High-dimensional data classification method, device, equipment and storage medium

Abstract

The present invention discloses a high-dimensional data classification method, device, equipment and storage medium. The method includes: obtaining high-dimensional data to be classified; normalizing the feature vectors of each dimension of the high-dimensional data to be classified; determining the projection samples of each training sample according to the feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training sample; determining the mean of the projection samples of each training sample in each binary classifier in the preset classifier structure; determining the projection data of the high-dimensional data to be classified according to the feature mapping model; comparing the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier, respectively obtaining the comparison results of each binary classifier; and determining the category of the high-dimensional data to be classified according to the comparison results of each binary classifier. The method can realize one-step fast and accurate prediction of multi-classification problems and improve the speed of high-dimensional data classification.

Description

Technical Field

[0001] The present invention relates to the field of data classification, and in particular to a high-dimensional data classification method, device, equipment and storage medium. Background Art

[0002] Currently, data classification has been used in many businesses, such as health risk prediction, underwriting conclusion prediction, claims risk prediction, financial invoice classification and recognition, etc., and with the popularization of technologies such as OCR (Optical Character Recognition) and text analysis, classification problems as a basic prerequisite technology are gradually becoming a technical point that cannot be ignored or circumvented. Therefore, research on classification technology has important significance and value both for business applications and basic technology research and development.

[0003] There are many methods available for multi-classification of high-dimensional data, such as support vector machines and deep neural networks. Support vector machines use the idea of ​​feature dimensionality increase to achieve classification, while deep neural networks use the idea of ​​statistical learning to achieve multi-classification. There are many problems with deep learning technology, such as bulky models, large storage space and computational requirements, which make it run very slowly under the traditional CPU (Central Processing Unit) architecture, while the cost of GPU (Graphics Processing Unit) is very high.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the invention

[0005] The present invention provides a high-dimensional data classification method, device, equipment and storage medium, which can achieve one-step fast and accurate prediction of multi-classification problems and improve the speed of high-dimensional data classification.

[0006] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0007] According to one aspect of the present invention, a high-dimensional data classification method is provided, which is characterized by comprising: obtaining high-dimensional data to be classified; normalizing the feature vectors of each dimension of the high-dimensional data to be classified; determining the projection samples of each training sample according to a feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples; determining the mean of the projection samples of each training sample in each binary classifier in a preset classifier structure; determining the projection data of the high-dimensional data to be classified according to the feature mapping model; comparing the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier to obtain the comparison results of each binary classifier respectively; and determining the category of the high-dimensional data to be classified according to the comparison results of each binary classifier.

[0008] According to an embodiment of the present invention, when the number of the categories is k, the number of the classifiers is k-1.

[0009] According to one embodiment of the present invention, the preset classifier structure is determined by the following formula:

[0010] Binary classifier 1: {C1}: {C2,C3,…C j ,…,C k}

[0011] Binary classifier 2: {C1,C2}: {C3,C4,…C j ,…,C k}

[0012] …

[0013] Binary classifier k-1: {C1,C2,…,C j ,…,C k-1}:{C k}

[0014] Among them, {C1,C2,…C j ,…,C k} is the set of categories, and “:” represents the interval of the binary classifier.

[0015] According to one embodiment of the present invention, the feature mapping model is obtained according to the following formula:

[0016]

[0017] Among them, S w is the total intra-class scatter matrix, S b is the total inter-class scatter matrix.

[0018] According to one embodiment of the present invention, the projection sample of each training sample is the product of the transposed matrix of the feature mapping model and each training sample, and the projection data of the high-dimensional data to be classified is the product of the transposed matrix of the feature mapping model and the high-dimensional data to be classified.

[0019] According to an embodiment of the present invention, the dimension of the projection data of the high-dimensional data to be classified is smaller than the dimension of the high-dimensional data to be classified.

[0020] According to one embodiment of the present invention, determining the category of the high-dimensional data to be classified based on the comparison results of each binary classifier includes: determining the scores of each category in each binary classifier based on the comparison results of each binary classifier; and accumulating the scores of each category in each binary classifier to determine the category with the highest score as the category of the high-dimensional data to be classified.

