Method and device for extracting MOM scene features and mining multi-view code semantic associations

Abstract

The present invention relates to a method and device for extracting MOM scene features and mining semantic associations of multi-view codes, the method comprising: modeling and refining the target scene's code and document data through a MOM system into sub-scenes of the target scene; extracting key features from the sub-scenes using a feature extraction algorithm and building a feature library; semantically parsing the codes under multiple scene views through multi-view code analysis, and mining potential association rules from the parsed codes using an association rule mining algorithm. The present invention refines sub-scenes through a clustering algorithm, extracts key features through a principal component analysis method, and mines frequent item sets through an FP‑Growth algorithm, and finally generates effective association rules.

Description

Technical Field

[0001] The present invention relates to the technical field of software engineering, and in particular to a method and device for extracting MOM scene features and mining multi-view code semantic associations. Background Art

[0002] The MOM (Manufacturing Operations Management) system is a comprehensive manufacturing management solution designed to optimize and coordinate the entire manufacturing process. The MOM system not only covers the traditional MES (Manufacturing Execution System), but also integrates other related systems such as MPM (Manufacturing Process Management), LIMS (Laboratory Information Management System) and EAM (Enterprise Asset Management), forming a comprehensive management platform covering the entire manufacturing process.

[0003] In the process of software development, understanding and analyzing the semantic associations of code is of great significance for improving software quality and optimizing software structure. However, traditional code analysis methods can usually only analyze code under a single view, and it is difficult to fully reveal the complex semantic associations between codes. In addition, as the scale of software increases and the complexity of scenarios increases, how to accurately extract key features and conduct effective semantic association mining has become an urgent problem to be solved. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a method and device for MOM scenario feature extraction and multi-view code semantic association mining, aiming to model the target scenario in a model-driven manner, perform semantic analysis and association rule mining on the code from multiple views, so as to improve the accuracy and comprehensiveness of code analysis.

[0005] The specific plan is as follows:

[0006] On the one hand, MOM scene feature extraction and multi-view code semantic association mining methods include:

[0007] A method for extracting MOM scene features and mining multi-view code semantic associations, characterized by comprising:

[0008] S1, based on the code data and document data of the target scenario, the target scenario is modeled through the MOM system, and the modeled target scenario is refined into sub-scenarios with unique feature sets;

[0009] S2, extract key features from sub-scenes with unique feature sets through feature extraction algorithms, and build a feature library based on the extracted key features;

[0010] S3, retrieves key feature vectors based on the feature library, performs semantic analysis on the codes under multiple scenario views based on the key feature vectors, and uses the association rule mining algorithm to mine potential association rules from the semantically parsed codes.

[0011] Furthermore, the modeled target scene is refined into sub-scenes with unique feature sets, specifically including:

[0012] Extracting feature data sets for cluster analysis from the modeled target scene;

[0013] Use the K-means clustering algorithm to divide the feature data set into K clusters, and randomly select M feature data from the feature data set as the initial sub-scene center, denoted as μ1, μ2, ...., μ m ;

[0014] For each point in the feature dataset, that is, an element x in the scene i , using the Euclidean distance to calculate its distance to each sub-scene center μ m The distance d(x i ,μ m ), the calculation formula is as follows:

[0015]

[0016] where x i represents a point in the feature data set; μ m represents the center of the mth sub-scene; x ij Represents point x i The jth feature of mj represents the sub-scene center μ m The jth feature of ; n represents the dimension of the feature;

[0017] For each point x i Assign to the cluster to which the nearest cluster center belongs, the formula is as follows:

[0018] c i = arg min m d(x i ,μ m )

[0019] Among them, c i Represents point x i The cluster labels of

[0020] Update each sub-scene center μ m It has a unique feature set, and the calculation formula is as follows:

[0021]

[0022] Among them, Cm Represents the set of all feature data of the i-th sub-scene.

[0023] Furthermore, the S2 specifically includes:

[0024] S201, select a feature extraction algorithm based on the characteristics of the sub-scenario and subsequent requirements;

[0025] S202, performing data conversion and data normalization on the sub-scene data to ensure data quality and consistency;

[0026] S203, inputting the sub-scene data after data conversion and data normalization into the selected feature extraction algorithm to extract key feature vectors;

[0027] S204, organizing the extracted key feature vectors into a feature library for use in multi-view code analysis and mining processes.

