding to the generalized pattern number “455” in the pattern table (count 1 shown in FIG. 3).

  In this way, the sequence pattern counting unit 140 adds a pattern counter using mapping to all the templates generated by the template generation unit 120.

  The details of the generalized pattern counting process by the series pattern counting unit 140 will be described later with reference to FIG.

  The sequence pattern finding unit 150 extracts a generalized pattern whose pattern counter counted by the sequence pattern counting unit 140 is equal to or greater than a predetermined threshold, and selects one of the arbitrary item values included in the extracted generalized pattern as a selection item value To generate a series pattern, and further refine the series pattern step by step by replacing the optional item value with the selected item value one by one, while eliminating the series pattern that does not need to be counted each time. It is a processing unit that counts the number of appearances of patterns.

  For example, when the generalized pattern <3 * 3 1 * 0> is refined using the selection item values {0, 1}, {2} and {3,4], the sequence pattern finding unit 150 The following six series patterns are generated.

<3 {0, 1} 3 1 * 0>,
<3 {2} 3 1 * 0>,
<3 {3, 4} 3 1 * 0>,
<3 * 3 1 {0, 1} 0>,
<3 * 3 1 {2} 0>,
<3 * 3 1 {3,4} 0>

  The sequence pattern finding unit 150 then expands the generated sequence pattern and the sequence data stored in the sequence data storage unit 110 with, for example, a Shift-AND method that has been extended to enable matching with a plurality of patterns. By matching using a character string matching algorithm, the number of occurrences of a sequence pattern that matches the sequence data is counted.

  Although the case where the Shift-AND method is used as the character string matching algorithm has been described here, the character string matching performed here is not limited to the Shift-AND method, and a plurality of values including arbitrary item values and selection item values are included. Other known character collation algorithms capable of collating the sequence pattern and the sequence data may be used.

  FIG. 4 is a diagram illustrating an example of a sequence pattern discovery process performed by the sequence pattern discovery unit 150. For example, as shown in the figure, the number of occurrences of <3 {0, 1} 3 1 * 0> is “50”, the number of occurrences of <3 {2} 3 1 * 0> is “10”, <3 { 3,4} 3 1 * 0> is “20”, <3 * 3 1 {0,1} 0> is “30”, <3 * 3 1 {2} 0> Is “10”, the number of occurrences of <3 * 3 1 {3,4} 0> is “10”, and the predetermined threshold specified by the user is “20”.

  In this case, the sequence pattern discovery unit 150, as shown in FIG. 4, <3 {0, 1} 3 1 * 0>, <3 {3, 4} 3 1 * 0>, <3 * 3 1 {0,1} 0> is extracted and multiplied by <3 {0,1} 3 1 * 0> and <3 * 3 1 {0,1} 0> (pair 1 shown in FIG. 4) ), Further refined refinement pattern <3 {0, 1} 3 1 {0, 1} 0> is defined, and <3 {3,4} 3 1 * 0> and <3 * 3 1 By multiplying {0,1} 0> (pair 2 shown in FIG. 4), a further refined refinement pattern <3 {3,4} 3 1 {0,1} 0> is defined. Since the number of occurrences of series patterns decreases monotonously by the procedure of replacing arbitrary item values with selection item values, the number of occurrences of series patterns obtained by replacement with other selection item values is less than the threshold value. It is possible to exclude it without counting the number of appearances by collation.

  Then, the sequence pattern discovery unit 150 generates the generated sequence patterns (<3 {0, 1} 3 1 {0, 1} 0> and <3 {3, 4} 3 1 {0, 1} 0>), The sequence data stored in the sequence data storage unit 110 is collated again using, for example, a character string collation algorithm such as the extended Shift-AND method, and the number of occurrences of the sequence pattern matching the sequence pattern is counted together.

  Details of the sequence pattern finding process by the sequence pattern finding unit 150 will be described later with reference to FIG.

  As described above, the sequence pattern finding unit 150 reads the sequence data stored in the sequence data storage unit 110 and uses the character string matching algorithm based on the extended Shift-And method, and the number of occurrences of the sequence pattern that matches the sequence data. First, count the number of occurrences of the generalized pattern, narrow down the generalized pattern using a predetermined threshold, and then replace one of the optional item values included in the narrowed down generalized pattern with the selected item value The series pattern is generated step by step, and the series pattern is refined step by step by replacing the optional item value with the selected item value one by one. Can be eliminated, and the number of occurrences of the sequence pattern is counted efficiently to extract the sequence pattern. It is possible to reduce the management time.

