TWI413913B - Method for mining subspace clusters from dna microarray data - Google Patents

Abstract

A method for mining subspace clusters from DNA microarray data is provided, in which a plurality of maximum dimension sets (MDSs) are calculated according to a minimum gene value, a minimum condition value and a clustering threshold value; the frequent pattern tree (FP-Tree) is used to calculate a plurality of conditional pattern bases (CPBs) according to each gene and corresponding MDS(s), and qualified subspace clusters are obtained according to the CPBs. Whereby, the complicated distribution process is simplified and the searching space and searching time are reduced. Furthermore, user(s) can utilize the method of the invention to cluster DNA microarray data of biological data base on the World Wide Web (WWW), and can utilize the clustering results to realize relationship between genes.

Description

Method for exploring subspace grouping in DNA microarray data

The invention relates to a method for exploring subspace grouping in a data, in particular, a method for exploring subspace grouping in DNA microarray data.

Sub-space grouping techniques for gene and experimental conditions in gene microarrays have been shown to help understand such functions as gene function, gene regulation, cellular processes, and cell subtypes. In general, in most cases, a disease is made up of multiple genes, so researchers are constantly trying to find out the similarities of certain genes under certain conditions as a basis for judging the disease.

In the prior art, in the method of supporting subspace grouping in gene microarrays, common methods such as pCluster and zCluster are subspace groupings for finding out that certain genes have consistent performance under certain conditions. However, both of these conventional methods involve very time consuming steps, namely constructing the largest set of dimensions of a gene pair and distributing the genetic information on each node of the prefix tree. Therefore, the conventional sub-space grouping method of the gene microarray of pCluster and zCluster has a complicated distribution process and must use a large amount of search space and search time.

Therefore, it is necessary to provide an innovative and progressive method for probing subspace grouping in DNA microarray data to solve the above problems.

The invention provides a method for exploring subspace grouping in DNA microarray data, wherein the DNA microarray is an M×N array composed of M genes and N conditional information of each gene, and the genes have a sequence. Gene number, which has a conditional number in the order. The method comprises the steps of: (a) setting a minimum gene value, a minimum condition information value, and a group threshold; (b) calculating a plurality of condition pairs based on the minimum gene value, the minimum condition information value, and the group threshold value. a set of maximum dimensions, and using a Frequent Pattern Tree (FP-Tree) method to calculate a plurality of conditional pattern bases based on the genes and their corresponding conditions for the largest set of dimensions; and (c) calculating based on the conditional patterns A plurality of subspace clusters.

In the method for exploring subspace grouping in the DNA microarray data, the method for calculating the subspace grouping in the gene microarray is to calculate the corresponding large item from the largest dimension set of the condition pair of the gene by using the high frequency model tree method. Set, from which to obtain a collection of genes related to the conditional pair. Therefore, the method for exploring subspace grouping in the DNA microarray data can avoid complicated distribution process, and the FP-Tree method can greatly reduce the search space and the search time.

Moreover, the method for exploring subspace grouping in the DNA microarray data of the present invention can be applied to a biological database analysis website set up on the World Wide Web (WWW), and the user can group the gene microarray data by the method of the present invention. Analysis, or research units (eg, biomedical companies that study gene microarrays) can use clustering results to help understand the relationships between genes.

The present invention provides a method of profiling subspace grouping in DNA microarray data. Referring to Fig. 1, the DNA microarray is an M×N array composed of M genes (Gene) and N conditional information of each gene, and the genes have sequential gene numbers Gene 1 to Gene M, The condition information has sequential condition numbers c1 to cN, and each condition information of each gene has a value. Preferably, the M system is between 10 ³ and 10 ⁶ and the N system is less than 100.

Figure 2 is a flow chart showing the method of exploring subspace grouping in the DNA microarray data of the present invention. First, referring to step S21, a minimum gene value, a minimum condition information value, and a group threshold are set. The minimum gene value is set according to the number of genes included in the subspace group of the exploration sub-score desired by the user. For example, if the user wants to obtain a sub-population of the exploration subspace containing at least 3 genes, the minimum gene value is set to 3.