[0028] Furthermore, the feature extraction algorithm is a principal component analysis method, and the sub-scene data is converted into a linearly independent representation of each dimension through a linear transformation, and a feature vector with a maximum eigenvalue in the sub-scene data is found based on the linearly independent representation;

[0029] The principal component analysis method comprises calculating the covariance matrix of the sub-scene data, solving the eigenvalue and eigenvector of the covariance matrix, and selecting the eigenvector with the largest eigenvalue as the principal component;

[0030] The number of samples obtained is s, and the dimension is n sub-scenes;

[0031] Store the sub-scene samples in an s×n matrix X, where each row represents a sample and each column represents a feature; calculate the sample mean vector The formula is as follows:

[0032]

[0033] in, represents an n-dimensional vector, each element of which is the mean of the corresponding feature; y i represents the i-th row of matrix X, i.e. the i-th sample;

[0034] Calculate the covariance matrix C, the formula is as follows;

[0035]

[0036] Among them, the covariance matrix C is an n×n matrix, whose elements C ij represents the covariance between the i-th feature and the j-th feature; s represents the number of samples; x i represents the i-th sample vector; represents the mean value of the sample vector;

[0037] Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ1,λ2,λ3,....,λ n and the corresponding eigenvectors ω1,ω2,ω3,....,ω n , the formula for eigenvalue decomposition is:

[0038] C ω =λω;

[0039] Where λ is the eigenvalue and ω is the corresponding eigenvector;

[0040] Select the eigenvectors corresponding to the first k largest eigenvalues ​​as principal components, and arrange the corresponding eigenvectors into a matrix P by columns; project the sub-scene data onto the selected principal components to obtain the sub-scene data after dimensionality reduction;

[0041] Y=P T (X-1 s μ T );

[0042] Among them, 1 s represents an s×1 column vector, Y represents the matrix after dimensionality reduction, μ T Represents the transposed mean vector of each feature of the sub-scene data, all elements of which are 1.

[0043] Furthermore, the S3 specifically includes:

[0044] S301, obtaining code data under multiple scene views;

[0045] S302, analyzing the code data using lexical analysis, syntactic analysis, pragmatic analysis, and context analysis, and semantically annotating each element in the code data based on the analysis results;

[0046] S303, based on the constructed feature library, matching and associating the key feature vectors in the feature library with the semantically annotated code data;

[0047] S304, scanning the code data after feature association annotation, finding frequently occurring semantic items, and obtaining a candidate item set;

[0048] S305, using the prior principle of the FP-Growth algorithm, removing semantic items that are unlikely to become frequent items from the candidate item set, and obtaining a frequent item set;

[0049] S306, setting a minimum support threshold and a confidence threshold through cross-validation;

[0050] S307, extracting association rules that meet the minimum support and confidence thresholds from the frequent item set.

[0051] Furthermore, the calculation formula for extracting the items satisfying the minimum support from the frequent item set is as follows:

[0052]

[0053] Among them, count(x) represents the number of times the frequent item set x appears, and N represents the total number of samples analyzed for the code data;

[0054] Based on the frequent item set Support(x), analyze all possible non-empty true subsets X and generate corresponding association rules:

[0055] X→(Support(x)-X)

[0056] Calculate the confidence of the association rules, and select strong association rules by using the minimum confidence threshold set by cross-validation:

[0057]

[0058] Support(X∪Y) represents the proportion of transactions that contain both X and Y in the sub-scenario dataset; Support(X) represents the proportion of transactions that contain a non-empty true subset X in the sub-scenario dataset;

[0059] The confidence of each rule is calculated. If the confidence is greater than or equal to the minimum confidence threshold, the rule is considered a valid association rule, otherwise it is discarded.

