A multi-label causal feature selection method and system

Abstract

The application provides a multi-label causal feature selection method and system, and the method comprises the following steps: obtaining a multi-label gene expression dataset, and preprocessing the multi-label dataset; based on the preprocessed multi-label gene expression dataset, using a unary approximation Markov blanket to find an approximation Markov blanket feature for each label; deleting other labels contained in the approximation Markov blanket of the label, and restoring equivalent features to obtain Markov blanket features of the label; using a symmetry test method to ensure that the selected features belong to the Markov blanket of the label; merging the Markov blanket features of all labels to obtain a gene expression data feature subset, and realizing feature selection of the gene expression data. The application adopts a feature selection method based on a causal structure and a unary approximation Markov blanket, and restores features that are mistakenly deleted due to equivalent information of the label, so that the problem of low accuracy of existing gene expression data feature selection methods is solved from the root.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing, and particularly relates to a multi-label causal feature selection method and system. BACKGROUND

[0002] At present, multi-label classification learning has been widely applied in the biomedical field, and gene expression data is commonly used in the biomedical field. However, with the rapid development, the dimension and quantity of the collected multi-label gene expression data are explosively increasing, and if not properly handled, the problem of "dimension disaster" will be faced, therefore, multi-label feature selection of gene expression data has attracted extensive attention. A large number of multi-label feature selection methods of gene expression data have been proposed by many researchers at home and abroad, according to the combination mode of the methods and learning models, the feature selection methods can be summarized as filter method, wrapper method and embedded method.

[0003] However, most of the existing multi-label feature selection methods of gene expression data only consider the correlation of variables, and do not consider the causality, resulting in lack of interpretability. However, the Markov blanket correlation algorithm has been widely applied in single-label tasks, but due to the complexity of multi-label gene expression data, there are few Markov algorithms for multi-label at present. At the same time, the existing feature selection methods ignore the feature omission caused by equivalent information when searching for the Markov blanket of multi-label. Therefore, the existing multi-label feature selection methods of gene expression data have the problem of inaccurate feature selection.

[0004] Therefore, how to improve the feature selection accuracy of gene expression data is a problem to be solved in the field. SUMMARY

[0005] The application aims at the defects of the prior art, and provides a multi-label causal feature selection method and system. The application adopts a feature selection method based on a causal structure and a unary approximate Markov blanket, considers three relationships between labels and features, features and features, and labels and labels, adds a causal relationship on the basis of the unary approximate Markov blanket, so that a complete Markov blanket can be found, and the features which are mistakenly deleted due to equivalent information of labels are restored, thereby fundamentally solving the problem of low accuracy of existing gene expression data feature selection methods.

[0006] To achieve the above object, the application adopts the following technical scheme:

[0007] The application provides a multi-label causal feature selection method, which specifically comprises the following steps.

[0008] S1, a multi-label data set is acquired, and the multi-label data set is preprocessed, wherein the multi-label data set is a multi-label gene expression data set;

[0009] S2, based on the preprocessed multi-label gene expression data set, an approximate Markov blanket is searched for each label by using a unary approximate Markov blanket; the specific steps comprise the following steps.

[0010] S21, a threshold value is determined by calculating the uncertainty between each label and a feature; specifically, the symmetric uncertainty between each label and each feature is calculated, and the average value of the top t% of all symmetric uncertainties is taken as the threshold value alpha, wherein t is a set threshold value;

[0011] S22, it is judged whether the feature is related to the label, if the symmetric uncertainty between the label and the feature is greater than alpha, it is indicated that the label and the feature are related, if the symmetric uncertainty is less than alpha, it is indicated that the label and the feature are not related, and the feature is deleted from the feature set;

[0012] S23, a unary approximate Markov blanket is used to delete the false MB feature in the feature set obtained in the step S22;

[0013] S24, a symmetry test is used to delete the false MB feature in the feature set obtained in the step S23, and the approximate Markov blanket of each label is obtained;

[0014] The unary approximate Markov blanket is specifically:

[0015] In a Bayesian network, given a label T i , two features F i , F j , if the following formula is satisfied:

[0016]

[0017] SU(F i ; T i ) > SU(F j ; T i ) && SU(F i ; F j ) > SU(F j ; T i ) (2)

[0018] I(T i ; F j | F i ) < I(T i ; F j ) (3)

[0019] For F j , F i is a unary approximate Markov blanket of T i ;

[0020] Wherein, SU represents the symmetric uncertainty, which is normalized by mutual information, representing the correlation between two variables, I(T i ; F i ) represents the size of mutual information between two variables, and I(T i ; F j | F i ) represents the conditional mutual information size between T i and F j given F i , which measures the degree of influence of F i on the correlation between T i and F j ; if for F j , F i is a unary approximate Markov blanket of T i , at this time F j must not be an MB feature of T i , so F j is deleted;

[0021] S3, delete other labels contained in the approximate Markov blanket of the label, and restore the equivalent features to obtain the Markov blanket features of the label;

[0022] S4, using a fast symmetry checking method to ensure that the features recovered in step S3 belong to the Markov blanket of the label, features passing the fast symmetry checking method will be retained, and features failing to pass will be deleted;

[0023] S5, merging the Markov blanket features of all labels to obtain a feature subset of the gene expression data, and realizing feature selection of the gene expression data.

[0024] Further, in step S1, the preprocessing includes data cleaning and discrete coding operation on different data.

[0025] Further, step S3 is specifically: deleting other labels contained in the approximate Markov blanket of the label, and recovering features mistakenly deleted due to equivalent information of the label, to obtain the Markov blanket features of the label.

[0026] Further, step S4 is specifically: if a feature X is part of the Markov blanket of a label T, then T should also be part of the Markov blanket of X; the symmetry ensures consistency and mutual inclusion between features and labels.

[0027] Further, after step S5, based on the feature subset of the gene expression data obtained by the multi-label causal feature selection method, a gene classification model is trained to obtain a trained gene classification model; the gene data to be classified is obtained, and based on the feature subset of the gene data to be classified, the trained gene classification model is used to classify the gene data to be classified, to obtain the label of the gene data to be classified.

[0028] The application also provides a multi-label causal feature selection system, characterized in that the feature selection system executes the multi-label causal feature selection method, and includes a data acquisition and preprocessing module, an approximate Markov blanket feature acquisition module, a Markov blanket feature determination module, a feature verification module, and a feature merging module.

[0029] The data acquisition and preprocessing module acquires a multi-label data set and pre-processes the multi-label data set, wherein the multi-label data set is a multi-label gene expression data set.

[0030] The approximate Markov blanket feature acquisition module acquires approximate Markov blanket features for each label based on the pre-processed multi-label gene expression data set by using a unary approximate Markov blanket.

[0031] The Markov blanket feature determination module deletes other labels contained in the approximate Markov blanket of the label, and recovers equivalent features to obtain Markov blanket features of the label.

[0032] The feature verification module uses a fast symmetry test method to ensure that the restored features belong to the Markov blanket of the label, features passing the fast symmetry test method will be retained, and features failing to pass will be deleted;

[0033] The feature merging module merges the Markov blanket features of all labels to obtain a feature subset of the gene expression data, thereby realizing feature selection of the gene expression data.

[0034] The application further provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor realizes the method when executing the computer program.

[0035] Compared with the prior art, the application has the following beneficial effects:

[0036] The multi-label causal feature selection method of the application is based on a preprocessed multi-label gene expression data set, uses a unary approximate Markov blanket to find an approximate Markov blanket feature for each label, deletes other labels contained in the approximate Markov blanket of the label, restores equivalent features, obtains Markov blanket features of the label, uses a symmetry test method to ensure that the selected features belong to the Markov blanket of the label, merges the Markov blanket features of all labels to obtain a feature subset of the gene expression data, and realizes feature selection of the gene expression data. The application adopts a feature selection method based on a causal structure and a unary approximate Markov blanket, considers three relationships between labels and features, features and features, and labels and labels, adds a causal relationship on the basis of the unary approximate Markov blanket, so that a complete Markov blanket can be found, and the features that are mistakenly deleted due to equivalent information of the label are restored, the accuracy of feature selection of the gene expression data is improved, and the accuracy of a gene classification model is improved. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 The multi-label causal feature selection method provided for the embodiments of the application is shown in the figure.