[0021] According to another aspect of the present invention, a high-dimensional data classification device is provided, characterized in that it includes: a data acquisition module, used to acquire high-dimensional data to be classified; a data processing module, used to normalize the feature vectors of each dimension of the high-dimensional data to be classified; a sample projection module, used to determine the projection samples of each training sample according to a feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples; a mean determination module, used to determine the mean of the projection samples of each training sample in each binary classifier in a preset classifier structure; a data projection module, used to determine the projection data of the high-dimensional data to be classified according to the feature mapping model; a data comparison module, used to compare the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier, and obtain the comparison results of each binary classifier respectively; and a category determination module, used to determine the category of the high-dimensional data to be classified according to the comparison results of each binary classifier.

[0022] According to another aspect of the present invention, a computer device is provided, comprising: a memory, a processor, and executable instructions stored in the memory and executable in the processor, wherein the processor implements any of the above methods when executing the executable instructions.

[0023] According to another aspect of the present invention, a computer-readable storage medium is provided, on which computer-executable instructions are stored, and when the executable instructions are executed by a processor, any of the above methods is implemented.

[0024] According to the high-dimensional data classification method provided by the present invention, the projection samples of the training samples and the projection data of the high-dimensional data to be classified are determined according to the feature mapping model, the mean of the projection samples is determined based on the preset classifier structure, and the comparison results of each binary classifier are obtained by comparing the projection data of the high-dimensional data to be classified and the mean of the projection samples, thereby determining the category of the high-dimensional data to be classified. For a classification task with a classification category of k, only k-1 classifiers are needed to obtain the final classification result, thereby realizing a one-step fast and accurate prediction of multi-classification problems and improving the speed of high-dimensional data classification.

[0025] It is to be understood that the foregoing general description and the following detailed description are exemplary only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above and other objects, features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0027] Figure 1 The figure is a flowchart of a high-dimensional data classification method according to an exemplary embodiment.

[0028] Figure 2 is a flowchart of another high-dimensional data classification method according to an exemplary embodiment.

[0029] Figure 3 It is a block diagram of a high-dimensional data classification device according to an exemplary embodiment.

[0030] Figure 4 is a schematic structural diagram of an electronic device according to an exemplary embodiment.

[0031] Figure 5 is a schematic diagram of a computer-readable storage medium according to an exemplary embodiment. DETAILED DESCRIPTION

[0032] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present invention will be more comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and thus their repeated description will be omitted.

[0033] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solution of the present invention may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known structures, methods, devices, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present invention.

[0034] Currently, health insurance has become the largest business of Internet insurance. The resulting risks such as adverse selection of customers and insurance with illness are rapidly increasing. In order to control underwriting risks, mid-to-high-end health insurance products have begun to require customers to provide physical examination reports issued by third parties. Insurance companies conduct underwriting risk assessments based on the analysis of physical examination reports, and then form a series of underwriting measures such as differentiated pricing and exclusions on this basis, thereby reducing the insurance company's claims risk to a certain extent.

[0035] The present invention provides a high-dimensional data classification method, which can realize automatic prediction of insurance risks based on customer physical examination data.

[0036] Figure 1 The figure is a flowchart of a high-dimensional data classification method according to an exemplary embodiment.

[0037] refer to Figure 1 , high-dimensional data classification methods 10 include:

[0038] In step S102, high-dimensional data to be classified is obtained.

[0039] The high-dimensional data to be classified may be, for example, image data, face pictures, voiceprint data, financial invoices, voice data, etc.

[0040] Take the physical examination report data provided by the customer as an example. The physical examination report data can be divided into the following four categories:

[0041] (1) Routine examination items: gender, age, height, weight, BMI, weight, vision, blood pressure, etc.;

[0042] (2) Laboratory tests: blood test, urine test, stool test, etc.;

[0043] (3) Ultrasound and chest X-ray examination items: liver examination conclusion, kidney examination conclusion, prostate examination conclusion, bladder examination conclusion, etc.;

[0044] (4) Electrocardiogram examination items: heart rate, abnormal values, etc.