[0060] On the other hand, the MOM scene feature extraction and multi-view code semantic association mining device includes:

[0061] The target scenario modeling module models the target scenario through the MOM system based on the code data and document data of the target scenario, and refines the modeled target scenario into sub-scenarios with unique feature sets;

[0062] The feature library construction module extracts key features from sub-scenes with unique feature sets through feature extraction algorithms, and constructs a feature library based on the extracted key features;

[0063] The association rule mining module retrieves key feature vectors based on the feature library, performs semantic analysis on the codes under multiple scenario views based on the key feature vectors, and uses the association rule mining algorithm to mine potential association rules from the semantically analyzed codes.

[0064] The present invention adopts the above technical solution and has the following beneficial effects:

[0065] (1) The present invention models the target scenario in a model-driven manner, and performs semantic analysis and association rule mining on the code under multiple views, which can fully reveal the semantic associations between codes and improve the accuracy and comprehensiveness of code analysis;

[0066] (2) The present invention subdivides the modeled scenario into multiple sub-scenarios, extracts key features from the sub-scenario using a feature extraction algorithm, accurately identifies the most critical features for understanding and analyzing the scenario, and provides strong support for subsequent code analysis and optimization;

[0067] (3) The present invention stores the extracted key features in a feature library and establishes an efficient indexing and query mechanism, so that the key features can be quickly retrieved and used, thereby improving the efficiency and accuracy of code analysis. At the same time, the feature library can also provide valuable reference information for subsequent software development and code optimization;

[0068] (4) The present invention uses a multi-view code analysis tool to perform in-depth semantic analysis on the code, extract key elements and their attributes in the code, and then perform association analysis on the parsed code elements by applying an association rule mining algorithm to discover potential semantic association rules between codes and reveal deep-level associations between codes, providing strong support for code optimization and reconstruction;

[0069] (5) The present invention has good scalability and can adapt to software development projects of different scales and complexities, and can be customized and optimized according to specific needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 The MOM scene feature extraction and multi-view code semantic association mining method process of the embodiment of the present invention;

[0071] Figure 2 A schematic diagram of the MOM scenario modeling and sub-scenario refinement process according to an embodiment of the present invention;

[0072] Figure 3 A schematic diagram of the feature extraction and feature library construction process of an embodiment of the present invention;

[0073] Figure 4 A schematic diagram of a multi-view code semantic association mining process according to an embodiment of the present invention;

[0074] Figure 5 This is a diagram of a device for extracting MOM scene features and mining multi-view code semantic associations according to an embodiment of the present invention. DETAILED DESCRIPTION

[0075] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0076] See also Figure 1 As shown, the MOM scene feature extraction and multi-view code semantic association mining method of the present invention includes the following steps

[0077] S1, based on the code data and document data of the target scenario, the target scenario is modeled through the MOM system, and the modeled target scenario is refined into sub-scenarios with unique feature sets. For details, see Figure 2 ;

[0078] Specifically, the modeled target scene is refined into sub-scenes with unique feature sets, specifically including:

[0079] Extracting feature data sets for cluster analysis from the modeled target scene;

[0080] Use the K-means clustering algorithm to divide the feature data set into K clusters, and randomly select M feature data from the feature data set as the initial sub-scene center, denoted as μ1, μ2, ...., μ m ;

[0081] For each point in the feature dataset, that is, an element x in the scene i , using the Euclidean distance to calculate its distance to each sub-scene center μ m The distance d(x i ,μ m ), the calculation formula is as follows:

[0082]

[0083] where x i represents a point in the feature data set; μ m represents the center of the mth sub-scene; x ij Represents point x i The jth feature of mj represents the sub-scene center μ m The jth feature of ; n represents the dimension of the feature;

[0084] For each point x i Assign to the cluster to which the nearest cluster center belongs, the formula is as follows:

[0085] c i = arg min m d(x i ,μ m )

[0086] Among them, c i Represents point x i The cluster labels of

[0087] Update each sub-scene center μ mIt has a unique feature set, and the calculation formula is as follows:

[0088]

[0089] Among them, C m Represents the set of all feature data of the i-th sub-scene.

[0090] In addition, the present invention utilizes the modeling tool of the MOM system to model and express the target scene.

