CN115905654A

CN115905654A - Service data processing method, device, equipment, storage medium and program product

Info

Publication number: CN115905654A
Application number: CN202211038510.9A
Authority: CN
Inventors: 李熠; 邬子庄; 赵心睿
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2022-08-29
Filing date: 2022-08-29
Publication date: 2023-04-04

Abstract

The present application relates to the field of big data technologies, and in particular, to a method and an apparatus for processing business data, a computer device, a storage medium, and a computer program product. Dividing the service data to be processed into multi-class view data according to a preset classification dimension; aiming at various view data, constructing a feature matrix by using the view data, and constructing a target optimization function based on the feature matrix and the view data; solving the target optimization function to obtain a feature extraction result of the service data; the feature extraction result is used for representing feature information of the service data. By adopting the method, the efficiency of obtaining the characteristic extraction result of the characteristic information representing the business data can be improved, namely, the efficiency of obtaining the characteristic information of the business data is improved through the process.

Description

Service data processing method, device, equipment, storage medium and program product

Technical Field

The present application relates to the field of big data technologies, and in particular, to a method, an apparatus, a device, a storage medium, and a program product for processing service data.

Background

With the development of information technology, various industries generate a large amount of business data, for example, in the financial industry. By effectively analyzing the business data, enterprises can be assisted to make more optimized operation and maintenance strategies based on the analysis results.

In the traditional technology, business data is mainly analyzed through a machine learning method. However, due to the influences of factors such as large data volume of the service data to be analyzed, high data dimension, complex data relation and the like, the traditional method has the problem of low service data analysis efficiency.

Disclosure of Invention

In view of the above, it is necessary to provide a business data processing method, a business data processing apparatus, a business data processing device, a business data processing apparatus, a storage medium, and a program product, which can improve the efficiency of analyzing business data.

In a first aspect, the present application provides a method for processing service data, where the method includes:

dividing the service data to be processed into multi-class view data according to a preset classification dimension;

aiming at various types of view data, constructing a feature matrix by using the view data, and constructing an objective optimization function based on the feature matrix and the view data;

solving the target optimization function to obtain a feature extraction result of the service data; the feature extraction result is used for representing feature information of the service data.

In one embodiment, the feature matrix is a first feature matrix or a second feature matrix; the constructing a feature matrix by using the view data for each type of the view data comprises:

determining whether the service data has a tag; the label is used for representing the type information of the service data;

if so, constructing the first feature matrix by using the view data and the label corresponding to the view data;

and if not, constructing the second feature matrix by using the view data.

In one embodiment, the constructing the first feature matrix by using the view data and the label corresponding to the view data includes:

sequencing the view data to obtain sequenced view data;

and constructing the first feature matrix according to the label corresponding to the sequenced view data and the position information of the sequenced view data.

In one embodiment, the constructing an objective optimization function based on the feature matrix and the view data includes:

and constructing the objective optimization function based on the first feature matrix and the view data.

In one embodiment, the method further comprises:

acquiring a target label with the largest ratio in target view data; the target view data are a plurality of view data, the distance between the target view data and the view data is smaller than a preset threshold value;

determining the target label as a reconstruction label of the view data, and constructing an optimized first feature matrix according to the view data and the reconstruction label corresponding to the view data;

constructing an adjusted target optimization function based on the optimized first feature matrix and the view data;

and solving the adjusted target optimization function to obtain the feature extraction result.

In one embodiment, the constructing the feature matrix using the view data includes:

and constructing the second feature matrix based on the neighbor relation information of the view data.

and constructing the objective optimization function based on the second feature matrix and the view data.

In one embodiment, the solving the objective optimization function to obtain the feature extraction result of the service data includes:

solving the target optimization function to obtain a feature extraction matrix of the service data;

and acquiring the feature extraction result based on the feature extraction matrix.

In one embodiment, the method further comprises:

and acquiring a feature extraction result of the new service data to be processed by using the feature extraction matrix and the new service data to be processed.

In a second aspect, the present application further provides a device for processing service data, where the device includes:

the dividing module is used for dividing the service data to be processed into multi-class view data according to a preset classification dimension;

the first construction module is used for constructing a feature matrix by utilizing the view data aiming at various types of the view data and constructing an objective optimization function based on the feature matrix and the view data;

the first acquisition module is used for solving the target optimization function and acquiring a feature extraction result of the service data; the feature extraction result is used for representing feature information of the service data.