[0039] Figure 2 The structure of the existing Markov blanket provided for the embodiments of the application is shown in the figure.

[0040] Figure 3 The equivalent information situation provided for the embodiments of the application is shown in the figure.

[0041] Figure 4 A multi-label causal feature selection system schematic diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0043] It should be understood that when used in the specification and the appended claims, the terms "comprise" and "include" indicate the presence of described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0044] The present application will be further described below with reference to the drawings and specific embodiments, but not as a limitation of the present application.

[0045] It should also be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0046] It should be further understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0047] The present application proposes a multi-label causal feature selection method. As shown in the figure, the multi-label causal feature selection method comprises the following steps S1 to S5. Figure 1

[0048] S1, a multi-label data set is obtained, and the multi-label data set is preprocessed, wherein the multi-label data set is a multi-label gene expression data set.

[0049] In step S1, the preprocessing includes data cleaning and discretization coding operation on different data.

[0050] ​In specific embodiments, different data sources come from different devices and collection platforms, and device failure or insufficient device performance can lead to problems such as missing data, recording errors, and duplicate data. Therefore, the data cleaning includes correction of error data and abnormal data, filling of missing road section data, and removal of redundant data.

[0051] Since mutual information is used to measure the causality between labels, between features, and between features and labels when feature selection is performed later, the use of mutual information requires that the dataset be discretized first. Therefore, different data are discretized and encoded.

[0052] In an embodiment, the multi-label dataset is a VirusGO virus dataset, which is used to predict the subcellular location of a protein from its sequence. It contains the sequences of 207 virus species. GO (Gene Ontology) features and PseAAC (including 20 amino acids, 20 pseudo-amino acids, and 400 dipeptide components) are provided. There are 6 subcellular locations (virus capsid, host cell membrane, host endoplasmic reticulum, host cytoplasm, host nucleus, and secretion).

[0053] In addition, in another embodiment, the multi-label dataset can be a GpositiveGO dataset, which is used to predict the subcellular location of a protein from its sequence. It contains the sequences of 519 gram-positive species. GO (Gene ontology) features and PseAAC (including 20 amino acids, 20 pseudo-amino acids, and 400 dipeptide components) are provided; there are 4 subcellular locations (cell membrane, cell wall, cytoplasm, and extracellular).

[0054] The characteristics of the relevant datasets are listed in the following table, including the number of instances (m), the number of features (d), the number of labels (q), the cardinality (Card), the density (Dens), the diversity (Div), and the average imbalance ratio (avgIR) for each label. The cardinality measures the average number of labels associated with each instance, and the density is defined as the cardinality divided by the number of labels. The diversity represents the percentage of the set of labels present in the dataset divided by the number of possible label sets. The avgIR measures the average imbalance of all labels, and the larger the avgIR, the more imbalanced the dataset.

[0055]

[0056] S2, based on the preprocessed multi-label gene expression dataset, using a unary approximation Markov blanket to find an approximate Markov blanket for each label;

[0057] This step is mainly to make a preliminary feature selection on the pre-processed features in the last step. Assume that D = (F, T) represents a d-dimensional instance space, where F = {F1, F2, …, Fm} represents a feature space with m possible features, and T = {T1, T2, …, Tq} represents a label space with q possible labels. Our goal is to select fewer features so that the classifier performs better. m} represents a feature space with m possible features, and T = {T1, T2, …, T q} represents a label space with q possible labels. Our goal is to select fewer features so that the classifier performs better.

[0058] Unlike single-label feature selection, multi-label feature selection needs to consider the relationships between labels, between labels and features, and between features. We use the idea of first finding the causal features of each label in multi-label data, and then merging them. Therefore, in order to take into account the various relationships between labels and features, when we first find the causal features of each label in multi-label data, we treat the remaining labels as feature variables to determine whether these labels are related to our label, and these relationships may cause some features to be misjudged.