Referring to step S22, a plurality of conditional pair maximal dimension sets (MDSs) are calculated according to the minimum gene value, the minimum condition information value, and the group threshold value, and a high frequency model tree (Frequent Pattern Tree, FP-) is used. The Tree method calculates a plurality of Conditional Pattern Bases for the largest set of dimensions based on the genes and their corresponding conditions.

In this embodiment, calculating the condition-to-maximum dimension set includes the steps of: calculating a difference of condition information of each numbered gene under each two condition number, wherein each difference is numbered by a corresponding smaller condition The difference between the condition information and the condition information of the corresponding larger condition number; the difference values are arranged from small to large as a difference order, and the corresponding genes are sorted according to the difference; and according to the minimum gene value, the The minimum condition information value, the group threshold value, the difference ranking, and the genes of the corresponding gene numbers are calculated, and the condition is calculated for the largest dimension set.

Referring to Figure 3, there is shown a schematic diagram of one of the DNA microarrays of the present invention and its associated parameters. In FIG. 3, only 7 genes g ₀ to g ₆ and 5 conditional information c ₁ to c _{5 are taken} as an example, wherein in the present embodiment, the minimum gene value (minG) is set to 3, and the minimum condition information is set. The value (minC) is set to 3, and the group threshold value (δ) is set to 1.

First, calculating the difference g ₀ to g ₆ in the c ₁ to c ₅ in each of the conditions under the second condition of the ID information, which is calculated under the same process conditions for each of the second condition information, and calculates a difference number of the largest dimension set of conditions The following is only exemplified by the conditional information c ₃ and c ₅ .

Referring to FIG. 4, calculating the conditional pattern base includes the following steps. First, calculating the difference (c ₃ -c ₅ ) of the condition information of each of the numbered genes g ₀ to g ₆ at c ₃ and c ₅ ; The differences are stored in a small to large order as a difference order (-3, -2, -2, -1, -1, 4, 5), and the corresponding genes are sorted according to the difference (g ₄ , g ₀ , g ₁ , g ₂ , g ₃ , g ₆ , g ₅ ).

Then, according to the group threshold value 1, each difference value in the difference ranking is sequentially used as a base point, and the difference between the maximum value and the minimum value is defined to be less than or equal to the part of the grouping threshold value ((-3, -2) , -2) and (-2, -2, -1, -1), as defined by the rectangular line) (g ₄ , g ₀ , g ₁ ) and (g ₀ , g ₁ , g ₂ , g ₃ )) and corresponding condition information (c ₃ , c ₅ ), according to which a plurality of conditions are calculated for the largest set of dimensions (g ₀ , g ₁ , g ₄ ) x (c ₃ , c ₅ ) and (g ₀ , g ₁ , g ₂ , g ₃ )x(c ₃ , c ₅ ), the condition includes a conditional pair (c ₃ , for the largest set of dimensions (g ₀ , g ₁ , g ₄ ) x(c ₃ , c ₅ ) c ₅ ) and a maximum dimension set (g ₀ , g ₁ , g ₄ ), the condition includes a condition pair for the largest dimension set (g ₀ , g ₁ , g ₂ , g ₃ ) x(c ₃ , c ₅ ) (c ₃ , c ₅ ) and a set of maximum dimensions (g ₀ , g ₁ , g ₂ , g ₃ ). The conditional pair (c ₃ , c ₅ ) is the condition information c ₃ and c ₅ corresponding to each difference, the largest set of dimensions (g ₀ , g ₁ , g ₄ ) and (g ₀ , g _{1 )} , g ₂ , g ₃ ) has a number of genes of at least the minimum gene value of 3.