In a third aspect, the present application further provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the following steps when executing the computer program:

for each type of view data, constructing a feature matrix by using the view data, and constructing an objective optimization function based on the feature matrix and the view data;

In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, which when executed by a processor, performs the steps of:

In a fifth aspect, the present application further provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of:

According to the business data processing method, the business data processing device, the business data to be processed can be divided into multiple types of view data according to the preset classification dimension, the view data can be used for constructing the feature matrix aiming at the various types of view data, the target optimization function is constructed on the basis of the feature matrix and each type of view data, the constructed feature matrix can represent feature information of the view data, low-dimensional features containing useful data information can be extracted from a high-dimensional data feature space through the constructed feature matrix, the dimension of the obtained feature matrix is lower than that of the view data, the target optimization function can be rapidly constructed on the basis of the feature matrix and the view data, the efficiency of constructing the target optimization function is improved, the feature extraction result of the business data is obtained by solving the target optimization function, the efficiency of the obtained feature extraction result representing the feature information of the business data is improved, and the efficiency of obtaining the feature information of the business data is improved through the process.

Drawings

Fig. 1 is a schematic flow chart illustrating a method for processing service data according to an embodiment;

fig. 2 is a schematic flowchart of a processing method of service data in another embodiment;

fig. 3 is a schematic flow chart of a processing method of service data in another embodiment;

fig. 4 is a schematic flow chart of a processing method of service data in another embodiment;

FIG. 5 is a flow diagram that illustrates a method for processing tagged business data, according to one embodiment;

FIG. 6 is a flow diagram that illustrates a method for processing traffic data that does not have tags, according to one embodiment;

FIG. 7 is a block diagram showing a configuration of a device for processing service data according to an embodiment;

fig. 8 is a block diagram showing a configuration of a service data processing apparatus according to another embodiment;

fig. 9 is a block diagram showing a configuration of a service data processing apparatus according to another embodiment;

fig. 10 is a block diagram showing a configuration of a service data processing apparatus according to another embodiment;

fig. 11 is a block diagram showing a configuration of a service data processing apparatus according to another embodiment;

fig. 12 is a block diagram showing a configuration of a service data processing apparatus according to another embodiment;

FIG. 13 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

It should be noted that the method, the apparatus, the device, the storage medium, and the program product for processing service data according to the present application may be applied to the field of big data, and may also be applied to other technical fields except the field of big data.

In an embodiment, as shown in fig. 1, a method for processing service data is provided, and this embodiment is illustrated by applying the method to a computer device, it is to be understood that the method may also be applied to a server, and may also be applied to a system including a computer device and a server, and is implemented by interaction between the computer device and the server. In this embodiment, the method includes the steps of:

s101, dividing the service data to be processed into multi-class view data according to a preset classification dimension.

The business data to be processed can be financial business data, enterprise approval business data, business data of a transformer substation and the like; the preset classification dimension can be the type of the service data, the user group corresponding to the service data, and the like. For example, taking the service data to be processed in this embodiment as financial service data as an example, the preset classification dimensions may be three classification dimensions, such as basic information of the user, asset information of the user, and behavior information of the user. It can be understood that, in this embodiment, the category of the view data corresponds to the classification dimension, for example, if the classification dimension includes four dimensions, the obtained view data is four types of view data; for another example, if the classification dimensions include five dimensions, the obtained view data is five types of view data. Optionally, the preset classification dimension may be determined based on a historical experience value, or may be determined according to a data type included in the service data to be processed.

S102, aiming at various types of view data, a feature matrix is constructed by using the view data, and an objective optimization function is constructed based on the feature matrix and the view data.

Optionally, in this embodiment, for each type of view data, a feature matrix may be constructed by using a neighbor relation between the view data and adjacent view data, or a feature matrix may also be constructed by using a label corresponding to the view data, or a preset feature extraction method may be used to perform feature extraction on the view data, and a feature matrix is constructed by using the extracted features.