[0059] When finding the causal features of a single label, we find the Markov blanket of the label. The existing structure of the Markov blanket is shown in Figure 2 As shown in Figure 2 , the Markov blanket refers to the set of parent nodes (A, B) around a variable X, child nodes (C, D), and spouse nodes E, which represent all the information that can represent X. Therefore, we need to find the causal features of label X, which can be equivalent to finding the Markov blanket features of X.

[0060] However, the traditional Markov blanket method will miss some of the multi-label data sets, so this application proposes to first use a unary approximation Markov blanket to find the approximate Markov blanket features of each label. Through the unary approximation Markov blanket, we can find the small number of Markov blanket features we need, thereby improving the accuracy of feature selection and further improving the accuracy and efficiency of the classification model.

[0061] The specific steps of step S2 include:

[0062] S21, determine the threshold by calculating the uncertainty between each label and feature; specifically, calculate the symmetric uncertainty between each label and feature, and take the average of the top t% of all symmetric uncertainties as the threshold a, where t is a set threshold, and the value is [0.01, 0.05];

[0063] where the threshold a represents the correlation between the label and the feature.

[0064] S22, judging whether the feature is related to the label, if the symmetry uncertainty between the label and the feature is greater than a, it indicates that the label is related to the feature, if less than a, it indicates that the label is not related to the feature, and the feature is deleted from the feature set;

[0065] Specifically, whether the feature is related to the label is judged, and specifically, considering that the labels of the multi-label will cause influence, when the related feature of a label is searched, the remaining labels are also added to the feature set as ordinary features; if the symmetry uncertainty between the label and the feature is greater than a, it indicates that the label is related to the feature, if less than a, it indicates that the label is not related to the feature, and the feature is deleted from the feature set.

[0066] S23, using a unary approximation Markov blanket to delete the false MB feature in the feature set obtained in step S22;

[0067] The false MB feature is deleted by using the unary approximation Markov blanket, so that the remaining features and the label form the Markov blanket structure required by us.

[0068] S24, using symmetry test to delete the false MB feature in the feature set obtained in step S23, to obtain the approximate Markov blanket of each label.

[0069] The unary approximation Markov blanket is specifically:

[0070] In a Bayesian network, given a label T i , two features F i , F j , if the following formula is satisfied:

[0071]

[0072] SU(F i ; T i ) > SU(F j ; T i ) && SU(F i ; F j ) > SU(F j ; T i ) (2)

[0073] I(T i ; F j | F i ) < I(T i ; F j ) (3)

[0074] For F j , F i is T ia one-approximate Markov blanket of T.

[0075] where SU represents the symmetric uncertainty, is normalized by mutual information, and represents the correlation between two variables, I(T i ; F j ) represents the size of mutual information between two variables, and I(T i ; F i | F j ) represents the conditional mutual information size between T i and F i given F i , which measures the degree of influence of F j on the correlation between T i and F i ; if F j is a one-approximate Markov blanket feature of T j for F i , then F i must not be an MB feature of T j , so F i is deleted.

[0076] In a faithful Bayesian network, if there is a F j feature that is a one-approximate Markov blanket feature of T i for F j , then F i must not be an MB feature of T j , so F i is deleted; we can illustrate this by assuming that F j is a one-approximate Markov blanket feature of T j for F i and F i belongs to the MB features of T j , then it will contradict our formula.

[0077] Case 1: F i is a parent-child subset feature in the MB features of T j . At this time, there are three cases for the relative positions among F i , F i , and T j : 1. F i is connected to T i through F j and T i -F i -F j , 2. F i is connected to F i through T i and F j Fj -T i -F i , 3, F i -XT i -F j ,F j Through the remaining variables and T i We can find that in cases 1 and 3, due to T i With F i Directly connected, T i With F i Indirectly connected, so T i With F j The correlation between i With F i The correlation between , thus violating our formula 2 SU (F i ,T i )>SU(F j ,T i ), Similarly, in the second case, due to T i With F j The correlation between them is greater than F i With F j The correlation between them violates the formula SU(F i ,F j )>SU(F j ,T i ). Therefore, for F j In terms of i T i When the unary approximate Markov blanket characteristics of F j Impossible for T i The parent-child set features in the MB features.