  The pattern output unit 160 extracts a sequence pattern in which the number of appearances counted by the sequence pattern discovery unit 150 is equal to or greater than a predetermined threshold, and uses the extracted sequence pattern as an extraction result sequence pattern, for example, to an output device such as a display device. It is a processing part to output.

  Next, a processing procedure of the generalized pattern counting process performed by the sequence pattern counting unit 140 according to the first embodiment will be described. FIG. 5 is a flowchart of the generalized pattern counting process performed by the sequence pattern counting unit 140 according to the first embodiment. As shown in the figure, first, the template generation unit 120 generates one template X (step S101).

When the template X is generated (step S102, Yes), the map generation unit 130 generates a map f _X (D) for calculating the generalized pattern number from the series data D based on the template X ( Step S103).

  Then, the series pattern counting unit 140 clears all the pattern counters in the pattern table (step S104). After clearing the pattern table, the sequence pattern counting unit 140 extracts one sequence data D from the sequence data storage unit 110 (step S105).

When the sequence data D can be extracted (step S106, No), the sequence pattern counting unit 140 uses the mapping f _X (D) generated by the mapping generation unit 130 to use the generalized pattern number from the extracted sequence data D. Is calculated (step S107), and the pattern counter corresponding to the calculated generalized pattern number is incremented (step S108). Thereafter, the sequence pattern counting unit 140 returns to step S105, retrieves the next sequence data D from the sequence data storage unit 110, and performs the processes of steps S107 and S108. The sequence pattern counting unit 140 repeats these processes until the sequence data D cannot be extracted from the sequence data storage unit 110.

  When the sequence data D cannot be extracted from the sequence data storage unit 110, that is, when the extraction of all the sequence data D stored in the sequence data storage unit 110 has already been completed, the sequence pattern counting unit 140 Returning to step S101, the next template X is generated, and the processes after step S102 are repeated until all the templates X are generated.

  Then, after generating all the templates X, the sequence pattern counting unit 140 ends the generalized pattern counting process (No in step S102).

  As described above, the mapping generation unit 130 generates a generalized map for calculating the generalized pattern number from the sequence data in advance, and the sequence pattern counting unit 140 sequentially reads the sequence data stored in the sequence data storage unit 110. In addition, by calculating the generalized pattern number using the generalized map, the number of appearances of the sequence pattern (generalized pattern) can be counted efficiently.

  In this way, the sequence pattern counting unit 140 includes a generalized pattern table that associates a generalized pattern number that uniquely identifies a generalized pattern with a pattern counter for counting the number of appearances of the generalized pattern, Map generation unit 130 generates a generalized map that is a map for deriving the generalized pattern number from the sequence data, and sequence pattern counting unit 140 sequentially reads the sequence data stored in sequence data storage unit 110. At the same time, the generalized pattern number is derived from the read sequence data using the generalized map, and the generalized pattern counter corresponding to the derived generalized pattern number in the generalized pattern table is counted up. Every time a certain series pattern group is generated, all the series data is read and the generalized pattern Instead of counting the number of appearances, each time to produce a generalized mapping by reading all the series data, the number of occurrences of all series pattern represented by the same mapping is counted collectively. As a result, the generalized pattern finding that appears frequently only by scanning the series data a predetermined number of times without being affected by the explosion of the number of candidate patterns is completed, so that the processing time is not affected by the threshold value.

  Next, a processing procedure of sequence pattern discovery processing by the sequence pattern discovery unit 150 will be described. FIG. 6 is a flowchart showing a processing procedure of sequence pattern discovery processing by the sequence pattern discovery unit 150. As shown in the figure, the sequence pattern finding unit 150 first acquires a generalized pattern P whose pattern counter counted by the sequence pattern counting unit 140 is equal to or greater than a predetermined threshold (step S201).

If the generalized pattern P can be acquired (step S202, Yes), the sequence pattern finding unit 150 sets “0” to the variable i (step S203) and includes only one acquired generalized pattern P. a set and held as a sequence pattern set L ₀ (step S204), further, all series data that match the general pattern P (hereinafter referred to as series data D _P) is extracted from the time series data storage unit 110 (step S205 ).