Optionally, in this embodiment, the target optimization function may be composed of a principal component term, a regression term, and a regularization term, where the principal component term may be constructed using the feature matrix, the regression term may be constructed based on the view data, and the regression term may be fit to the nonlinear mapping by linear projection, so that the feature extraction method has a nonlinear dimension reduction capability, and can effectively improve a feature extraction effect, and the regularization term may also be constructed based on the view data, so as to alleviate an over-fitting problem of the target optimization function and improve robustness of the target optimization function. Illustratively, the constructed objective optimization function may be:

in the formula, B represents a main component item constructed based on a characteristic matrix; v is the number of views, and gamma is a hyper-parameter, and is used for adjusting the influence of a regularization term on the whole optimization problem;

the regression term can be constructed by fitting nonlinear mapping through linear projection, so that the feature extraction method has the nonlinear dimension reduction capability and can effectively improve the feature extraction effect; />

The method is a regular term, so that the overfitting problem can be relieved, and the robustness of the model is improved; />

In order to maximize the divergence of the variables after the feature extraction, the correlation among the variables after the feature extraction can be reduced, and the learning accuracy of a subsequent model can be effectively improved; alpha is alpha _i ，Y，P ⁽ⁱ⁾ Optimization objective for optimization problem, wherein, α _i Weighting the balance factor for the adaptive view; p _(i) Extracting a dimension reduction matrix for the features under each view; and Y is the feature extraction result of the current data.

S103, solving the target optimization function to obtain a feature extraction result of the service data; the feature extraction result is used for representing feature information of the service data.

Optionally, in this embodiment, the hyper-parameters and the initialization parameters of the objective optimization function may be set based on the optimization conditions corresponding to the objective optimization function, the objective optimization function is solved to obtain a feature extraction matrix corresponding to the service data, and a feature extraction result of feature information representing the service data is obtained based on the feature extraction matrix.

In the method for processing the service data, the service data to be processed can be divided into multiple types of view data according to preset classification dimensions, a feature matrix can be constructed by using the view data aiming at the various types of view data, a target optimization function is constructed on the basis of the feature matrix and each type of view data, the constructed feature matrix can represent feature information of the view data, low-dimensional features containing useful data information can be extracted from a high-dimensional data feature space by the constructed feature matrix, the dimension of the obtained feature matrix is lower than that of the view data, a target optimization function can be rapidly constructed on the basis of the feature matrix and the view data, the efficiency of constructing the target optimization function is improved, the feature extraction result of the service data is obtained by solving the target optimization function, the efficiency of the obtained feature extraction result representing the feature information of the service data is improved, and the efficiency of obtaining the feature information of the service data is improved through the process.

In some scenarios, the service data may be service data with a tag or service data without a tag, and the feature matrix may be constructed by using different methods for different service data. In the above scenario of constructing a feature matrix by using view data for various types of view data, the constructed feature matrix may be a first feature matrix or a second feature matrix, and in one embodiment, as shown in fig. 2, the "constructing a feature matrix by using view data for various types of view data" in S102 includes:

s201, determining whether the service data has a label; the label is used for representing the type information of the service data.

The label of the service data can be used for characterizing the type information of the service data. Optionally, the service data may be service data with a tag, or may also be service data without a tag. Optionally, in this embodiment, the computer device may determine whether the service data has the tag according to whether the service data has the corresponding type information, for example, if the service data is a kth-class service data, that is, the service data has the corresponding type information, the computer device may determine that the service data is the service data having the tag.

And S202, if yes, constructing a first feature matrix by using the view data and the label corresponding to the view data.

In this embodiment, for the case that the service data has a label, the computer device may construct the first feature matrix by using the view data and the label corresponding to the view data. Optionally, in an embodiment, the view data may be sorted first to obtain sorted view data, and a first feature matrix is constructed according to the labels corresponding to the sorted view data and the position information of the sorted view data. Optionally, the first feature matrix constructed in this embodiment may be an indication matrix, for example, so as to

And representing the indicating matrix of the kth class, constructing the indicating matrix as follows:

and S203, if not, constructing a second feature matrix by using the view data.

