KR20210041185A - Pattern Dictionary Establish Method by Data Classify and Analysis - Google Patents

Abstract

According to the present invention, a method of building a pattern dictionary in accordance with data classification analysis includes the following steps of: (a) sampling and classifying data from a sample document in accordance with subject words by user-defined classification system; (b) conducting morpheme analysis with respect to an analysis target word combination included in the data sampled by the step (a); (c) classifying the analysis target word combination as an analysis failure combination, an analysis exception combination and a normalization combination in accordance with the morpheme analysis result of the step (b); (d) conducting follow-up analysis with respect to the normalization combination of the combinations classified by the step (c); and (e) building a pattern dictionary through the normalization combination to which the follow-up analysis has been conducted in the step (d). Therefore, the present invention is capable of considerably increasing the efficiency of analysis.

Description

Pattern Dictionary Establish Method by Data Classify and Analysis}

The present invention relates to a pattern dictionary construction method according to data classification analysis, and more specifically, data for improving work efficiency by accurately searching for desired data after classifying/analyzing a large amount of data by classification system defined by a user. It relates to a method of constructing a pattern dictionary according to classification analysis.

Today, the amount of electronic documents is vast, and it is difficult to quickly search for desired information of users, and a representative example of such electronic documents is an electronic document that is generated and distributed in real time through the Internet.

As a method of providing such a large amount of electronic documents in a form that is easy for users to search and use, many companies have developed and serviced Semantic Search, Sentiment Analysis Search, and Classification Search, etc. Or in development.

Text mining is an essential technique for such search techniques, and many classification extraction methods are used.

In the conventional classification analysis system as described above, a large amount of data is directly reviewed by a person to generate an analysis pattern and a pattern dictionary is constructed, so a lot of time and labor costs were incurred in the construction process.

However, there is a limit to the data that can be checked by humans, so it is difficult to establish a broad pattern dictionary, and the accuracy and reproducibility of the analysis results are also low, so the application of work is limited, and a combination of adjectives and nouns such as proper nouns and movie titles should be used in combination. In this case, there is a problem that an error occurs in the analysis process.

Therefore, a method for solving these problems is required.

Korean Patent Publication No. 10-1999-0087858

The present invention is an invention conceived to solve the problems of the prior art described above. After classifying/analyzing a large amount of data by classification system defined by a user, the present invention is used to accurately search for desired data to improve work efficiency. It has the purpose of providing a pattern dictionary construction method according to.

The problems of the present invention are not limited to the problems mentioned above, and other problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

The method for constructing a pattern dictionary according to the data classification analysis of the present invention for achieving the above object comprises the steps (a) and (a) of sampling and classifying data from sample documents according to the subject words for each user-defined classification system. According to the morpheme analysis results of step (b) and step (b), which perform morpheme analysis on the sampled data, the analysis target word combinations are converted into analysis failure combinations, analysis exclusion combinations, and normalization combinations. The pattern dictionary is analyzed through the step (c) of classifying, the step (d) of performing a follow-up analysis on the normalized combination among the combinations classified by the step (c), and the normalization combination performed in the step (d). It includes the step (e) of building.

In this case, the step (a) includes the step (a-1) of matching the data included in the sample document with the subject word for each user-defined classification system, and the ( It may include step a-2).

In addition, the step (a) includes the step (a-3) of matching the data matched in step (a-1) with the excluded words for each user-defined classification system, and the data matched in step (a-3). A step (a-4) of classifying the data as valueless data, and a step (a-5) of sampling data not matched in the step (a-4) may be included.

In addition, in step (b), morpheme analysis may be performed by formulating the analysis target word combination.

In addition, the morpheme analysis may include a pattern method formula including a unique search for searching the sampled data for the analysis target word combination itself, and an adjacent search for combining adjacent analysis target word combinations.

In addition, the morpheme analysis includes a forward search for searching for a pattern from a word exposed to the right of a reference word, a reverse search for searching for a pattern from a word exposed to the left of the reference word, and a right side of the reference word. And a search direction formula including a full search for searching for a pattern in words exposed from both left sides.

In addition, the morpheme analysis may include a search range formula for searching for a pattern within n adjacent ranges around a reference word.