Thereafter, the sequence pattern finding unit 150 adds “1” to the variable i (step S206), and when the variable i is smaller than the number k of arbitrary item values or selection item values included in the generalized pattern (step S207). , Yes), a series pattern is generated by setting the selection item value to any one of the arbitrary item values included in the held series pattern set L _i-1 (step S208). Here, for example, when there are three types of selection item values, | Π _m | × (k−i + 1) sequence patterns are generated for each sequence pattern belonging to the set Li−1.

The sequential pattern discovery unit 150, a generated sequence pattern, and a series data D _P, for example, by matching with a string matching algorithm such as extended Shift-the AND method, it summarizes the number of occurrences counted (Step S209).

The sequence pattern mining unit 150, the number of occurrences counted extracts a sequence pattern which exceeds a predetermined threshold, the extracted frequent sequence pattern holds as a sequence pattern set L _i (step S210).

Thereafter, the sequence pattern finding unit 150 counts the variable i while counting up the variable i until the variable i becomes equal to or greater than the number k of arbitrary item values or selection item values included in the generalized pattern (step S207, No). The extraction of the set L _i is repeated.

When the variable i becomes equal to or larger than the number k of arbitrary item values or selection item values included in the generalized pattern, the sequence pattern finding unit 150 extracts the sequence pattern sets L _{0 to} L _k−1 extracted so far. Are stored as a set L (P) (step S211).

  Then, the sequence pattern finding unit 150 returns to step S201 to further acquire a generalized pattern whose number of appearances is equal to or greater than a predetermined threshold, and finishes acquiring all generalized patterns whose number of appearances is equal to or greater than the predetermined threshold. Until then, the processing from step S201 is repeated.

  Then, after processing all the generalized patterns, the sequence pattern finding unit 150 ends this sequence pattern finding process, and outputs all the stored sequence pattern sets L (P) as extraction results as a pattern output. The data is output via the unit 160 (step S212).

  As described above, in the first embodiment, the sequence data storage unit 110 stores the sequence data that is the target of sequence pattern extraction, and the template generation unit 120 allows arbitrary item values and item values that allow arbitrary values. Is defined as a generalized pattern, and the sequence pattern counting unit 140 sequentially reads out the stored sequence data and sets the number of sequence data matching each generalized pattern as the generalized pattern appearance number. Then, the sequence pattern finding unit 150 extracts the generalized pattern whose counted generalized pattern appearance number is equal to or greater than a predetermined threshold, and selects the selected item value as an arbitrary item value included in the extracted generalized pattern. Is defined, and the stored sequence data is read sequentially, and the number of sequence data that matches each sequence pattern is determined. Since the sequence pattern output unit 160 counts as the number of column pattern appearances and outputs the sequence pattern whose sequence pattern appearance number counted by the sequence pattern discovery unit 150 is equal to or greater than a predetermined threshold as an extraction result. It is possible to count the number of appearances of generalized patterns under conditions, narrow down with a predetermined threshold, and count the number of appearances of series patterns under more detailed conditions. Thus, by efficiently counting the number of appearances of the generalized pattern, the processing time required for extracting the sequence pattern can be shortened.

  Further, the mapping generation unit 130 generates a generalized map for deriving a generalized pattern number for uniquely identifying the generalized pattern by converting the item value included in the sequence data, and the sequence pattern counting unit 140 Generating a pattern table in which a generalized pattern, a generalized pattern number derived using a generalized mapping, and a pattern counter that counts the number of series data that matches the generalized pattern are associated with each generalized pattern; Since the sequence pattern counting unit 140 sequentially reads the sequence data, derives the generalized pattern number using the generalized mapping, and counts the pattern counter of the pattern table based on the derived generalized pattern number. As a result, it is possible to easily determine the generalized pattern that matches the read series data. By efficiently count the number of occurrences of the general pattern, it is possible to shorten the processing time required for the extraction of the generalized pattern.

  By the way, in the first embodiment, series data is read, the number of appearances of each generalized pattern is counted using a generalized map generated for each generalized pattern based on a predetermined set value, and the number of appearances is a predetermined threshold value. After extracting the above generalized pattern, generate a series pattern that refines the generalized pattern by setting the selection item value to an arbitrary item value, read the series data again, and count the number of occurrences of the series pattern In the above, the case where frequent sequence patterns are found by narrowing down the sequence pattern by a predetermined threshold has been explained. However, a generalized map is generated at the same time as a detailed pattern map, and these two maps are combined. By using them, only the number of occurrences of detailed patterns may be counted, and only frequent detailed patterns may be found.