In this embodiment, for the case that the service data does not have a label, the computer device may construct the second feature matrix using the view data. Optionally, in an embodiment, the second feature matrix may be constructed based on neighbor relation information of the view data, where the neighbor relation information of the view data may be neighbor relations between the view data under all views. Optionally, the second feature matrix constructed in this embodiment may be a global local structure retention matrix W, and for example, the global local structure retention matrix W constructed by using the neighbor relation of the view data may be:

if v is _i And v _j All the views are neighboring points

It should be noted that the global local structure maintaining matrix W constructed in the present embodiment considers the neighbor relationship between view data under all views and the distance between global neighbor view data under different views.

In the embodiment, different feature matrices are constructed by judging whether the service data has a label, if the service data has the label, a first feature matrix is constructed by using the view data and the label corresponding to the view data, and if the service data does not have the label, a second feature matrix is constructed based on neighbor relation information of the view data; when the second feature matrix is constructed for the service data without the label, the neighbor relation between the view data under all views and the distance between the global neighbor view data under different views are considered at the same time, and the method has important significance for maintaining the topology relation based on the global after the dimension reduction of the data.

Further, in the above scenario of constructing the objective optimization function based on the feature matrix and the view data, if the constructed feature matrix is the first feature matrix, in an embodiment, the "constructing the objective optimization function based on the feature matrix and the view data" in S102 includes: and constructing an objective optimization function based on the first feature matrix and the view data.

Optionally, the constructed target optimization function may be a supervised optimization function constructed based on the first feature matrix and the multi-view data, and as an optional implementation manner, in this embodiment, the target optimization function may be a supervised optimization function, and is respectively composed of an information discriminant term, a regression term, and a regularization term, and the view weight is adjusted by an adaptive method. For example, in this embodiment, based on the first feature matrix and the view data, the constructed objective optimization function may be:

in the formula (I), the compound is shown in the specification,

for the information discrimination item constructed based on the first feature matrix, the feature extraction effect can be improved in a mode of aggregating the same kind of view data as closely as possible, wherein alpha is _i ，Y，P ⁽ⁱ⁾ Is an optimization objective of an optimization problem, wherein，α _i Weighting the balance factor for the adaptive view; p _(i) Extracting a dimension reduction matrix for the features under each view; and Y is the feature extraction result of the current data.

In this embodiment, for service data with a label, according to the first feature matrix and the view data, the constructed target optimization function can improve the feature extraction effect by aggregating similar sample points as closely as possible, so that the service data after dimension reduction is more comprehensive, and the constructed target optimization function is more accurate.

In the scenario that the service data is service data with tags, the tags of the service data may have an inaccuracy problem, so that the tags of the service data may be reconstructed, and an optimized objective optimization function may be constructed by using the service data and the reconstructed tags. In one embodiment, as shown in fig. 3, the method further comprises:

s301, acquiring a target label with the largest proportion in target view data; the target view data is a plurality of view data with a distance from the view data smaller than a preset threshold.

The target view data may be a plurality of view data having a distance from the view data smaller than a preset threshold, and it is understood that the preset threshold may be set in advance based on an empirical value. Optionally, in this embodiment, the computer device may determine, by using a K-nearest neighbor method, a plurality of target view data whose distance from the view data is smaller than a preset threshold. Further, for the target label with the largest proportion in the target view data, the obtaining manner may be to count all the label data in the target view to obtain the target label with the largest proportion. It should be noted that, if the target view data has a plurality of labels with the same percentage, a label corresponding to the target view data closest to the view data in the target view data corresponding to the labels may be selected as the target label.

S302, determining the target label as a reconstruction label of the view data, and constructing an optimized first feature matrix according to the view data and the reconstruction label corresponding to the view data.

In this embodiment, the computer device determines the obtained target tag as a reconstruction tag of the view data, and further, the computer device may construct an optimized first feature matrix according to the view data and the reconstruction tag corresponding to the view data by using the method for constructing the first feature matrix in S202. It should be noted that a process of constructing the optimized first feature matrix is the same as the process of constructing the first feature matrix described in the foregoing S202, and details are not repeated here in this embodiment.

S303, constructing an adjusted target optimization function based on the optimized first feature matrix and the view data.