[0091] S2, extract key features from sub-scenes with unique feature sets through feature extraction algorithms, and build a feature library based on the extracted key features. For details, see Figure 3 ;

[0092] The S2 specifically includes:

[0093] S201, select a feature extraction algorithm based on the characteristics of the sub-scenario and subsequent requirements;

[0094] S202, performing data conversion and data normalization on the sub-scene data to ensure data quality and consistency;

[0095] S203, inputting the sub-scene data after data conversion and data normalization into the selected feature extraction algorithm to extract key feature vectors;

[0096] S204, organizing the extracted key feature vectors into a feature library for use in multi-view code analysis and mining processes.

[0097] Specifically, the feature extraction algorithm is a principal component analysis method, and the sub-scene data is converted into a linearly independent representation of each dimension through a linear transformation, and a feature vector with a maximum eigenvalue in the sub-scene data is found based on the linearly independent representation;

[0098] The principal component analysis method includes calculating a covariance matrix of sub-scene data, solving eigenvalues ​​and eigenvectors of the covariance matrix, and selecting an eigenvector with a maximum eigenvalue as a principal component;

[0099] The number of samples obtained is s, and the dimension is n sub-scenes;

[0100] Store the sub-scene samples in an s×n matrix X, where each row represents a sample and each column represents a feature; calculate the sample mean vector The formula is as follows:

[0101]

[0102] in, represents an n-dimensional vector, each element of which is the mean of the corresponding feature; yi represents the i-th row of matrix X, i.e. the i-th sample;

[0103] Calculate the covariance matrix C, the formula is as follows;

[0104]

[0105] Among them, the covariance matrix C is an n×n matrix, whose elements C ij represents the covariance between the i-th feature and the j-th feature; s represents the number of samples; x i represents the i-th sample vector; represents the mean value of the sample vector;

[0106] Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ1,λ2,λ3,....,λ n and the corresponding eigenvectors ω1,ω2,ω3,....,ω n , the formula for eigenvalue decomposition is:

[0107] C ω =λω;

[0108] Where λ is the eigenvalue and ω is the corresponding eigenvector;

[0109] Select the eigenvectors corresponding to the first k largest eigenvalues ​​as principal components, and arrange the corresponding eigenvectors into a matrix P by columns; project the sub-scene data onto the selected principal components to obtain the sub-scene data after dimensionality reduction;

[0110] Y=P T (X-1 s μ T );

[0111] Among them, 1 s represents an s×1 column vector, Y represents the matrix after dimensionality reduction, μ T Represents the transposed mean vector of each feature of the sub-scene data, all elements of which are 1.

[0112] Specifically, the present invention determines key features by analyzing the feature set of each sub-scene; then uses a feature extraction algorithm to extract key features from the sub-scene; finally, stores the extracted key features in a feature library, and establishes an index and query mechanism.

[0113] Specifically, the eigenvector (main variation direction) with the largest eigenvalue can retain the information in the original scene sub-data to the greatest extent.

[0114] S3, retrieves key feature vectors based on the feature library, performs semantic analysis on the code under multiple scene views based on the key feature vectors, and uses the association rule mining algorithm to mine potential association rules from the semantically parsed code. For details, see Figure 4 .

[0115] The S3 specifically includes:

[0116] S301, obtaining code data under multiple scene views;

[0117] S302, analyzing the code data using lexical analysis, syntactic analysis, pragmatic analysis, and context analysis, and semantically annotating each element in the code data based on the analysis results;

[0118] S303, based on the constructed feature library, matching and associating the key feature vectors in the feature library with the semantically annotated code data;

[0119] S304, scanning the code data after feature association annotation, finding frequently occurring semantic items, and obtaining a candidate item set;

[0120] S305, using the prior principle of the FP-Growth algorithm, removing semantic items that are unlikely to become frequent items from the candidate item set, and obtaining a frequent item set;

[0121] S306, setting a minimum support threshold and a confidence threshold through cross-validation;

[0122] S307, extracting association rules that meet the minimum support and confidence thresholds from the frequent item set.