Referring to FIG. 5, a schematic diagram of a method for calculating a transaction database according to the present invention is shown. The table on the left side of FIG. 5 is all condition pairs (Condition-pair) and maximum dimension set (MDSs) calculated based on the data of FIG. . First, the number of occurrences of each gene in the largest dimension set is calculated, and a gene order is calculated based on the number of occurrences of the genes, and is recorded in the middle table of FIG. 5, and the Large 1-itemset represents a gene length item. , Support indicates the number of occurrences of each gene, wherein the number of occurrences of the gene g ₆ is less than the minimum condition information value of 3, and the deletion of the gene g _{6 is} not considered, which can be expressed as: The deletion is not considered, wherein the minimum condition information value minC is 3, and the minimum condition information value minC is equal to the minimum gene value minG.

Then, a transaction database (transaction DB, refer to the table on the right side of FIG. 5) is calculated according to the conditions and the number of occurrences of each gene, and the transaction database includes a plurality of transaction rankings (TID: T _{1 -} T ₁₂ ) and corresponding The condition pair and the plurality of transaction maximum dimension sets (transMDSs), the gene systems in the largest dimension set of each transaction are sorted according to at least the number of occurrences. Wherein, if a condition pairs a corresponding plurality of maximum dimension sets, each maximum dimension set is paired with the condition pair to form a plurality of condition pairs and a transaction maximum dimension set, for example: condition pairs (c ₀ , c ₃ ) corresponding to 2 The largest set of dimensions (g ₀ , g ₂ , g ₃ , g ₄ ) and (g ₁ , g ₃ , g ₄ ) are therefore paired into the conditional pair of T ₂ and T ₃ and the largest set of dimensions of the transaction.

Wherein, since the number of occurrences of the gene g ₆ is less than the minimum condition information value 3 and the deletion is not considered, the condition pair (c ₀ , c ₄ ) corresponds to the largest dimension set only considering (g ₃ , g ₄ , g ₅ ), the transaction The conditional pair of T ₄ and the largest dimension set of the transaction are (c ₀ , c ₄ ) and (g ₃ , g ₄ , g ₅ ); (c ₄ , c ₅ ) the corresponding maximum dimension set (g ₀ , g ₅ , g ₆ ) Consider only (g ₀ , g ₅ ), the conditional pair of transaction order T ₁₃ and the largest dimension set of transactions are (c ₄ , c ₅ ) and (g ₀ , g ₅ ).

Figure 6 shows a schematic diagram of the high frequency pattern tree established by the present invention based on the data of Figure 3. Each of the nodes records a gene and the number of times it passes through the node, for example, g ₃ : 10 in the first child node directly below the root node, which indicates that the number of times the gene g ₃ passes through the node is 10 times. . Referring to FIG. 4 to FIG. 6, the high frequency pattern tree (FP-Tree) is established according to the genetic order (g ₄ , g ₀ , g ₁ , g ₂ , g ₃ , g ₅ ) and the transaction database. Each of the lowest-order nodes of the frequency pattern tree has at least one corresponding transaction condition pair. For example, the lowest-order node in the lower left of the high-frequency pattern tree is recorded as (g ₂ : 3), and (g ₂ : 3) represents the gene g. ₂ appears 3 times on the branch of (g ₃ , g ₀ , g ₄ , g ₂ ) of the high-frequency model tree, and the corresponding trading condition pair is T ₁ :(c ₀ , c ₁ ), T ₂ :( c ₀ , c ₃ ), T ₇ : (c ₁ , c ₃ ). For a detailed description of the high-frequency model tree, refer to the journal article: Han, J., Pei, J., and Yin, Y. 2000. Mining frequent patterns without candidate generation. Proc. ACM SIGMOD Int. Conf. Manage. Data 1-12, no further description here.