Exemplarily, in the present embodiment, to

For example, if the optimized first feature matrix is represented, the adjusted objective optimization function constructed based on the optimized first feature matrix and the view data may be:

in the formula (I), the compound is shown in the specification,

represents an adjusted target optimization function>

For the explanation of other parameters in the formula, please refer to the description in S202, which is not repeated herein.

And S304, solving the adjusted optimization function to obtain a feature extraction result.

In this embodiment, the process of obtaining the feature extraction result may be to set an initial value of the hyper-parameter γ firstInitialization parameter α _i =1/v, then solving the objective optimization function and calculating intermediate data

And setting the cycle end conditions: a is less than or equal to sigma or the cycle number reaches k times, and the following formula is applied to loop iteration calculation results:

(1) Solving by using generalized eigenvalue solving method

The solution of Y is a matrix formed by the eigenvectors corresponding to the first M minimum eigenvalues.

(2) Solving for

Wherein the obtained alpha _i For view weight, further, a feature extraction matrix is obtained based on the solving result

Wherein Y is a feature extraction result.

In the embodiment, the target label with the largest ratio in the view data with the distance from the view data smaller than the preset threshold is obtained, the target label is determined as the reconstruction label of the view data, the optimized first feature matrix is constructed according to the view data and the reconstruction label corresponding to the view data, the adjusted target optimization function is constructed based on the optimized first feature matrix and the view data, the adjusted target optimization function is solved, and the feature extraction result is obtained.

Further, in the above scenario of constructing the objective optimization function based on the feature matrix and the view data, if the constructed feature matrix is the second feature matrix, in an embodiment, the "constructing the objective optimization function based on the feature matrix and the view data" in S102 includes: and constructing an objective optimization function based on the second feature matrix and the view data.

As an optional implementation, in this embodiment, the target optimization function may be an unsupervised optimization function that is constructed based on the second feature matrix and the multi-view data, and is respectively composed of a global local structure retention term, a regression term, and a regularization term, and the view weight is adjusted by an adaptive method. Illustratively, in this embodiment, based on the second feature matrix and the view data, the constructed objective optimization function may be:

in the formula (I), the compound is shown in the specification,

for the global local structure maintenance item constructed based on the second feature matrix, the local neighbor relation between the view data can be measured from the perspective of the overall view data; v is the number of views, and gamma is a hyper-parameter, and is used for adjusting the influence of a regularization term on the whole optimization problem; />

For regression terms, linear projection fitting nonlinear mapping can be used for construction, so that the feature extraction methodThe method has the capability of nonlinear dimension reduction, and can effectively improve the feature extraction effect; />

In order to maximize the divergence of the variables after the feature extraction, the correlation among the variables after the feature extraction can be reduced, and the learning accuracy of a subsequent model can be effectively improved; alpha is alpha _i ，Y，P ⁽ⁱ⁾ For the optimization goal of the optimization problem, where α _i Weighting the balance factor for the adaptive view; p is _(i) Extracting a dimension reduction matrix for the features under each view; and Y is the feature extraction result of the current data. />

In the embodiment, for the service data without a label, because the target optimization function constructed based on the second feature matrix can measure the local neighbor relation between the view data from the perspective of the whole view data, the information from different view data can be fully utilized, the core characteristics of each view data are retained, the missing information is made up, and the effect that the service data after dimension reduction is more comprehensive and more accurate can be achieved.

In the scenario of solving the objective optimization function to obtain the feature extraction result of the service data, the objective optimization function may be solved to obtain the feature extraction matrix of the service data, and the feature extraction result of the service data may be obtained based on the feature extraction matrix. In one embodiment, as shown in fig. 4, the step S103 includes:

s401, solving the objective optimization function, and acquiring a feature extraction matrix of the service data.

In this embodiment, the process of solving the objective optimization function may be: setting the initial value of the hyper-parameter gamma and initializing the parameter alpha _i And =1/v, calculating intermediate data so as to obtain a feature extraction matrix of the service data. It can be understood that the objective optimization functions constructed for tagged business data and untagged business data are differentTherefore, the formulas of the intermediate data solved by the two are also different, and the two cases will be described separately below. Optionally, the service data with the label can be processed by a formula

Calculating intermediate data thereof; for traffic data without labels, the formula for calculating the intermediate data may be

In the formula, L is a Laplace matrix based on a global local structure maintaining matrix; further, a cycle end condition may be set: a is less than or equal to sigma or the cycle number reaches k times, and the following formula is applied to loop iteration calculation results:

(1) Solving by using generalized eigenvalue solving method

The solution of Y is a matrix composed of eigenvectors corresponding to the first M minimum eigenvalues.