[0078] Case 2: F j T i When the spouse set feature is in the MB feature, according to the definition of the spouse set feature, we can always find an F i Can make F j Satisfy F j →F i ←T i The structure has a characteristic, that is, when F is given i When F j With T i The correlation between them becomes larger, thus violating our formula 3. So F j It can't be us T i The spouse set features in the MB features.

[0079] In summary, for F j In terms of F iFor T i 's unary approximation Markov blanket features, F j j → F i → T i , F j → T i This special structure, where F j may still be a parent of T i . So a formula 1 is set, that is, only when the correlation between F i and F j is greater than the average threshold δ, it means that the redundancy between F i and F j is too large, and we will make the deletion decision of F j .

[0080] S3, delete other labels contained in the approximation Markov blanket of the label, and restore the equivalent features to obtain the Markov blanket features of the label.

[0081] Further, step S3 is specifically: deleting other labels contained in the approximation Markov blanket of the label, and restoring the features that are mistakenly deleted due to the equivalent information of the label, to obtain the Markov blanket features of the label.

[0082] Step S3 is mainly to delete the case that the Markov blanket of the label in step S2 contains other labels, and at the same time, due to the fact that the real data often violates the assumption of fidelity, the equivalent information appears, which further leads to the false deletion of part of the true Markov blanket features. When the equivalent information case as shown in Figure 3 is present, T j and {F i , F j} both contain the same information of T i , that is, SU(T i , T j ) = SU(T i , {F i , F j}), but F i and F j will be mistakenly deleted due to the first stage algorithm, and finally lead to when we delete the label T i in the feature set selected by the label T j , F i and F j ​Also no selected. The application deletes other labels contained in the feature set of the label, and restores the features deleted by mistake in the deletion process due to the addition of the label, so that the multi-label feature selection process can obtain the correct Markov blanket feature set.

[0083] S4, using a fast symmetry test method to ensure that the features recovered in step S3 belong to the Markov blanket of the label, the features passing the fast symmetry test method will be retained, and the features failing to pass will be deleted.

[0084] Further, step S4 is specifically: if a feature X is part of the Markov blanket of a label T, then T should also be part of the Markov blanket of X; the symmetry ensures the consistency and mutual inclusion between the features and the labels.

[0085] It can be specifically expressed by the following formula: The symmetry can ensure the mutual influence between the features and the labels, so after recovering the features, we use the symmetry test on the labels and their Markov blanket set to ensure that the selected features are our real Markov blanket features.

[0086] S5, merging the Markov blanket features of all labels to obtain a feature subset of the gene expression data, and realizing feature selection of the gene expression data.

[0087] After the Markov blanket feature test of each label, the feature selection is not really completed, and all the features of the labels need to be merged, and the merged features are the features retained by the method.

[0088] The application processes the multi-label data, maintains the uniformity of the data caliber, encodes by using a suitable discrete coding mode, forms different types of features, then uses the unary approximate Markov blanket based method for preliminary selection for each label, selects the features belonging to the approximate Markov blanket of the label and other labels, then screens the features in the approximate Markov blanket and the labels, deletes other labels contained in the approximate Markov blanket of the label, and restores the features deleted by mistake in the deletion process due to the equivalent information of the label, then performs symmetry test on the features selected by each label to ensure the accuracy, and then merges to obtain the required features.

[0089] The present invention is specifically aimed at multi-label datasets, and implements a feature selection method CFAB based on causal structure and unary approximate Markov blanket for multi-label feature selection. This method takes into account three relationships: between labels and features, between features and features, and between labels and labels. It adds causal relationships on the basis of unary approximate Markov blanket, so that a complete Markov blanket can be found. At the same time, in response to the situation where equivalent information of labels appearing in multi-label datasets leads to the accidental deletion of features, the present invention proposes a method to delete other labels contained in the approximate Markov blanket of labels and restore equivalent features to solve the problem. Compared with traditional multi-label feature selection methods, the multi-label causal feature selection method of the present invention has higher accuracy and is more interpretable.