[0123] Using the prior principle of the FP-Growth algorithm, remove the semantic items that cannot become frequent items in the candidate item set to obtain the frequent item set. The calculation formula is as follows:

[0124]

[0125] Where count(x) represents the number of times the frequent item set x appears, and N represents the total number of samples in the analysis of code data;

[0126] Based on the frequent item set Support(x), analyze all possible non-empty true subsets X and generate corresponding association rules:

[0127] X→(Support(x)-X)

[0128] Calculate the confidence of the association rules, and select strong association rules by using the minimum confidence threshold set by cross-validation:

[0129]

[0130] Support(X∪Y) represents the proportion of transactions that contain both X and Y in the sub-scenario dataset; Support(X) represents the proportion of transactions that contain a non-empty true subset X in the sub-scenario dataset;

[0131] The confidence of each rule is calculated. If the confidence is greater than or equal to the minimum confidence threshold, the rule is considered a valid association rule, otherwise it is discarded.

[0132] See also Figure 5 As shown, this embodiment also discloses a MOM scene feature extraction and multi-view code semantic association mining device, including:

[0133] A target scenario modeling module 51 models the target scenario through the MOM system based on the code data and document data of the target scenario, and refines the modeled target scenario into sub-scenarios with unique feature sets;

[0134] A feature library construction module 52 extracts key features from a sub-scene with a unique feature set through a feature extraction algorithm, and constructs a feature library based on the extracted key features;

[0135] The association rule mining module 53 performs semantic analysis on the codes under multiple scene views through multi-view code analysis, and uses an association rule mining algorithm to mine potential association rules from the semantically analyzed codes.

[0136] The specific implementation of a MOM scene feature extraction and multi-view code semantic association mining device is the same as the MOM scene feature extraction and multi-view code semantic association mining method, and this embodiment will not be repeated.

[0137] Although the present invention has been specifically shown and described in conjunction with the preferred embodiments, it should be understood by those skilled in the art that various changes may be made to the present invention in form and details without departing from the spirit and scope of the present invention as defined by the appended claims, all of which are within the scope of protection of the present invention.

Claims

1. A method for extracting MOM scene features and mining multi-view code semantic associations, characterized in that: include: S1, based on the code data and document data of the target scenario, the target scenario is modeled through the MOM system, and the modeled target scenario is refined into sub-scenarios with unique feature sets; S2, extract key features from sub-scenes with unique feature sets through feature extraction algorithms, and build a feature library based on the extracted key features; S3, retrieves key feature vectors based on the feature library, performs semantic analysis on the codes under multiple scenario views based on the key feature vectors, and uses the association rule mining algorithm to mine potential association rules from the semantically parsed codes.

2. The MOM scene feature extraction and multi-view code semantic association mining method according to claim 1, characterized in that: In S1, the modeled target scene is refined into sub-scenes with unique feature sets, specifically including: Extracting feature data sets for cluster analysis from the modeled target scene; Use the K-means clustering algorithm to divide the feature data set into K clusters, and randomly select M feature data from the feature data set as the initial sub-scene center, denoted as μ1, μ2, ...., μ m ; For each point in the feature dataset, that is, an element x in the scene i , using the Euclidean distance to calculate its distance to each sub-scene center μ m The distance d(x i ,μ m ), the calculation formula is as follows: where x i represents a point in the feature data set; μ m represents the center of the mth sub-scene; x ij Represents point x i The jth feature of mj represents the sub-scene center μ m The jth feature of ; n represents the dimension of the feature; For each point x i Assign to the cluster to which the nearest cluster center belongs, the formula is as follows: c i =very little m d(x i ,μ m ) Among them, c i Represents point x i The cluster labels of Update each sub-scene center μ m It has a unique feature set, and the calculation formula is as follows: Among them, C m Represents the set of all feature data of the i-th sub-scene.

3. The MOM scene feature extraction and multi-view code semantic association mining method according to claim 1, characterized in that, The S2 specifically includes: S201, select a feature extraction algorithm based on the characteristics of the sub-scenario and subsequent requirements; S202, performing data conversion and data normalization on the sub-scene data to ensure data quality and consistency; S203, inputting the sub-scene data after data conversion and data normalization into the selected feature extraction algorithm to extract key feature vectors; S204, organizing the extracted key feature vectors into a feature library for use in multi-view code analysis and mining processes.