  Therefore, in the second embodiment, a case will be described in which only the number of appearances of the detailed pattern is counted by using two maps in combination.

  First, the configuration of the sequence pattern finding apparatus according to the second embodiment will be described. Here, for convenience of explanation, functional units that play the same functions as the respective units shown in FIG. FIG. 7 is a functional block diagram of the configuration of the sequence pattern finding apparatus according to the second embodiment. As shown in the figure, the sequence pattern finding apparatus 200 includes a sequence data storage unit 110, a template generation unit 120, a mapping generation unit 230, a sequence pattern counting unit 240, and a sequence pattern output unit 160.

  Based on the template generated by the template generation unit 120, the mapping generation unit 230 generates a mapping (hereinafter referred to as a generalized mapping) for calculating a generalized pattern number from the sequence data, and a predetermined selection item value. Is a processing unit that generates a map (hereinafter referred to as a detailed map) for calculating a detailed pattern number. Here, the generalized pattern number is an identification number for uniquely identifying the generalized pattern defined by the template, and the detailed pattern number is a detail of the generalized pattern based on a predetermined selection item value. This is an identification number for uniquely identifying the refinement pattern obtained by making the data. Note that the generalized map generated here is the same as the generalized map described in the map generation unit 130 described in the first embodiment, and thus the description thereof is omitted. Here, a detailed map generation method is described. .

The mapping generation unit 230 first acquires the template X generated by the template generation unit 120, generates the generalized mapping f _X (D) based on the acquired template X, and then the detailed mapping g _X (D). Is generated.

For example, a _refined map g _XE (D) generated based on the template X _E shown below is defined by the following equation.

In the above formula (X _E formula), Σ represents that any one of “0”, “1”, “2”, “3”, “4” is set (formula (2)). (See Σ). [1] and [0] are serial numbers starting from 0 assigned to the symbol “*” representing an arbitrary item value or a selected item value in order from right to left, and the above g _XE (D _E ) D ^[1] and d ^{[0] in} the equation ⁽¹⁾ are the item values of the series data D _{E at} the same positions as * ^[1] and * ^[0] defined in the template X _E , that is, d ₄ , It shows d _1.

Further, the above m (x) formula is for assigning a serial number starting from 0 to the item value x according to the selection item value (see Equation (2)) included in the selection item value set Π _m . When the item value x is included in the selection item value {0, 1}, m (x) = “0”, and when x is included in {2}, m (x) = “1”. , X is included in {3, 4}, it indicates that m (x) = “2”. For example, when x = d ₄ = 1, m (d ₄ ) = 0, and when x = d ₁ = 4, m (d ₁ ) = 2.

In the above equation of g _XE (D _E ), “3” is the selected item value for 3 ^j (j = 1, 0) multiplied by the item values m (d ₄ ) and m (d ₁ ). Is determined by the number of selection item values (three of {0, 1}, {2}, {3, 4}) included in the set Π _m (see Π _m in equation (2)), and the index j is , serial number granted to the symbol "*" representing any item values or selection items value in template _{X E ([1], [} 0]) is determined by.

With this mapping example g _XE (D _E ), for example, the series data is

D _E = <3 1 3 1 4 0>

If it was

g _XE (D _E ) = 0 × 3 ¹ + 2 × 3 ⁰
= 2

Thus, the detailed pattern number is “2”.

Here, the length l of the sequence pattern to be extracted is “6”, the number k of arbitrary item values or selection item values included in the sequence pattern to be extracted is “2”, and is included in the set 選 択_m of selection item values. Although the generation of the detailed mapping when the number of selection item values is “3” has been described, the present invention is not limited to this, and as described above, the detailed mapping includes l, k, and Π. _According to the value set in _m , it is generated appropriately.

The sequence pattern counting unit 240 reads the sequence data stored in the sequence data storage unit 110, and uses the generalized map and the detailed map generated by the map generation unit 230 to use the detailed pattern that matches the read sequence data. It is a processing part which counts the appearance number of. Specifically, the series pattern counting unit 240 generates a pattern table in which the generalized pattern number and the detailed pattern reference pointer are associated with each other for each template X generated by the template generation unit 120. Here, the detailed pattern reference pointer is a pointer for associating a generalized pattern number with a detailed pattern table to be described later. Here, the generalized pattern number corresponds to the generalized pattern number derived from the above-described generalized map f _X (D), and is a serial number starting from 0. The pattern counter is a counter for counting the number of appearances of detailed patterns in the sequence data stored in the sequence data storage unit 110 for each set of generalized pattern numbers and detailed pattern numbers.