[0090] After step S5, the gene classification model is trained based on the feature subset of the gene expression data obtained by the multi-label causal feature selection method CFAB to obtain a trained gene classification model; the gene data to be classified is obtained, and based on the feature subset of the gene data to be classified, the gene data to be classified is classified using the trained gene classification model to obtain a label for the gene data to be classified.

[0091] The multi-label causal feature selection method CFAB is used to reduce feature dimensionality and improve the accuracy of feature selection and classification. Since our invention does not rely on a specific classifier, we randomly selected the MLKNN classifier commonly used in multi-label tasks as the gene classification model.

[0092] Specifically, during the model training process, six evaluation indicators were selected: Hamming Loss, Average Precision, Coverage, Ranking Loss, SubsetAccuracy, and F-Micro.

[0093] The proposed method CFAB is compared with six existing methods and ranked for each indicator, as shown in the table below. The data in the table below show that the proposed method CFAB has better feature selection effect on the two gene expression datasets.

[0094]

[0095]

[0096] Figure 4 This is a multi-label causal feature selection system provided by an embodiment of the present invention. Figure 4 As shown, the multi-label causal feature selection system includes a data acquisition and preprocessing module, an approximate Markov blanket feature acquisition module, a Markov blanket feature determination module, a feature verification module, and a feature merging module;

[0097] a data acquisition and preprocessing module, configured to acquire a multi-label dataset and preprocess the multi-label dataset, wherein the multi-label dataset is a multi-label gene expression dataset;

[0098] an approximate Markov blanket feature acquisition module, configured to find approximate Markov blanket features for each label based on the preprocessed multi-label gene expression dataset using a unary approximate Markov blanket;

[0099] a Markov blanket feature determination module, configured to delete other labels contained in the approximate Markov blanket of a label and restore equivalent features to obtain Markov blanket features of the label;

[0100] a feature verification module, configured to use a fast symmetry test method to ensure that the restored features belong to the Markov blanket of the label, wherein features passing the fast symmetry test method will be retained and features failing the fast symmetry test method will be deleted;

[0101] a feature merging module, configured to merge Markov blanket features of all labels to obtain a gene expression data feature subset, thereby realizing feature selection of the gene expression data.

[0102] The multi-label causal feature selection system can be implemented in the form of a computer program, which can run on a computer device.

[0103] The computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory can include a non-volatile storage medium and an internal memory.

[0104] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, which, when executed, can cause the processor to execute a multi-label causal feature selection method.

[0105] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0106] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, can cause the processor to execute a multi-label causal feature selection method.

[0107] The network interface is configured to communicate with other devices over a network. Those skilled in the art can understand that the above computer device structure is only part of the structure related to the scheme of the present application and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0108] The processor is configured to run a computer program stored in the memory, and the computer program is configured to implement the multi-label causal feature selection method in the first embodiment.

[0109] It should be understood that, in the embodiments of the present application, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0110] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above-mentioned embodiments.

[0111] The present application also provides a storage medium. The storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program is executed by a processor to make the processor execute a multi-label causal feature selection method in the first embodiment.

[0112] The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, and various computer-readable storage media that can store program codes.

[0113] It can be understood by those skilled in the art that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0114] In several embodiments provided by the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. For example, the described apparatus embodiments are merely schematic. For example, the division of the units is merely a logical function division. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In this way, the inventive idea can be implemented.