Referring to FIG. 6 and FIG. 7 , a Conditional Pattern Base (CPB) of each gene is calculated according to the high frequency model tree, the minimum gene value, and the pair of transaction conditions. In the present example, the method of the present invention calculates the conditional pattern basis of the respective genes in the order of the genetic ordering (Fig. 4). The same basic pattern calculated for each condition, the following only the most high-frequency pattern Genome Project tree of the left branch path (gItem) to g ₄ illustrates. Since the minimum gene value minG is 3, the two gene items (g ₃ and g ₀ ) located above the path in the high-frequency pattern tree are considered, and according to the gene items g ₃ , g ₀ and the gene item g ₄ Corresponding transaction conditions for the lower path for T ₁ :(c ₀ , c ₁ ), T ₂ :(c ₀ , c ₃ ), T ₇ :(c ₁ , c ₃ ), calculate the corresponding conditional pattern basis (g ₃ , g ₀ ) x (c ₀ , c ₁ ) & (c ₀ , c ₃ ) & (c ₁ , c ₃ ).

Referring to Figure 8, there is shown a schematic diagram of all corresponding genetic items and conditional pattern foundations established by the present invention based on the data of Figure 3. For example, the genetic item g ₂ and its corresponding conditional pattern basis include: (g ₃ , g ₀ , g ₄ ) x (c ₀ , c ₁ ) & (c ₀ , c ₃ ) & (c ₁ , c ₃ ) (g ₀ , g4) x (c ₀ , c ₅ ), (g ₃ , g ₀ ) x (c ₃ , c ₅ ) & (c ₁ , c ₅ ).

Referring to step S23, a plurality of eligible subspace clusters are calculated according to the conditional pattern basis. In this embodiment, step S23 includes: step S231, calculating a plurality of candidate sets according to the genes and the conditional pattern basis, each candidate set includes a candidate gene pair and a candidate condition pair; and step S232, according to the Genes and the candidate sets calculate a plurality of corresponding first genomes and first condition groups, and calculate a plurality of first large item sets according to the genes, the corresponding first genomes, and the first condition group, and output the first subspace group; Step S233, calculating a plurality of corresponding second genomes and a second condition group according to the corresponding first genome and the first condition group, and calculating a plurality of second largest items according to the genes, the corresponding second genome, and the second condition group And the output is a second subspace group, wherein the number of genes in each second genome is the number of genes in the first genome plus one; and in step S234, the corresponding second genome and the second condition group are defined as The candidate set of the next cycle, and repeating steps S232 and S233, calculates the first subspace group and the second subspace group of the next cycle. In this embodiment, the number of genes of the candidate gene pair of the first genome is the minimum gene value minus one.

Wherein, in step S232, the corresponding first genome and the first condition group are calculated according to the genes and the candidate set having the same gene item; and in step S233, the group is calculated according to the set having the same first condition group. Waiting for the second condition group, and calculating the second genome according to the first genome of the corresponding first condition group; further comprising a screening step in step S234, deleting the first child that is identical to the second large item set Space grouping.

The subspace grouping of each gene project is calculated in the same manner. The following is only an example of the maximum dimension set by the condition of the corresponding gene project g ₂ . Referring to FIG. 8, FIG. 9, FIG. 10 and the above description about step S231, a plurality of candidate sets (CandidateSet[2], wherein the value of [2] is the minimum gene value minG minus 1 is calculated according to the conditional pattern basis respectively. ). In the present embodiment, (g ₃ , g ₀ , g ₄ ) in the conditional pattern basis is divided into candidate gene pairs (g ₃ , g ₀ ), (g ₃ , g ₄ ), and (g ₀ , g ₄ ), And the candidate gene pairs (g ₃ , g ₀ ), (g ₃ , g ₄ ) and (g ₀ , g ₄ ) and the candidate condition pairs (c ₀ , c ₁ ), (c ₀ , c ₃ ) and (c, respectively ₁ , c ₃ ) pairing to form a plurality of candidate sets; (g ₀ , g ₄ ) in the conditional pattern basis is a candidate gene pair, which is paired with the candidate condition pair (c ₀ , c ₅ ) to form a candidate set; conditional pattern The (g ₃ , g ₀ ) in the base is the candidate gene pair, which is paired with the candidate condition pairs (c ₃ , c ₅ ) and (c ₁ , c ₅ ) to form a second candidate set. Candidate sets of corresponding gene items g ₁ , g ₂ , g ₄ , g _{5 are} completely recorded in the intermediate table of FIG.