  For example, the template mentioned above

X _E = <x ₅ * x ₃ x ₂ * x ₀ >

In this pattern table, 5 ⁴ (= 625) pattern numbers, that is, pattern numbers “0” to “624” are associated with the pattern counter. Then, the number of such pattern number and the pattern counter is a pattern as many generalized pattern generated from the template X _E.

  Further, the series pattern counting unit 240 includes a detailed pattern table in which a detailed pattern number and a detailed pattern counter are associated with each generalized pattern number set in the pattern table. The detailed pattern table is associated with each generalized pattern number of the pattern table by the above-described detailed pattern reference pointer.

Here, the detailed pattern number corresponds to the detailed pattern number derived from the above-described detailed map g _XE (D _E ), and is a serial number starting from 0. The detailed pattern counter is a counter for counting the number of appearances of the detailed pattern in the sequence data stored in the sequence data storage unit 110 for each detail pattern number.

  FIG. 8 is a diagram illustrating an example of a detailed pattern table used by the sequence pattern counting unit 240 according to the second embodiment. For example, as shown in FIG.

X _E = <x ₅ * x ₃ x ₂ * x ₀ >

In the refined pattern table for the pattern field tool 455 included in the pattern table for, 3 ² (= 9) refined pattern numbers, that is, refined pattern numbers of “0” to “8”, and the refined pattern counter Are associated with each other. Then, the number of such details pattern number and detailed patterns counter, a pattern as many details of patterns generated from the template X _E.

The sequence pattern counting unit 240 initializes the detail pattern counter of the detail pattern table (sets “0”), then sequentially reads the sequence data D stored in the sequence data storage unit 110, and the mapping generation unit 230 The refined pattern number is derived using the generated generalized map f _X (D) and the refined map g _X (D), and the refined pattern counter of the refined pattern table corresponding to the derived refined pattern number is used. Add “1”.

  For example, the series data mentioned above

D _E = <3 1 3 1 4 0>

Is read out,

f _XE (D _E ) = 455, g _XE (D _E ) = 2

Thus, the sequence pattern counting unit 240 refers to the refined pattern reference pointer corresponding to the generalized pattern number “455” in the pattern table, and refines the refined pattern table associated with the pointer. “1” is added to the detailed pattern counter corresponding to the pattern number “2” (count 2 shown in FIG. 3).

  In this way, the pattern counting unit 240 generates the above-described pattern table and detailed pattern table, and the detailed pattern counter using the generalized map and the detailed map for all templates generated by the template generating unit 120. Is added.

  The sequence pattern output unit 160 extracts a detailed pattern whose detailed pattern counter is equal to or greater than a predetermined threshold from the detailed pattern table generated by the sequence pattern counting unit 240, and uses the extracted detailed pattern as an extraction result. For example, it is a processing unit that outputs to an output device such as a display device.

  As described above, in the second embodiment, the mapping generation unit 230 generates a detailed mapping for deriving a detailed pattern number for uniquely identifying the detailed pattern from the item values included in the sequence data, and the sequence pattern The counting unit 240 corresponds to the refinement pattern, the refinement pattern number derived using the refinement mapping, and the refinement pattern counter that counts the number of series data matching the refinement pattern for each refinement pattern. The detailed pattern table attached is cleared, the stored series data is read sequentially, the detailed pattern number is derived using the detailed mapping, and the details of the detailed pattern table based on the derived detailed pattern number Every time a series pattern group that may occur frequently is generated. Instead of counting out and counting the number of occurrences of the sequence pattern, by reading out all the sequence data every time a generalized map and a detailed map are generated, all of the detailed patterns represented by the same set of maps The number of appearances is counted at once. As a result, the generalized pattern finding that appears frequently only by scanning the series data a predetermined number of times without being affected by the explosion of the number of candidate patterns is completed, so that the processing time is not affected by the threshold value.

  In addition, each component of each apparatus illustrated in the first and second embodiments is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  Further, in the first and second embodiments, the sequence pattern finding devices 100 and 200 have been described. However, by realizing the configuration of the sequence pattern finding devices 100 and 200 with software, a sequence pattern finding program having the same function can be obtained. Obtainable. A computer that executes this sequence pattern finding program will be described.