[0115] The steps in the method embodiments of the present application can be adjusted, combined and deleted in sequence according to actual needs. The units in the apparatus embodiments of the present application can be combined, divided and deleted according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0116] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0117] Note that the above are merely preferred embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and those skilled in the art can make various obvious changes, re-adjustments and substitutions without departing from the scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A multi-label causal feature selection method, characterized in that: Including steps: S1. Acquire a multi-label dataset and preprocess the multi-label dataset, wherein the multi-label dataset is a multi-label gene expression dataset; S2. Based on the preprocessed multi-label gene expression dataset, use a unary approximate Markov blanket to find an approximate Markov blanket for each label. The specific steps include: S21. Determine a threshold by calculating the uncertainty between each label and feature. Specifically, calculate the symmetric uncertainty between each label and feature, and take the average of the first t% of all symmetric uncertainties as the threshold α, where t is the set threshold. S22. Determine whether the feature is related to the label. If the symmetric uncertainty between the label and the feature is greater than α, it means that the label and the feature are related. If it is less than α, it means that the label and the feature are not related, and the feature is deleted from the feature set. S23, using a unary approximate Markov blanket to delete the false MB features in the feature set obtained in step S22; S24. Using a symmetry check, remove the false MB features in the feature set obtained in step S23 to obtain an approximate Markov blanket for each label. The unary approximate Markov blanket is specifically: In a Bayesian network, given a label T i , two features F i 、F j , if the following formula is satisfied: SU(F i ;T i )>SU(F j ;T i )&&SU(F i ;F j )>SU(F j ;T i ) (2) I(T i ;F j |F i )<I(T i ;F j ) (3) Then for F j In terms of i T i The unary approximate Markov blanket of Among them, SU represents the symmetric uncertainty, which is normalized by mutual information and represents the correlation between two variables, I(T i ; F i ) represents the size of the mutual information between two variables, and I(T i ; F j |F i ) represents a given F i After the variable, T i With F j The size of the conditional mutual information between F i For T i With F j The degree of influence of the correlation between j In terms of i T i The unary approximate Markov blanket feature, at this time F j It must not be T i MB features, so delete F j ; S3, deleting other labels contained in the approximate Markov blanket of labels and restoring equivalent features to obtain the Markov blanket features of labels; S4. Using a fast symmetry check method to ensure that the features recovered in step S3 belong to the Markov blanket of labels, features that pass the fast symmetry check method will be retained, and features that fail the fast symmetry check method will be deleted; S5. Merge the Markov blanket features of all labels to obtain a feature subset of gene expression data, thereby achieving feature selection of gene expression data.

2. The method according to claim 1, characterized in that In step S1, the preprocessing includes: data cleaning and discretization encoding operations on different data.

3. The method according to claim 1, characterized in that Step S3 specifically includes: deleting other labels contained in the approximate Markov blanket of labels, and restoring features that are mistakenly deleted due to equivalent information of labels, to obtain the Markov blanket features of labels.

4. The method according to claim 1, wherein Step S4 is specifically as follows: if a feature X is part of the Markov blanket of a label T, then T should also be part of the Markov blanket of X; the symmetry ensures the consistency and mutual inclusion between the feature and the label.

5. The method according to claim 1, wherein After step S5, the gene classification model is trained based on the gene expression data feature subset obtained by the multi-label causal feature selection method to obtain a trained gene classification model; Obtaining gene data to be classified, and classifying the gene data to be classified using a trained gene classification model based on the feature subset of the gene data to be classified to obtain a label for the gene data to be classified.

6. A multi-label causal feature selection system, characterized in that: The feature selection system executes the multi-label causal feature selection method according to claim 1, comprising: a data acquisition and preprocessing module, an approximate Markov blanket feature acquisition module, a Markov blanket feature determination module, a feature verification module, and a feature merging module; A data acquisition and preprocessing module acquires a multi-label dataset and preprocesses the multi-label dataset, wherein the multi-label dataset is a multi-label gene expression dataset; The approximate Markov blanket feature acquisition module uses a unary approximate Markov blanket to find the approximate Markov blanket feature for each label based on the preprocessed multi-label gene expression dataset; The Markov blanket feature determination module deletes other labels contained in the approximate Markov blanket of the label and restores the equivalent features to obtain the Markov blanket feature of the label; The feature verification module uses a fast symmetry check method to ensure that the recovered features belong to the Markov blanket of the label. Features that pass the fast symmetry check method will be retained, and features that fail will be deleted; The feature merging module merges the Markov blanket features of all labels to obtain the feature subset of gene expression data, thereby realizing feature selection of gene expression data.

Images