[0045] The above four types of data are digitized, among which:

[0046] (1) For routine examination items, the original examination results (the values ​​on the physical examination report) shall be used;

[0047] (2) The standard dimension range is used for the test items, and the actual test values ​​are normalized, mainly (maximum value - test value) / maximum range;

[0048] (3) Ultrasound and chest X-ray examination items: Use the actual test description to check the medical risk level table (internal data knowledge base) to query the risk level corresponding to the test conclusion. There are four main levels: 0, 1, 2, and 3. The larger the value, the greater the risk.

[0049] (4) ECG examination items: The heart rate is normalized using (maximum value - detection value) / maximum range. The abnormal value is checked using the actual detection description to check the medical risk level table (internal data knowledge base) to query the risk level corresponding to the detection conclusion. There are four main levels: 0, 1, 2, and 3. The larger the value, the greater the risk. In step S104, the feature vectors of each dimension of the high-dimensional data to be classified are normalized.

[0050] For high-dimensional data, the scales corresponding to the elements inside it are often different. It contains vector data and scalar data, and the dimensions and importance of the data are also different. In order to be able to merge data of different dimensions into a data vector and use it for subsequent calculations, it is necessary to normalize these data so that all the data can be normalized to the same scalar space with the same measurement scale.

[0051] The feature vector composed of the high-dimensional data to be classified can be, for example, X = {x1, x2, ..., x n}, where x i represents the i-th element of the feature vector X, and n represents the data dimension, then the normalization calculation formula is:

[0052]

[0053] in,

[0054] The normalized eigenvector is X'={x1',x2',…,x n '}.

[0055] Taking the physical examination report data provided by the above customer as an example, the digitized physical examination data is normalized to obtain a normalized feature vector.

[0056] In step S106, the projection sample of each training sample is determined according to the feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples.

[0057] The data set of training samples can be expressed as D = {(X1, Y1), (X2, Y2), ..., (X m ,Y m )}, where any sample X i Represents the feature vector of n-dimensional training sample data, m represents the number of samples, and any sample X i The corresponding category is Y i ∈{C1,C2,…,C k}, k represents the number of categories. i ,Y i ) indicates that the feature vector representation of the i-th sample is X i， The corresponding category is Y i , the categories are manually labeled.

[0058] The mean vector of the j-th class sample is:

[0059]

[0060] Where N j represents the number of samples of the jth class, X j represents the set of samples of the jth class, class j = {1, 2, ..., k}, X i,j For the set X j The i-th element in , that is, i is the serial number code of sample X under category j, indicating the nth eigenvector.

[0061] The intra-class covariance matrix of the j-th class sample is:

[0062]

[0063] Where T represents the transpose of the matrix.

[0064] The total intra-class scatter matrix of each type of sample is:

[0065]

[0066] The total inter-class scatter matrix of each type of sample is:

[0067]

[0068] Where μ is the mean vector of all samples.

[0069] In some embodiments, the feature mapping model is obtained according to the following formula:

[0070]

[0071] Among them, S w is the total intra-class scatter matrix, S bis the between-class scatter matrix. W = {w1, w2, …, w d} is an n×d matrix, which is the feature mapping model matrix to be solved.

[0072] To achieve good classification results, the feature mapping model needs to group the data with the same class label as closely as possible and separate the data with different class labels as much as possible. Only in this way can the obtained feature mapping hyperplane have better ability to distinguish samples. The degree of aggregation of intra-class data can be represented by the within-class scatter matrix, and the degree of classification of inter-class data can be represented by the between-class scatter matrix.

[0073] The goal of the feature projection model is to find the optimal classification hyperplane for the data, then project the high-dimensional data to be classified onto this hyperplane, and finally predict the data class through a classification method.

[0074] Based on the Rayleigh Quotient theory, the feature mapping model matrix W can be obtained by combining the eigenvectors corresponding to the first d largest eigenvalues of the matrix . The size of d is determined according to the number of classification categories, and d ≤ k - 1. If you want to achieve feature dimensionality reduction while classifying, then you can make d < k - 1, but it must be ensured that the removed eigenvectors are those corresponding to the smallest sorted eigenvalues. Reducing the eigenvectors means reducing the dimension of the feature mapping model, and the corresponding target classification accuracy will be slightly reduced, but generally the impact is not significant.

[0075] In some embodiments, the projection samples of each training sample are the product of the transpose matrix of the feature mapping model and each training sample.