  FIG. 9 is a functional block diagram illustrating the configuration of the computer that executes the sequence pattern finding program according to the first and second embodiments. As shown in the figure, the computer 300 includes a RAM 310, a CPU 320, an HDD 330, a LAN interface 340, an input / output interface 350, and a DVD drive 360.

  The RAM 310 is a memory that stores a program, a program execution result, and the like. The CPU 320 is a central processing unit that reads a program from the RAM 310 and executes the program.

  The HDD 330 is a disk device that stores programs and data, and the LAN interface 340 is an interface for connecting the computer 300 to other computers via the LAN.

  The input / output interface 350 is an interface for connecting an input device such as a mouse or a keyboard and a display device, and the DVD drive 360 is a device for reading / writing a DVD.

  The sequence pattern finding program 311 executed in the computer 300 is stored in the DVD, read from the DVD by the DVD drive 360, and installed in the computer 300.

  Alternatively, this sequence pattern finding program 311 is stored in a database or the like of another computer system connected via the LAN interface 340, read from these databases, and installed in the computer 300.

  The installed sequence pattern finding program 311 is stored in the HDD 330, read out to the RAM 310, and executed as a sequence pattern finding process 321 by the CPU 320.

  Further, in the first and second embodiments, the case where the series pattern is extracted from the time series data in which the closing price of the daily stock price is accumulated has been described. However, the present invention is not limited to this. The same can be applied to the case of extracting the series pattern from the series data such as the sales data and the credit card usage history.

(Supplementary note 1) A sequence pattern finding program for extracting a sequence pattern frequently appearing in the sequence data from the sequence data in which fluctuating item values are arranged in order,
A sequence data storage procedure for storing sequence data to be sequence pattern extracted in a storage device;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored by the sequence data storage procedure is sequentially read and the series data matches A sequence pattern counting procedure for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of occurrences counted by the sequence pattern counting procedure is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Sequence pattern generation, and sequentially reading the sequence data stored by the sequence data storage procedure and collating it with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data Procedure and
A sequence pattern output procedure for outputting a sequence pattern whose number of occurrences counted by the sequence pattern discovery procedure is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding program characterized in that a computer is executed.

(Supplementary Note 2) The sequence pattern counting procedure is as follows:
A generalized pattern table associating a generalized pattern number for uniquely identifying the sequence pattern with a generalized pattern counter for counting the number of occurrences of the generalized pattern;
A generalized map generation procedure for generating a generalized map that is a map for deriving the generalized pattern number from the sequence data;
The sequence data stored by the sequence data storage procedure is sequentially read out, the generalized pattern number is derived from the read sequence data using the generalized mapping, and the derived generalization of the generalized pattern table is derived. A generalized pattern counter counting procedure for counting up the generalized pattern counter corresponding to the pattern number;
The sequence pattern finding program according to appendix 1, wherein the computer is executed.

(Supplementary Note 3) The sequence pattern discovery procedure is as follows:
A generalized pattern in which the number of occurrences counted by the sequence pattern counting procedure is equal to or greater than a predetermined threshold is extracted, and a sequence is set by setting a selection item value as one of the arbitrary item values included in the extracted generalized pattern Generate patterns sequentially,
By comparing the sequentially generated sequence pattern with the sequence data stored by the sequence data storage procedure, the number of occurrences of the sequence pattern matching the sequence data is counted,
Extracting a series pattern in which the counted number of occurrences is equal to or greater than a predetermined threshold, further refine the series pattern by setting a selection item value to one of the arbitrary item values included in the extracted series pattern,
By collating the detailed sequence pattern with the sequence data stored by the sequence data storage procedure, the number of occurrences of the sequence pattern matching the sequence data is counted,
Supplementary note 1 or 2 characterized in that the series pattern detailing and counting of the number of occurrences of the series pattern are repeated until selection item values are set for all arbitrary item values included in the series pattern. Series pattern discovery program.

(Supplementary Note 4) The sequence pattern counting procedure is as follows:
Define a series pattern that includes only the selected item value and the item value that have selected acceptable values as a refinement pattern,
A detailed pattern table in which a detailed pattern number for uniquely identifying the detailed pattern is associated with a detailed pattern counter for counting the number of appearances of the detailed pattern;
A detailed map generation procedure for generating a detailed map that is a map for deriving the detailed pattern number from the item value included in the series data;
The sequence data stored by the sequence data storage procedure is sequentially read out, the detailed pattern number is derived from the read sequence data by using the detailed mapping, and the derived detail of the detailed pattern table is derived. A detailed pattern counter counting procedure for counting up the detailed pattern counter corresponding to the pattern number;
The sequence pattern finding program according to appendix 1 or 2, wherein the program is executed by a computer.