Referring again to the gene project g ₂ and its corresponding conditional pattern basis (g ₃ , g ₀ , g ₄ ), a plurality of first and second subspace clusters are calculated based on the gene g ₂ and its corresponding candidate set. Referring to FIG. 11 and the above description of step S232, the candidate sets of columns 1-3 in the candidate set (CandidateSet [2]) column have the same candidate gene pair (g ₃ , g ₀ ), the candidate gene pair (g ₃ , g ₀ ) is defined as the first genome, and a plurality of first large item sets are calculated according to the corresponding gene item g ₂ , the first genome and the first condition group and output as the first subspace group. The candidate condition pairs (c ₀ , c ₁ ), (c ₀ , c ₃ ), and (c ₁ , c ₃ ) of the candidate sets are combined into a first condition group (c ₀ , c ₁ , c ₃ ), That is, the combination of the first condition group (c ₀ , c ₁ , c ₃ ) includes (c ₀ , c ₁ ), (c ₀ , c ₃ ), and (c ₁ , c ₃ ), and the corresponding gene item g _{2. The} first genome (g ₃ , g ₀ ) and the first condition group (c ₀ , c ₁ , c ₃ ) form a first large item set L[2]: (g ₂ , g ₃ , g ₀ ) x (c ₀ , c ₁ , c ₃ ); Similarly, in the candidate set (CandidateSet [2]) column, the candidate set of the third column and the last 1-2 columns have the same candidate gene pair (g ₃ , g ₀ The candidate gene pair (g ₃ , g ₀ ) is defined as a first genome, and the candidate condition pairs of the candidate sets are combined with (c ₁ , c ₃ ), (c ₃ , c ₅ ) and (c ₁ , c ₅ ) Is a first condition group (c ₁ , c ₃ , c ₅ ), that is, the combination of the first condition group (c ₁ , c ₃ , c ₅ ) comprises (c ₁ , c ₃ ), (c _{3 )} , c ₅ ) and (c ₁ , c ₅ ), the corresponding gene project g ₂ , the first genome (g ₃ , g ₀ ) and the first condition group (c ₁ , c ₃ , c ₅ ) form another first largest Item set L[2]: (g ₂ , g ₃ , g ₀ )x(c ₁ , c ₃ , c ₅ ). After calculating the first large item set L[2], the first large item set L[2] is output as the first subspace group.

Referring to FIG. 12 and the above description of step S233, the first genome and the first condition group of the subset (g ₃ , g ₀ ) of the corresponding condition pattern basis (g ₃ , g ₀ , g ₄ ) are (g ₃ , g ₀ ) x(c ₀ , c ₁ , c ₃ ) and (g ₃ , g ₀ )x(c ₁ , c ₃ , c ₅ ); a subset of (g ₃ , g ₀ , g ₄ ) in the corresponding conditional pattern basis (g _3, g ₄₎ and the first genome of a first set of conditions _{(g 3, g 4) x} (c 0, c 1, c 3); the basic pattern corresponding conditions _{(g 3, g 0, g} 4 The first genome of the set of children (g ₀ , g ₄ ) and the first condition set are (g ₀ , g ₄ ) x (c ₀ , c ₁ , c ₃ ). Here, (g ₃ , g ₀ )x(c ₀ , c ₁ , c ₃ ), (g ₃ , g ₄ )x(c ₀ , c ₁ , c ₃ ) and (g ₀ , g ₄ )x ( _{c 0, c 1, c 3} ) are all of a first set of conditions _{(c 0, c 1, c 3} ) , i.e., a first set of the three sets of conditions may be formed as a second set of conditions (c ₀ , c ₁ , c ₃ ). Then, combining the first genomes (g ₃ , g ₀ ), (g ₃ , g ₄ ), and (g ₀ , g ₄ ) corresponding to the above three first condition groups into a genome set (g ₃ , g ₀ , g ₄ ) (second genome). In addition, the second genome (g ₃ , g ₀ , g ₄ ) of the second condition group (c ₀ , c ₁ , c ₃ ) is a combination of C[3]: (g ₃ , g ₀ , g _{4 )} x(c ₀ , c ₁ , c ₃ ), and the second genome (g ₃ , g ₀ , g ₄ ), the second condition group (c ₀ , c ₁ , c ₃ ) and the corresponding gene The item g _{2 is} combined into a second largest item set L[3], and the second large item set L[3] is output as the second subspace group (g ₀ , g ₂ , g ₃ , g ₄ )x ( c ₀ , c ₁ , c ₃ ).