In addition, the morpheme analysis may include a matching analysis for searching the sampled data for an arbitrary word itself, and a search method formula including a dictionary analysis for recognizing whether an arbitrary word is a word in a dictionary.

And in the step (d), for the normalization combination selected by the step (d-1) and the normalization combination selected by the step (d-1) and the step (d-1) of performing TF-IDF analysis on the normalized combination, n- As the gram analysis is performed, a step (d-2) of calculating the exposure level in the sample document may be included.

Further, in step (e), a pattern dictionary may be constructed by arranging the normalization combinations in descending order according to the exposure level in the sample document.

Meanwhile, the present invention may further include a step (f) of setting a first predicted new word for an analysis failure combination among the combinations classified by step (c) and recommending it to the user.

At this time, the step (f) includes the step (f-1) of setting the first predicted new word for the failed analysis combination among the combinations classified by the step (c), and the TF-IDF analysis for the first predicted new word. Step (f-2) of performing and screening, step (f-3) of identifying and removing the type and location of the survey for the first predicted new word selected by step (f-2), the (f- Unanalyzed words for the first predicted new words for which the morpheme analysis was re-performed by steps (f-4) and (f-4), in which the morpheme analysis was re-performed on the first predicted new words from which the investigation was removed by step 3) If it is determined that there is an unanalyzed word as a result of the determination of step (f-5) and step (f-5) of determining whether or not is present, the number of searches within the analysis period is extracted for the corresponding first predicted new word. If the search quantity extracted during the analysis period by step (f-6) and step (f-6) shows an increase rate of n% or more with respect to the search quantity in the previous period, the corresponding first predicted new word is used as a new word recommended by the user. It may include the step (f-7) of setting and providing.

In addition, in step (f), when the number of searches extracted in the analysis period by step (f-6) shows an increase rate of less than n% of the number of searches in the previous period, the corresponding first predicted new word is second predicted. After setting the new word, the step (f-8) of setting and providing the new word recommended by the user may be further included.

Meanwhile, the present invention may further include a step (g) of performing sentiment analysis on the normalization combination included in the pattern dictionary constructed by the step (e).

In this case, the step (g) includes the step (g-1) of assigning emotion to the normalization combination included in the pattern dictionary according to the positive rate formula, and assigning a correlation score to the normalization combination included in the pattern dictionary. It may include step (g-2).

Here, as the positive rate formula,

Positive rate = (positive pattern quantity/(positive pattern quantity + negative pattern quantity))*100

The formula of can be applied.

In addition, the correlation score is,

Association score = (Total exposure pattern quantity/Normalization combination quantity)*100

The formula of can be applied.

The method of constructing a pattern dictionary according to the data classification analysis of the present invention for solving the above problems has the advantage that a user can quickly and accurately search a large amount of data, and easily access the data by removing unnecessary contents from the large amount of data. There is this.

In addition, the present invention has advantages in that the application of new words newly posted on the Internet is fast, the expansion of the pattern dictionary is easy, and the expansion of the user-defined classification system is easy.

In addition, the present invention has an advantage in that the accuracy increases according to the size of the dictionary and various classifications can be performed, and thus analysis efficiency can be greatly improved by simultaneously analyzing various subjects.

The effects of the present invention are not limited to the above-mentioned effects, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

1 and 2 are diagrams and flow charts showing an overall process of a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention;
3 is a diagram showing a detailed process of a step (a) of sampling and classifying data from a sample document in a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention;
4 is a table showing types of equations used for morpheme analysis in a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention;
FIG. 5 is a diagram showing a detailed process of step (d) of performing a subsequent analysis on a normalized combination in a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention; FIG.
6 and 7 are diagrams and flow charts showing a detailed process of step (f) in which a first predicted new word is set and recommended to a user in a pattern dictionary construction method according to data classification analysis according to an embodiment of the present invention;
8 is a survey considered in step (f-3) of identifying and removing the type and location of a survey for a first predicted new word in a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention A table showing the types of; And
9 and 10 are diagrams and flow charts showing a detailed process of step (g) of performing sentiment analysis on a normalized combination among a pattern dictionary construction method according to data classification analysis according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention in which the object of the present invention can be realized in detail will be described with reference to the accompanying drawings. In the description of the present embodiment, the same names and the same reference numerals are used for the same components, and additional descriptions thereof will be omitted.