[0076] Still taking the above-mentioned training sample D as an example, all sample vectors X i are all subjected to the transformation of W T X i to obtain the projected feature vector Z i . The set of projected samples in the new feature space is where the projected feature vector Z i corresponds one-to-one to the sample data vector X i in the aforementioned D.

[0077] In step S108, determine the mean value of the projection samples of each training sample in each binary classifier in the preset classifier structure.

[0078] In some embodiments, when the number of classes is k, the number of classifiers is k - 1.

[0079] As can be seen from the above, the feature mapping model matrix W is an n×d matrix, where d ≤ k - 1, which can play the role of dimensionality reduction. Therefore, it is more reasonable to set the number of classifiers to k - 1.

[0080] In some embodiments, the preset classifier structure is determined by the following formula:

[0081] Binary classifier 1: {C1}: {C2,C3,…C j ,…,C k}

[0082] Binary classifier 2: {C1,C2}: {C3,C4,…C j ,…,C k}

[0083] …

[0084] Binary classifier k-1: {C1,C2,…,C j ,…,C k-1}:{C k}

[0085] Among them, {C1,C2,…C j ,…,C k} is a set of categories, and “:” represents the interval of the binary classifier.

[0086] There are four risk levels used in insurance underwriting business, namely K=4, from low to high: normal (underwriting), low risk (additional inspection), medium risk (extension or exclusion), and high risk (rejection of insurance).

[0087] Still taking the above training sample D as an example, after projection, the projected sample is Calculate the mean of the projected samples of each training sample in each binary classifier.

[0088] Taking binary classifier 1 as an example, according to the formula Calculate the mean of {C1} and get P, where N j represents the number of samples in the jth class, Z j represents the set of samples of the jth category, category j = {1, 2, ..., k}, Z i,j For the set Z j The i-th element in , that is, i is the serial number code of sample Z under category j, indicating the nth eigenvector.

[0089] According to the formula Calculate {C2,C3,…C j ,…,C k}, and get Q.

[0090] The mean of the projected samples of each training sample in the binary classifier 1 is: f1 = (P + Q) / 2.

[0091] According to the above method, we can get the mean values ​​of binary classifier 2, binary classifier 3, ... binary classifier k-1 as f2, f3, ... f k-1 . Thus, the mean vector F of the projection samples of each training sample in each binary classifier is obtained. F is a vector with a dimension less than or equal to k-1, expressed as:

[0092] F={f1,f2,…,f k-1}

[0093] In step S110, projection data of the high-dimensional data to be classified is determined according to the feature mapping model.

[0094] In some embodiments, the projection data of the high-dimensional data to be classified is the product of the transposed matrix of the feature mapping model and the high-dimensional data to be classified.

[0095] The feature vector X={x1,x2,…,x n} as an example, the normalized eigenvector is X'={x1',x2',…,x n '}. The projection data of the high-dimensional data to be classified is Among them, W is the feature mapping model matrix.

[0096] In some embodiments, the dimension of the projection data of the high-dimensional data to be classified is smaller than the dimension of the high-dimensional data to be classified.

[0097] Taking the projection data of the above-mentioned high-dimensional data to be classified as an example, the dimension of the projection data of the high-dimensional data to be classified is k-1, and the dimension of the high-dimensional data to be classified is k, that is, the dimension of the projection data of the high-dimensional data to be classified is smaller than the dimension of the high-dimensional data to be classified.

[0098] In step S112, the projection data of the high-dimensional data to be classified is compared with the mean of the projection samples of the training samples in each binary classifier to obtain the comparison results of each binary classifier.

[0099] The above projection data In The mean f of the projection samples of the training samples in the above two classifiers i By comparison, it is concluded and f i The size relationship.

[0100] In step S114, the category of the high-dimensional data to be classified is determined according to the comparison results of each binary classifier.

[0101] according to and f i The size relationship is determined by selecting the category in the binary classifier for scoring, and the category of the high-dimensional data to be classified is determined according to the scoring result.

[0102] In insurance underwriting, the results corresponding to the category with the highest score can be used as the risk underwritten by the client.