For other gene items (eg, g ₁ , g ₄ , g ₅ ) corresponding candidate sets, first and second subspace groupings, and second subspace grouping calculation methods, corresponding to the gene item g ₂ candidate sets, first and second The calculation methods of subspace grouping and second subspace grouping are the same, and will not be repeated here.

The calculated plurality of second genomes and the second condition group are defined as candidate gene pairs and candidate condition pairs in the next stage, and the steps of calculating the first subspace group and the second subspace group are repeated until the combination cannot be performed. The first large project set of the next stage (L[2+1], L[2+2], L[2+3],...) or a combination of the corresponding second genome and the second condition group ( C[3+1], C[3+2], C[3+3], ...), which is a method for performing subspace grouping in the DNA microarray data of the present invention. In the method for exploring subspace grouping in DNA microarray data, the problem of subspace grouping in gene microarray is adopted, and a clustering algorithm based on a large item set is used, and a high frequency model tree is used (Frequent Pattern Tree). , FP-Tree) method calculates the corresponding large item set from the largest dimension set of the condition pair of genes, and obtains the gene set related to the condition pair. Therefore, the method for exploring subspace grouping in the DNA microarray data can avoid complicated distribution process, and the FP-Tree method can greatly reduce the search space and the search time.

Moreover, the method for exploring subspace grouping in the DNA microarray data of the present invention can be applied to a biological database analysis website set up on the World Wide Web (WWW), and the user can group the gene microarray data by the method of the present invention. Analysis, or research units (eg, biomedical companies that study gene microarrays) can use clustering results to help understand the relationships between genes.

The above embodiments are merely illustrative of the principles and effects of the invention and are not intended to limit the invention. Therefore, those skilled in the art can make modifications and changes to the above embodiments without departing from the spirit of the invention. The scope of the invention should be as set forth in the appended claims.

Figure 1 shows a schematic diagram of a DNA microarray;

2 is a flow chart showing a method for exploring subspace grouping in the DNA microarray data of the present invention;

Figure 3 is a schematic view showing a DNA microarray of the present invention and related parameters;

Figure 4 is a schematic view showing the process of calculating the condition pattern base of the present invention;

Figure 5 is a schematic diagram showing the method of calculating a transaction database of the present invention;

6 and FIG. 7 are schematic diagrams showing the conditional basis of the corresponding genes in the present invention;

Figure 8 is a schematic diagram showing the basis of all corresponding genetic items and conditional patterns established by the present invention based on the data of Figure 3;

9 and 10 are schematic diagrams showing the calculation candidate set of the present invention;

FIG. 11 is a schematic diagram showing the calculation of the first subspace grouping by the present invention; and

Figure 12 is a diagram showing the calculation of the second subspace grouping of the present invention.