The present invention is a system for automatically classifying/analyzing a large amount of data, and may provide a more accurate analysis result to a user by using functions such as a user-defined classification/analysis pattern dictionary construction procedure, pattern formulating, and new word recommendation. Hereinafter, each process and function for carrying out the present invention will be described in detail.

1 and 2 are diagrams and flowcharts showing the overall process of a method for constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention, and all descriptions below are based on FIGS. 1 and 2 showing the overall process as described above. It will be described with reference to the other drawings together.

As shown in FIGS. 1 and 2, in the method of constructing a pattern dictionary according to data classification analysis according to an embodiment of the present invention, a step (a) of sampling and classifying data from a sample document according to a subject word for each user-defined classification system Wow, according to the morpheme analysis results of step (b) and (b), which performs morphological analysis on the word combinations to be analyzed included in the data sampled by step (a), the analysis failure combination In step (c), which is classified into non-analytical combinations and normalized combinations, and (d), which performs a follow-up analysis on the normalized combinations among the combinations classified by step (c), and in step (d), follow-up analysis is performed. It includes the step (e) of constructing a pattern dictionary through the normalized combination.

On the other hand, this embodiment is included in the pattern dictionary constructed by step (f) and step (e) in which the analysis failure combination is set as the first predicted new word and recommended to the user among the combinations classified by step (c). It may further include the step (g) of performing sentiment analysis on the normalized combination.

In step (a), in order to sample and classify data from sample documents according to the subject words for each user-defined classification system, the data is sampled by selecting a representative subject word for each classification from the sample document.

At this time, it is desirable to set the exclusion words that are used together with the corresponding representative subject words and to be classified with different meanings and set them in detail so that they can be accurately sampled in the corresponding classification. In this way, documents classified for each classification are extracted by morpheme analysis in step (b), and the words that have been successfully analyzed are extracted, and the extracted words are normalized through the removal of the analysis exclusion combination. In this case, the analysis-excluded combination means a combination of words that are successful in analysis or a combination of general words that will not be used when generating patterns.

More specifically, as shown in FIG. 3, step (a) includes step (a-1) of matching the data included in the sample document with a subject word for each user-defined classification system, and the unmatched step (a-1). Steps (a-2) of classifying data as valueless data, step (a-3) of matching the data matched in step (a-1) with excluded words for each user-defined classification system, and (a-3) It includes a step (a-4) of classifying the data matched in step (a-4) as non-value data, and a step (a-5) of sampling the unmatched data in step (a-4).

In other words, in step (a), the data included in the sample document is matched with the subject words for each user-defined classification system, and the unmatched data is removed as valueless data, and the matched data is matched again with the excluded words for each user-defined classification system. This data is similarly classified and filtered as valueless data.

Then, by sampling the data that is finally left through the above process, one or more combinations of words to be analyzed are selected.

Thereafter, in step (b), a process of performing morpheme analysis by formulating the analysis target word combination selected in step (a) is performed.

In general, the main problems that occurred in the conventional data classification and analysis process were patterns composed of complex combinations of parts of speech such as movie titles and proper nouns. Since such patterns include words that cannot be analyzed morphemes, there is a problem that it is difficult to reflect the combined words in the pattern.

Therefore, in order to solve the above problems, the present invention is designed to combine by changing the combination of the set values and words of each pattern into an equation. The advantage of this formula is that the pattern is excellent in extensibility, and it is possible to use a combination of several parts of speech, so that it is easy to use the pattern setting value during automatic analysis.

In addition, the types of equations used for morpheme analysis in this example are shown in the table of FIG. 4.

As shown in FIG. 4, in the present embodiment, the morpheme analysis in step (b) includes a unique search (G) in which the analysis target word combination selected in step (a) is searched from the sampled data, and an analysis adjacent to each other. It includes a pattern method formula including an adjacent search (N) for searching by combining target word combinations.

In addition, when using such an adjacent search (N), a forward search (F) searches for a pattern from a word exposed to the right of a reference word, and a reverse search (F) searches for a pattern from a word exposed to the left of the reference word. A search direction formula including a search (R) and a full search (A) for searching for a pattern in words exposed from both the right and left sides of the reference word may be optionally additionally used.