[0103] According to the high-dimensional data classification method provided by the embodiment of the present invention, the projection samples of the training samples and the projection data of the high-dimensional data to be classified are determined according to the feature mapping model, and the mean of the projection samples is determined based on the preset classifier structure. By comparing the projection data of the high-dimensional data to be classified and the mean of the projection samples, the comparison results of each binary classifier are obtained, thereby determining the category of the high-dimensional data to be classified. For a classification task with a classification category of k, only k-1 classifiers are needed to obtain the final classification result, thereby realizing one-step rapid and accurate prediction of multi-classification problems and improving the speed of high-dimensional data classification.

[0104] The high-dimensional data classification method provided by the embodiment of the present invention can be quickly integrated into various software systems. For example, it can be integrated into the underwriting conclusion software system to provide underwriting conclusion prediction in real time. The operating personnel can decide the amount of effort to be invested in the review according to the underwriting prediction conclusion, thereby improving work efficiency and reducing underwriting risks.

[0105] It should be clearly understood that the present invention describes how to form and use specific examples, but the principles of the present invention are not limited to any details of these examples. On the contrary, based on the teachings of the content disclosed in the present invention, these principles can be applied to many other embodiments.

[0106] Figure 2 is a flowchart of another high-dimensional data classification method according to an exemplary embodiment. Figure 1 The method 10 shown differs in that Figure 2 It further provides how to determine the category of the high-dimensional data to be classified based on the comparison results of each binary classifier, that is, Figure 2 The method shown further provides Figure 1 An embodiment of step S114 in .

[0107] refer to Figure 2 , step S114 includes:

[0108] In step S1142, the scores of each category in each binary classifier are determined respectively according to the comparison results of each binary classifier.

[0109] Take the number of classification categories as 4 and the category of high-dimensional data to be classified as C3 as an example. According to Figure 1 The method 10 shown in FIG. Then the binary classifier 1 considers that the high-dimensional data to be classified belongs to the right category, that is, each category in {C2, C3, C4} is scored as 1, and the category in {C1} is scored as 0; according to Figure 1 The method 10 shown in FIG. Then the binary classifier 2 considers that the high-dimensional data to be classified belongs to the right category, that is, each category in {C3, C4} is scored as 1, and the category score in {C1, C2} is 0; according to Figure 1 The method 10 shown in FIG. Then the binary classifier 3 considers that the high-dimensional data to be classified belongs to the category on the left, that is, each category in {C1, C2, C3} is scored as 1, and the category score in {C4} is 0.

[0110] In step S1144, the scores of each category in each binary classifier are accumulated to determine the category with the highest score as the category of the high-dimensional data to be classified.

[0111] By adding up the scores of each category in the above binary classifiers, we can get the total score of category C1 is 1, the total score of category C2 is 2, the total score of category C3 is 3, and the total score of category C4 is 2. The highest score is category C3, so the category of the high-dimensional data to be classified is C3.

[0112] Those skilled in the art will appreciate that all or part of the steps for implementing the above embodiments are implemented as a computer program executed by a CPU. When the computer program is executed by the CPU, the above functions defined by the above method provided by the present invention are performed. The program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk, or an optical disk.

[0113] In addition, it should be noted that the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0114] The following are embodiments of the device of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments of the present invention, please refer to the method embodiments of the present invention.

[0115] Figure 3 It is a block diagram of a high-dimensional data classification device according to an exemplary embodiment.

[0116] refer to Figure 3 The high-dimensional data classification device 30 includes: a data acquisition module 302, a data processing module 304, a sample projection module 306, a mean determination module 308, a data projection module 310, a data comparison module 312 and a category determination module 314.

[0117] The data acquisition module 302 is used to acquire high-dimensional data to be classified.

[0118] The data processing module 304 is used to normalize the feature vectors of each dimension of the high-dimensional data to be classified.

[0119] The sample projection module 306 is used to determine the projection sample of each training sample according to the feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples. In some embodiments, the feature mapping model is obtained according to the following formula:

[0120]

[0121] Among them, S w is the total intra-class scatter matrix, S b is the total between-class scatter matrix.

[0122] The mean value determination module 308 is used to determine the mean value of the projection samples of each training sample in each binary classifier in the preset classifier structure.

[0123] In some embodiments, when the number of categories is k, the number of classifiers is k-1.