(no component symbol description)

Claims (12)

A method for exploring subspace grouping in DNA microarray data, wherein the DNA microarray is an M×N array composed of M genes and N conditional information of each gene, and the genes have sequential gene numbers. The condition information has a sequential condition number, and the method comprises the steps of: (a) setting a minimum gene value, a minimum condition information value, and a group threshold; (b) according to the minimum gene value, the minimum condition information value And the group threshold value is used to calculate a plurality of condition-to-maximum dimension sets, and a Frequent Pattern Tree (FP-Tree) method is used to calculate a plurality of conditional pattern bases according to the genes and corresponding conditions for the largest dimension set; and (c) Calculating a plurality of eligible subspace clusters based on the conditional pattern basis. The method of claim 1, wherein in the step (a), the M system is between 10 ³ and 10 ⁶ and the N system is less than 100. The method of claim 1, wherein calculating the condition-to-maximum dimension set in step (b) comprises the step of: (b1) calculating a difference in condition information of each numbered gene under each two condition number, wherein each The difference is the difference between the condition information of the corresponding smaller condition number and the condition information of the corresponding larger condition number; (b2) the difference is ranked from small to large as a difference, and the difference is sorted according to the difference Arrange the corresponding genes; and (b3) calculating the condition-to-maximum dimension set according to the minimum gene value, the minimum condition information value, the grouping threshold value, the difference ranking, and the genes of the corresponding gene numbers. The method of claim 3, wherein in step (b3), the difference between the maximum value and the minimum value is defined as less than or equal to the difference between the difference value and the minimum value according to the grouping threshold value. The corresponding gene of the group threshold and the corresponding condition information are used to calculate a plurality of conditional pair maximum dimension sets, wherein each condition pair maximum dimension set includes a condition pair and a maximum dimension set, and each condition pair is a corresponding each difference value The second condition information, each maximum dimension set has at least the number of genes with the smallest gene value. The method of claim 4, wherein calculating the conditional pattern base after the step (b3) comprises the steps of: (b31) calculating a genetic order according to the number of occurrences of each gene in the maximum dimension set; (b32) according to the conditions Calculating a transaction database for the number of occurrences of each gene, the transaction database includes a plurality of transaction rankings and corresponding condition pairs and a plurality of transaction maximum dimension sets; (b33) establishing the height according to the genetic ranking and the transaction database a frequency pattern tree, each of the lowest order nodes of the high frequency pattern tree having at least one corresponding pair of transaction conditions; and (b34) calculating corresponding genes according to the high frequency model tree, the minimum gene value, and the pair of transaction conditions Conditional Pattern Base. The method of claim 5, wherein in step (b32), the maximum dimension of each transaction The gene lines in the degree of concentration are ordered at least according to the number of occurrences. The method of claim 5, wherein in step (b34), the conditional pattern basis of the respective genes is calculated in the order in which the genes are sorted. The method of claim 5, wherein the step (c) comprises the steps of: (c1) calculating a plurality of candidate sets based on the genes and the conditional pattern basis, each candidate set comprising a candidate gene pair and a candidate condition pair; (c2) calculating a plurality of corresponding first genomes and first condition groups based on the genes and the candidate sets, and calculating a plurality of first large item sets based on the genes, the corresponding first genomes, and the first condition group, and outputting First subspace grouping; (c3) calculating a plurality of corresponding second genomes and second condition groups according to the corresponding first genome and the first condition group, and calculating according to the genes, the corresponding second genome, and the second condition group a plurality of second largest item sets and outputting the second subspace group, wherein the number of genes in each second genome is the number of genes in the first genome plus one; and (c4) defining the corresponding second genome And the second condition group is a candidate set of the next cycle, and steps (c2) and (c3) are repeated, and the first subspace group and the second subspace group of the next cycle are calculated. The method of claim 8, wherein in the step (c1), the number of genes of the candidate gene pair is the minimum gene value minus one. The method of claim 8, wherein in step (c2), the corresponding first genome is calculated based on the genes and candidate sets having the same gene project First condition group. The method of claim 8, wherein in step (c3), the second condition group is calculated according to a set having the same first condition group, and the first group is calculated according to the first genome group corresponding to the first condition group The second genome. The method of claim 8, wherein in the step (c4), a screening step is further included to delete the first subspace grouping that is the same as the second largest item set.