And when using such a search direction formula, a search range formula (1-n) that searches for a pattern within n contiguous ranges centered on the reference word can be used further, and the arbitrary word itself is sampled as it is. A search method formula including a matching analysis (M) for searching the generated data and a dictionary analysis (D) for recognizing whether an arbitrary word is a word in a dictionary may be further used.

The usage method of each of these formulas is described below by way of example.

[G] (Afternoon on a dog-like day)

[N, F, 5] ([M] Murder memories, [D] number of spectators)

[N, R, 5] ([D] domestic, [D] semiconductor, [D] status)

[N, A, 5] ([M] Baskin Robbins, [D] Almond Bonbon, [M] John Flavor)

That is, the present invention has excellent expandability by formulating the set value of each pattern and combinations of words, and it is easy to utilize the pattern set value during automatic analysis because several parts of speech can be used in combination.

Next, according to the result of the morpheme analysis, step (c) of classifying the analysis target word combination into an analysis failure combination, an analysis-exclusion combination, and a normalization combination is performed.

Here, the analysis failure combination refers to a combination of words that fail morpheme analysis and will not be used as a new word in the future, and the analysis-exclusion combination refers to a combination of general words that are not used when generating patterns, although the words that have been successful in morphological analysis. Combination refers to the remaining combinations after removing the failed analysis combination and the non-analysis combination.

Meanwhile, the analysis failure combination is classified as the first predicted new word and used in the automatic new word recommendation process to be described later, which will be described later.

After step (c), step (d) of performing a subsequent analysis on the normalized combinations among the combinations classified by step (c) is performed.

As shown in Fig. 5, in the present embodiment, step (d) is selected by step (d-1) and step (d-1) of performing TF-IDF analysis on the normalized combination in detail. It includes the step (d-2) of estimating the exposure level in the sample document by performing n-gram analysis with other words around the normalized combination.

Here, in the case of TF-IDF analysis, it is to remove the analyzed words, that is, the user general words that were not excluded in the previous step among the normalization combinations, and the meaningless words in the user classification, and TF is the exposure of specific words from one data. It is an index indicating degree, and IDF indicates an index indicating the importance of a specific word in the entire document.

And the TF-IDF analysis process can be selected through the following equation.

(N: total number of documents, DF: number of documents with a specific word)

In addition, in the case of n-gram analysis, it is to calculate the exposure level in the sample document by generating a group of n words.The value of n is used from 1 to 3, and words appearing consecutively among the surrounding words of the corresponding normalization combination are used. Bundled together to create a sample pattern.

Next, step (e) of constructing a pattern dictionary is performed through the normalization combination that has been subjected to subsequent analysis in step (d).

In this step, using the TF-IDF score and the n-gram exposure, the normalization combinations are sorted in descending order according to the exposure in the sample document and recommended to the user to construct a pattern dictionary. At this time, by adding an analysis-excluded combination, a more accurate recommendation pattern can be created when the next pattern dictionary is additionally constructed.

Meanwhile, as described above, the analysis failure combination classified in step (c) may be classified as a first predicted new word and used in a new word automatic recommendation process to be described later. To this end, the present embodiment may further include a step (f) of setting an analysis failure combination as a first predicted new word among combinations classified by step (c) and recommending it to a user.

6 and 7 are diagrams and flow charts showing a detailed process of step (f) in which a first predicted new word is set and recommended to a user in a method of pre-establishing a pattern according to data classification analysis according to an embodiment of the present invention.

As shown in FIGS. 6 and 7, step (f) in the present embodiment includes step (f-1) of setting the first predicted new word for the analysis failure combination among the combinations classified by step (c). , Identify and remove the type and location of the survey for the first predicted new words selected by step (f-2) and (f-2), which performs TF-IDF analysis on the first predicted new words. The morpheme analysis is performed by steps (f-3), (f-4) and (f-4), re-performing the morpheme analysis on the first predicted new words from which the investigation has been removed by the (f-3) and (f-3) steps. It includes a step (f-5) of determining whether or not an unanalyzed word exists for the re-performed primary predicted new word.