(Supplementary note 5) The sequence data storing procedure includes discretizing continuously changing values, and storing the sequence data in which the values for the discrete intervals are arranged in order as item values, as a reference for pattern extraction. The sequence pattern discovery program according to any one of Supplementary notes 1 to 4.

(Supplementary Note 6) The detailed pattern in which the sequence pattern finding procedure reads the sequence data stored by the sequence data storage procedure and matches the sequence data using a character string matching algorithm based on the extended Shift-And method. 4. The sequence pattern finding program according to appendix 1, 2, or 3, characterized in that the number of occurrences is counted.

(Supplementary note 7) A sequence pattern finding method for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which fluctuating item values are arranged in order,
A sequence data storage step of storing in the storage device the sequence data to be sequence pattern extracted;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored in the series data storage step is sequentially read and the series data is matched. A sequence pattern counting step for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of occurrences counted in the sequence pattern counting step is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Then, the sequence pattern is sequentially generated, and the sequence data stored in the sequence data storing step is sequentially read out and collated with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data. Process,
A sequence pattern output step of outputting a sequence pattern whose number of occurrences counted by the sequence pattern discovery step is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding method characterized by including

(Supplementary note 8) A sequence pattern finding device for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which the fluctuating item values are arranged in order,
Sequence data storage means for storing sequence data to be sequence pattern extracted in a storage device;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored by the series data storage unit is sequentially read and the series data is matched. A sequence pattern counting means for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of appearances counted by the sequence pattern counting unit is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Then, the sequence pattern is sequentially generated, and the sequence data stored in the sequence data storage means is sequentially read out and collated with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data. Means,
Sequence pattern output means for outputting a sequence pattern in which the number of occurrences counted by the sequence pattern finding means is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding device characterized by comprising:

  As described above, the sequence pattern discovery program, the sequence pattern discovery method, and the sequence pattern discovery device according to the present invention provide a sequence pattern that frequently appears in sequence data from sequence data in which fluctuating item values are arranged in order. Is particularly useful when the amount of data to be extracted is enormous.

It is a figure which shows an example of the sequence data used as the extraction target of a sequence pattern. 1 is a functional block diagram illustrating a configuration of a sequence pattern finding apparatus according to a first embodiment. It is a figure which shows an example of the pattern table used by the pattern counting part which concerns on the present Example 1. FIG. It is a figure which shows an example of the detailed pattern count process by the detailed pattern counting part which concerns on the present Example 1. FIG. 7 is a flowchart illustrating a processing procedure of a generalized pattern counting process by a pattern counting unit according to the first embodiment. 6 is a flowchart illustrating a processing procedure of a detailed pattern counting process by a detailed pattern counting unit according to the first embodiment. It is a functional block diagram which shows the structure of the sequence pattern discovery apparatus which concerns on the present Example 2. It is a figure which shows an example of the detailed pattern table used by the pattern counting part which concerns on the present Example 2. FIG. FIG. 3 is a functional block diagram illustrating a configuration of a computer that executes a sequence pattern finding program according to the first and second embodiments.

Explanation of symbols

100, 200 Sequence pattern discovery device 110 Sequence data storage unit 120 Pattern template generation unit 130, 230 Mapping generation unit 140, 240 Pattern counting unit 150 Detailed pattern counting unit 160, 260 Pattern output unit 300 Computer 310 RAM
311 Sequence Pattern Discovery Program 320 CPU
321 Sequence pattern discovery process 330 HDD
340 LAN interface 350 I / O interface 360 DVD drive

Claims (7)