[0124] In some embodiments, the preset classifier structure is determined by the following formula:

[0125] Binary classifier 1: {C1}: {C2,C3,…C j ,…,C k}

[0126] Binary classifier 2: {C1,C2}: {C3,C4,…C j ,…,C k}

[0127] …

[0128] Binary classifier k-1: {C1,C2,…,C j ,…,C k-1}:{C k}

[0129] Among them, {C1,C2,…C j ,…,C k} is a set of categories, and “:” represents the interval of the binary classifier.

[0130] The data projection module 310 is used to determine the projection data of the high-dimensional data to be classified according to the feature mapping model.

[0131] In some embodiments, the projection sample of each training sample is the product of the transposed matrix of the feature mapping model and each training sample, and the projection data of the high-dimensional data to be classified is the product of the transposed matrix of the feature mapping model and the high-dimensional data to be classified.

[0132] In some embodiments, the dimension of the projection data of the high-dimensional data to be classified is smaller than the dimension of the high-dimensional data to be classified.

[0133] The data comparison module 312 is used to compare the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier, and obtain the comparison results of each binary classifier respectively.

[0134] The category determination module 314 is used to determine the category of the high-dimensional data to be classified according to the comparison results of each binary classifier.

[0135] In some embodiments, the category determination module 314 includes a score determination module and a score accumulation module, wherein the score determination module is used to determine the scores of each category in each binary classifier according to the comparison results of each binary classifier, and the score accumulation module is used to accumulate the scores of each category in each binary classifier to determine the category with the highest score as the category of the high-dimensional data to be classified.

[0136] According to the high-dimensional data classification device provided by the embodiment of the present invention, the projection samples of the training samples and the projection data of the high-dimensional data to be classified are determined according to the feature mapping model, and the mean of the projection samples is determined based on the preset classifier structure. By comparing the projection data of the high-dimensional data to be classified and the mean of the projection samples, the comparison results of each binary classifier are obtained, thereby determining the category of the high-dimensional data to be classified. For a classification task with a classification category of k, only k-1 classifiers are needed to obtain the final classification result, thereby realizing one-step rapid and accurate prediction of multi-classification problems and improving the speed of high-dimensional data classification.

[0137] It should be noted that the block diagrams shown in the above figures are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0138] Figure 4 is a schematic diagram of a structure of an electronic device according to an exemplary embodiment. It should be noted that: Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0139] like Figure 4As shown, the electronic device 800 is in the form of a general-purpose computer device. The components of the electronic device 800 include: at least one central processing unit (CPU) 801, which can perform various appropriate actions and processes according to program codes stored in a read-only memory (ROM) 802 or program codes loaded from at least one storage unit 808 to a random access memory (RAM) 803.

[0140] In particular, according to an embodiment of the present invention, the program code can be executed by the central processing unit 801, so that the central processing unit 801 performs the steps according to various exemplary embodiments of the present invention described in the above method embodiment section of this specification. For example, the central processing unit 801 can perform the following steps: Figure 1 or Figure 2 Follow the steps shown in .

[0141] In the RAM 803, various programs and data necessary for the operation of the electronic device 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0142] The following components are connected to the I / O interface 805: an input unit 806 including a keyboard, a mouse, etc.; an output unit 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage unit 808 including a hard disk, etc.; and a communication unit 809 including a network interface card such as a LAN card, a modem, etc. The communication unit 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage unit 808 as needed.

[0143] Figure 5 is a schematic diagram of a computer-readable storage medium according to an exemplary embodiment.

[0144] refer to Figure 5 As shown, a program product 900 configured to implement the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0145] The computer-readable medium carries one or more programs. When the one or more programs are executed by a device, the computer-readable medium implements the following Figure 1 or Figure 2 The functionality shown in .

[0146] The exemplary embodiments of the present invention are specifically shown and described above. It should be understood that the present invention is not limited to the detailed structure, configuration or implementation method described herein; on the contrary, the present invention is intended to cover various modifications and equivalent configurations included in the spirit and scope of the appended claims.