That is, in this step, words that have not been previously analyzed are recommended to the user through a new word recommendation process using the first predicted new words extracted when constructing the pattern dictionary.

At this time, the first predicted new word can be started in a state that excludes commonly used words through the TF-IDF analysis technique, and the extracted words are processed in a state that basically removes the investigation, and then morpheme analysis can be performed.

Here, the TF-IDF analysis in step (f-2) and the morpheme analysis in step (f-4) can be performed in the same manner as the TF-IDF analysis in step (d) and the morpheme analysis in step (b) described above. Therefore, detailed description will be omitted.

In the case of step (f-3), based on the survey dictionary as shown in the table shown in FIG. 8, the type of survey and the connection location of the survey are identified and removed. As shown in FIG. 8, in the case of a narrative questionnaire, it appears only at the end position (E) of the sentence, and all other surveys may appear at both the middle position (M) and the end position (E) of the sentence.

In the present embodiment, step (f) is, when it is determined that an unanalyzed word exists as a result of the determination in step (f-5), extracting the number of searches within the analysis period for the corresponding first predicted new word (f-6). ) Step may be further included, and if the search quantity extracted within the analysis period by this step (f-6) shows an increase rate of n% or more with respect to the search quantity in the previous period, the corresponding first predicted new word is recommended by the user. If the search quantity extracted in the analysis period by steps (f-7) and (f-6) provided by setting a new word shows an increase rate of less than n% of the search quantity in the previous period, the corresponding primary prediction It may further include a step (f-8) of setting the new word as a second predicted new word, and then setting and providing the new word recommended by the user.

Here, n% may be a value set according to an arbitrary criterion, and in this embodiment, it is assumed to be 200%. However, it goes without saying that this may be variously set differently from the present embodiment.

The recommended new words can be separately registered and utilized in the pattern dictionary, and accordingly, the present invention has the advantage that the application of the new words newly posted on the Internet is very fast.

Meanwhile, as described above, the present embodiment may further include step (g) of performing sentiment analysis on the normalization combination included in the pattern dictionary constructed by step (e).

9 and 10 are diagrams and flow charts showing a detailed process of step (g) of performing sentiment analysis on a normalized combination among a pattern dictionary construction method according to data classification analysis according to an embodiment of the present invention.

As shown in FIGS. 9 and 10, step (g) in the present embodiment includes step (g-1) of imparting sensibility to the normalization combination included in the pattern dictionary according to the positive rate formula, and included in the pattern dictionary. It may include a step (g-2) of assigning an association score to the normalized combination.

Here, as the positive rate formula,

Positive rate = (positive pattern quantity/(positive pattern quantity + negative pattern quantity))*100

The formula of can be applied.

Also, the relevance score is,

Association score = (Total exposure pattern quantity/Normalization combination quantity)*100

The formula of can be applied.

As described above, in this embodiment, emotion is given to the normalization combination through the positive rate formula, and more accurate data is searched by allowing the user to sort in the order of data with high relevance when searching for data or analysis results through the relevance score. To be able to do it.

In conclusion, the present invention has an advantage in that a user can quickly and accurately search for a large amount of data, and easily access data from which unnecessary contents are removed from the large amount of data.

As described above, preferred embodiments according to the present invention have been examined, and the fact that the present invention can be embodied in other specific forms without departing from its spirit or scope in addition to the above-described embodiments is known to those skilled in the art. It is self-evident to them. Therefore, the above-described embodiments are to be regarded as illustrative rather than restrictive, and accordingly, the present invention is not limited to the above description and may be modified within the scope of the appended claims and their equivalents.

Claims (17)