4. The MOM scene feature extraction and multi-view code semantic association mining method according to claim 3, characterized in that, The feature extraction algorithm is a principal component analysis method, and the sub-scene data is converted into a linearly independent representation of each dimension through a linear transformation, and a feature vector with a maximum eigenvalue in the sub-scene data is found based on the linearly independent representation; The principal component analysis method comprises calculating the covariance matrix of the sub-scene data, solving the eigenvalue and eigenvector of the covariance matrix, and selecting the eigenvector with the largest eigenvalue as the principal component; The number of samples obtained is s, and the dimension is n sub-scenes; Store the sub-scene samples in an s×n matrix X, where each row represents a sample and each column represents a feature; Calculate the sample mean vector The formula is as follows: in, represents an n-dimensional vector, each element of which is the mean of the corresponding feature; y i represents the i-th row of matrix X, i.e. the i-th sample; Calculate the covariance matrix C, the formula is as follows; Among them, the covariance matrix C is an n×n matrix, whose elements C ij represents the covariance between the i-th feature and the j-th feature; s represents the number of samples; x i represents the i-th sample vector; represents the mean value of the sample vector; Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​λ1,λ2,λ3,....,λ n and the corresponding eigenvectors ω1,ω2,ω3,....,ω n , the formula for eigenvalue decomposition is: C ω =I say; Where λ is the eigenvalue and ω is the corresponding eigenvector; Select the eigenvectors corresponding to the first k largest eigenvalues ​​as principal components, and arrange the corresponding eigenvectors into a matrix P by column; Project the sub-scene data onto the selected principal component to obtain the sub-scene data after dimensionality reduction; Y=P T (X-1 s μ T ); Among them, 1 s represents an s×1 column vector, Y represents the matrix after dimensionality reduction, μ T Represents the transposed mean vector of each feature of the sub-scene data, all elements of which are 1.

5. The MOM scene feature extraction and multi-view code semantic association mining method according to claim 1, characterized in that, The S3 specifically includes: S301, obtaining code data under multiple scene views; S302, analyzing the code data using lexical analysis, syntactic analysis, pragmatic analysis, and context analysis, and semantically annotating each element in the code data based on the analysis results; S303, based on the constructed feature library, matching and associating the key feature vectors in the feature library with the semantically annotated code data; S304, scanning the code data after feature association annotation, finding frequently occurring semantic items, and obtaining a candidate item set; S305, using the prior principle of the FP-Growth algorithm, removing semantic items that are unlikely to become frequent items from the candidate item set, and obtaining a frequent item set; S306, setting a minimum support threshold and a confidence threshold through cross-validation; S307, extracting association rules that meet the minimum support and confidence thresholds from the frequent item set.

6. The MOM scene feature extraction and multi-view code semantic association mining method according to claim 5, characterized in that, The calculation formula for extracting the minimum support from the frequent item set is as follows: Among them, count(x) represents the number of times the frequent item set x appears, and N represents the total number of samples analyzed for the code data; Based on the frequent item set Support(x), analyze all possible non-empty true subsets X and generate corresponding association rules: X→(Support(x)-X) Calculate the confidence of the association rules, and select strong association rules by using the minimum confidence threshold set by cross-validation: Support(X∪Y) represents the proportion of transactions that contain both X and Y in the sub-scenario dataset; Support(X) represents the proportion of transactions that contain a non-empty true subset X in the sub-scenario dataset; The confidence of each rule is calculated. If the confidence is greater than or equal to the minimum confidence threshold, the rule is considered a valid association rule, otherwise it is discarded. 7.MOM scene feature extraction and multi-view code semantic association mining device, characterized in that: include: The target scenario modeling module models the target scenario through the MOM system based on the code data and document data of the target scenario, and refines the modeled target scenario into sub-scenarios with unique feature sets; The feature library construction module extracts key features from sub-scenes with unique feature sets through feature extraction algorithms, and constructs a feature library based on the extracted key features; The association rule mining module retrieves key feature vectors based on the feature library, performs semantic analysis on the codes under multiple scenario views based on the key feature vectors, and uses the association rule mining algorithm to mine potential association rules from the semantically analyzed codes.

Images