Claims

1. A high-dimensional data classification method, characterized in that: include: Obtain high-dimensional data to be classified; Normalizing the feature vectors of each dimension of the high-dimensional data to be classified; Determine the projection sample of each training sample according to a feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples; Determine the mean of the projected samples of each training sample in each binary classifier in the preset classifier structure; Determining projection data of the high-dimensional data to be classified according to the feature mapping model; Compare the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier to obtain the comparison results of each binary classifier respectively; as well as Determining the category of the high-dimensional data to be classified according to the comparison results of each binary classifier; Wherein, determining the category of the high-dimensional data to be classified according to the comparison results of each binary classifier includes: determining the score of each category in each binary classifier according to the comparison results of each binary classifier; and accumulating the scores of each category in each binary classifier to determine the category with the highest score as the category of the high-dimensional data to be classified; When the number of categories of the training samples is k, the number of classifiers is k-1, and the preset classifier structure is determined by the following formula: Binary classifier 1: {C1}: {C2,C3,…C j ,…,C k } Binary classifier 2: {C1,C2}: {C3,C4,…C j ,…, k } …… Binary classifier k-1: {C1,C2,…,C j ,…,C k-1 }:{C k } Among them, {C1,C2,…C j ,…,C k } is the set of categories, ":" represents the interval of the binary classifier; The high-dimensional data to be classified is physical examination report data, which is data with four major dimensional features consisting of routine examination items, laboratory examination items, B-ultrasound and chest X-ray examination items, and electrocardiogram examination items.

2. The method according to claim 1, characterized in that The feature mapping model is obtained according to the following formula: Among them, S w is the total intra-class scatter matrix, S b is the total inter-class divergence matrix, J(W) represents the generalized Rayleigh quotient of the feature mapping model matrix W to be solved, W T represents the transpose of the feature mapping model matrix W to be solved, w i T Represents the transpose of the i-th column of the feature mapping model matrix W to be solved.

3. The method according to claim 2, characterized in that The projection sample of each training sample is the product of the transposed matrix of the feature mapping model and each training sample, and the projection data of the high-dimensional data to be classified is the product of the transposed matrix of the feature mapping model and the high-dimensional data to be classified.

4. The method according to claim 3, characterized in that The dimension of the projection data of the high-dimensional data to be classified is smaller than the dimension of the high-dimensional data to be classified.

5. A high-dimensional data classification device, characterized in that: include: A data acquisition module, used to acquire high-dimensional data to be classified; A data processing module, used for normalizing the feature vectors of each dimension of the high-dimensional data to be classified; A sample projection module, used to determine the projection sample of each training sample according to a feature mapping model constructed based on the total intra-class scatter matrix and the total inter-class scatter matrix of the training samples; A mean value determination module, used to determine the mean value of the projection samples of each training sample in each binary classifier in a preset classifier structure; A data projection module, used to determine the projection data of the high-dimensional data to be classified according to the feature mapping model; A data comparison module, used for comparing the projection data of the high-dimensional data to be classified with the mean of the projection samples of the training samples in each binary classifier, and obtaining the comparison results of each binary classifier respectively; as well as A category determination module, used to determine the category of the high-dimensional data to be classified according to the comparison results of each binary classifier; The category determination module includes: a score determination module, which is used to determine the scores of each category in each binary classifier according to the comparison results of each binary classifier; a score accumulation module, which is used to accumulate the scores of each category in each binary classifier and determine the category with the highest score as the category of the high-dimensional data to be classified; When the number of categories of the training samples is k, the number of classifiers is k-1, and the preset classifier structure is determined by the following formula: Binary classifier 1: {C1}: {C2,C3,…C j ,…,C k } Binary classifier 2: {C1,C2}: {C3,C4,…C j ,…,C k } …… Binary classifier k-1: {C1,C2,…,C j ,…,C k-1 }:{C k } Among them, {C1,C2,…C j ,…,C k } is the set of categories, ":" represents the interval of the binary classifier; The high-dimensional data to be classified is physical examination report data, which is data with four major dimensional features consisting of routine examination items, laboratory examination items, B-ultrasound and chest X-ray examination items, and electrocardiogram examination items.

6. A computer device comprising: A memory, a processor, and executable instructions stored in the memory and executable in the processor, wherein the processor implements the method according to any one of claims 1 to 4 when executing the executable instructions.

7. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that: When the executable instructions are executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images