A sequence pattern discovery program for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which the fluctuating item values are arranged in order,
A sequence data storage procedure for storing sequence data to be sequence pattern extracted in a storage device;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored by the sequence data storage procedure is sequentially read and the series data matches A sequence pattern counting procedure for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of occurrences counted by the sequence pattern counting procedure is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Sequence pattern generation, and sequentially reading the sequence data stored by the sequence data storage procedure and collating it with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data Procedure and
A sequence pattern output procedure for outputting a sequence pattern whose number of occurrences counted by the sequence pattern discovery procedure is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding program characterized in that a computer is executed. The sequence pattern counting procedure includes:
A generalized pattern table associating a generalized pattern number for uniquely identifying the sequence pattern with a generalized pattern counter for counting the number of occurrences of the generalized pattern;
A generalized map generation procedure for generating a generalized map that is a map for deriving the generalized pattern number from the sequence data;
The sequence data stored by the sequence data storage procedure is sequentially read out, the generalized pattern number is derived from the read sequence data using the generalized mapping, and the derived generalization of the generalized pattern table is derived. A generalized pattern counter counting procedure for counting up the generalized pattern counter corresponding to the pattern number;
The sequence pattern finding program according to claim 1, wherein: The sequence pattern discovery procedure includes:
A generalized pattern in which the number of occurrences counted by the sequence pattern counting procedure is equal to or greater than a predetermined threshold is extracted, and a sequence is set by setting a selection item value as one of the arbitrary item values included in the extracted generalized pattern Generate patterns sequentially,
By comparing the sequentially generated sequence pattern with the sequence data stored by the sequence data storage procedure, the number of occurrences of the sequence pattern matching the sequence data is counted,
Extracting a series pattern in which the counted number of occurrences is equal to or greater than a predetermined threshold, further refine the series pattern by setting a selection item value to one of the arbitrary item values included in the extracted series pattern,
By collating the detailed sequence pattern with the sequence data stored by the sequence data storage procedure, the number of occurrences of the sequence pattern matching the sequence data is counted,
3. The method according to claim 1, wherein detailing of the sequence pattern and counting of the number of appearances of the sequence pattern are repeated until selection item values are set for all arbitrary item values included in the sequence pattern. The described series pattern discovery program. The sequence pattern counting procedure includes:
Define a series pattern that includes only the selected item value and the item value that have selected acceptable values as a refinement pattern,
A detailed pattern table in which a detailed pattern number for uniquely identifying the detailed pattern is associated with a detailed pattern counter for counting the number of appearances of the detailed pattern;
A detailed map generation procedure for generating a detailed map that is a map for deriving the detailed pattern number from the item value included in the series data;
The sequence data stored by the sequence data storage procedure is sequentially read out, the detailed pattern number is derived from the read sequence data by using the detailed mapping, and the derived detail of the detailed pattern table is derived. A detailed pattern counter counting procedure for counting up the detailed pattern counter corresponding to the pattern number;
3. The sequence pattern finding program according to claim 1, wherein:   The sequence data storing procedure discretizes continuously changing values, and stores the sequence data in which the values for the discretized intervals are sequentially arranged as item values as a reference for pattern extraction. The sequence pattern discovery program as described in any one of 1-4. A sequence pattern finding method for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which fluctuating item values are arranged in order,
A sequence data storage step of storing in the storage device the sequence data to be sequence pattern extracted;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored in the series data storage step is sequentially read and the series data is matched. A sequence pattern counting step for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of occurrences counted in the sequence pattern counting step is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Then, the sequence pattern is sequentially generated, and the sequence data stored in the sequence data storing step is sequentially read out and collated with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data. Process,
A sequence pattern output step of outputting a sequence pattern whose number of occurrences counted by the sequence pattern discovery step is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding method characterized by including A sequence pattern finding device for extracting a sequence pattern that frequently appears in the sequence data from the sequence data in which the fluctuating item values are arranged in order,
Sequence data storage means for storing sequence data to be sequence pattern extracted in a storage device;
A general pattern in which an arbitrary item value that allows an arbitrary item value and a series pattern including only the item value is defined as a generalized pattern, and the series data stored by the series data storage unit is sequentially read and the series data is matched. A sequence pattern counting means for counting the number of occurrences of the digitized pattern;
A generalized pattern in which the number of appearances counted by the sequence pattern counting unit is equal to or greater than a predetermined threshold is extracted, and a selection item value in which an allowable value is selected is set as an arbitrary item value included in the extracted generalized pattern Then, the sequence pattern is sequentially generated, and the sequence data stored in the sequence data storage means is sequentially read out and collated with the sequence pattern, thereby counting the number of occurrences of the sequence pattern matching the sequence data. Means,
Sequence pattern output means for outputting a sequence pattern in which the number of occurrences counted by the sequence pattern finding means is equal to or greater than a predetermined threshold as a sequence pattern of an extraction result;
A sequence pattern finding device characterized by comprising:

Images