(A) sampling and classifying data from the sample document according to the subject word for each user-defined classification system;
(B) performing a morpheme analysis on a combination of words to be analyzed included in the data sampled by the step (a);
(C) classifying the analysis target word combination into an analysis failure combination, an analysis-exclusion combination, and a normalization combination according to the result of the morpheme analysis in step (b);
(D) performing a subsequent analysis on the normalized combination among the combinations classified by the (c) step; And
Step (e) of constructing a pattern dictionary through the normalization combination performed subsequent analysis in step (d);
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 1,
The step (a),
(A-1) matching data included in the sample document with a subject word for each user-defined classification system; And
(A-2) classifying data not matched in step (a-1) as valueless data;
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 2,
The step (a),
Step (a-3) matching the data matched in step (a-1) with excluded words for each user-defined classification system;
(A-4) classifying the data matched in step (a-3) as the valueless data; And
(A-5) sampling the data not matched in the (a-4) step;
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 1,
The step (b),
A method of constructing a pattern dictionary according to data classification analysis for performing morpheme analysis by formulating the analysis target word combination. The method of claim 4,
The morpheme analysis,
A pattern dictionary construction method according to data classification analysis, including a pattern method formula including a unique search for searching for the analysis target word combination itself from sampled data and an adjacent search for combining adjacent analysis target word combinations. The method of claim 4,
The morpheme analysis,
A forward search that searches for a pattern from a word exposed to the right of the reference word, a reverse search that searches for a pattern from a word that is exposed to the left of the reference word, and an exposed from both the right and left of the reference word. A method of constructing a pattern dictionary according to data classification analysis including a search direction formula including a full search for searching for patterns in words. The method of claim 4,
The morpheme analysis,
A method of constructing a pattern dictionary according to data classification analysis including a search range formula that searches for patterns within n adjacent ranges based on a reference word. The method of claim 4,
The morpheme analysis,
A method of constructing a pattern dictionary according to data classification analysis including a matching analysis that searches for an arbitrary word itself from the sampled data, and a search method formula including a dictionary analysis that recognizes whether an arbitrary word is a word in the dictionary. The method of claim 1,
The step (d),
(D-1) selecting the normalized combination by performing TF-IDF analysis; And
(D-2) calculating the exposure level in the sample document by performing n-gram analysis with other words around the normalized combination selected by the (d-1) step;
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 9,
The step (e),
A pattern dictionary construction method according to data classification analysis for constructing a pattern dictionary by arranging the normalization combinations in descending order according to the exposure level in the sample document. The method of claim 1,
The method of constructing a pattern dictionary according to data classification analysis, further comprising step (f) of setting and recommending a first predicted new word to a user for an analysis failure combination among the combinations classified by step (c). The method of claim 11,
The step (f),
Step (f-1) of setting the first predicted new word for the failed analysis combination among the combinations classified by step (c);
(F-2) selecting by performing TF-IDF analysis on the first predicted new word;
(F-3) step of identifying and removing the type and location of the survey for the first predicted new word selected by step (f-2);
(F-4) re-performing morpheme analysis on the first predicted new words from which the investigation was removed by the (f-3) step;
(f-5) determining whether or not an unanalyzed word exists for the first predicted new word for which the morpheme analysis was re-performed in step (f-4);
(F-6) extracting the number of searches within the analysis period for the first predicted new word when it is determined that the unanalyzed word exists as a result of the determination in step (f-5); And
If the search quantity extracted in the analysis period by step (f-6) shows an increase rate of n% or more with respect to the search quantity in the previous period, the first predicted new word is set and provided as a user recommended new word (f-7). )step;
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 12,
The step (f),
If the search quantity extracted in the analysis period in step (f-6) shows an increase rate of less than n% of the search quantity in the previous period, the first predicted new word is set as the second predicted new word, and then the user is recommended. A method of constructing a pattern dictionary according to data classification analysis further comprising the step (f-8) of setting and providing a new word. The method of claim 1,
A method of constructing a pattern dictionary according to data classification analysis, further comprising a step (g) of performing sentiment analysis on the normalization combination included in the pattern dictionary constructed by the step (e). The method of claim 14,
The step (g),
(G-1) giving emotion to the normalization combination included in the pattern dictionary according to a positive rate formula; And
(G-2) assigning an association score to the normalized combination included in the pattern dictionary;
Pattern dictionary construction method according to data classification analysis comprising a. The method of claim 15,
With the above positive rate formula,
Positive rate = (positive pattern quantity/(positive pattern quantity + negative pattern quantity))*100
A method of constructing a pattern dictionary according to data classification analysis to which the formula of is applied. The method of claim 15,
The association score is,
Association score = (Total exposure pattern quantity/Normalization combination quantity)*100
A method of constructing a pattern dictionary according to data classification analysis to which the formula of is applied.

Images