(2) Solving for

In the formula, alpha _i For view weight, further, a feature extraction matrix can be obtained based on the solution result

S402, acquiring a feature extraction result based on the feature extraction matrix.

In this embodiment, the computer device obtains the feature extraction matrix of the service data through the solving process as

Y in the formula is a feature extraction result corresponding to the service data, and after the computer equipment obtains the feature extraction matrix, the feature extraction can be extractedAnd Y in the matrix is used for obtaining the feature extraction result of the service data.

In this embodiment, the constructed objective optimization function is solved, so that the feature extraction matrix of the service data can be accurately obtained, the feature extraction result of the service data can be accurately obtained based on the feature extraction matrix, and the accuracy of the obtained feature extraction result is improved.

After the feature extraction matrix is obtained, if new to-be-processed service data needs to be processed, the obtained feature extraction matrix may be directly used to process the new to-be-processed service data, and the process will be described in detail below. In one embodiment, the method further comprises: and acquiring a feature extraction result of the new service data to be processed by using the feature extraction matrix and the new service data to be processed.

Specifically, in this embodiment, the new to-be-processed service data may be service data with a tag or service data without a tag, and it can be understood that, if the new to-be-processed service data is service data with a tag, the computer device may calculate the new to-be-processed service data by using a computer

Quickly obtaining corresponding characteristic extraction result, wherein X is ⁽ⁱ⁾ Represents new pending traffic data, based on the sum of the data values in the database and the sum of the data values in the database>

The method comprises the steps of representing a characteristic extraction matrix corresponding to service data with labels; if the new service data to be processed is service data without a label, the computer device can calculate ≥ whether or not the service data is more or less labeled>

Quickly obtaining corresponding characteristic extraction result, wherein X is ⁽ⁱ⁾ Represents new pending traffic data, is asserted>

Whether or not to indicateAnd extracting a characteristic extraction matrix corresponding to the signed service data.

In this embodiment, for new to-be-processed service data, the computer device can quickly obtain a feature extraction result of the new to-be-processed service data by using the obtained feature extraction matrix and the new to-be-processed service data, so as to improve efficiency of obtaining the feature extraction result of the new to-be-processed service data.

Exemplarily, when the service data to be processed is service data with a tag, please refer to fig. 5, and the process for processing the service data may refer to the process described in the following example: taking banking business data with 500 dimensions as example of business data to be processed, firstly, the 500-dimensional business data can be split into 5 view data according to customer basic information, transaction information, asset information, product holding information and behavior information, and then an indication matrix C is constructed _k And an objective optimization function A, setting the hyperparameter gamma =0.5, the cycle ending condition A is less than or equal to 0.001 or the cycle frequency reaches 50 times, solving the objective optimization function, and obtaining M through calculation ⁽ⁱ⁾ And after H, iteratively solving the low-dimensional representation and the view weight of the view data until a loop stopping condition is met, wherein the obtained Y is a first feature extraction result,

extracting the matrix, alpha, for the first feature _i Is the view weight. Further, label reconstruction can be performed on each view data in the initial feature extraction matrix, an optimized indication matrix is constructed based on original view data labels and reconstruction labels, a target optimization function is constructed based on the optimized indication matrix, hyper-parameters and iteration stop conditions are set to solve the target optimization function, and a new feature extraction result Y is obtained>

And alpha _i . Furthermore, if new service data to be processed is available, the calculation can be carried out on ≥ er>

And rapidly obtaining the feature extraction result of the newly added service data to be processed.

For example, when the service data to be processed is service data without a tag, please refer to fig. 6, and for the processing procedure of the service data, refer to the procedure described in the following example: taking banking business data with 500 dimensions as business data to be processed as an example, firstly, 500-dimensional data can be split into 5 view data according to customer basic information, transaction information, asset information, product holding information and behavior information, then a global local structure retention matrix W and a target optimization function are constructed, a hyper-parameter gamma =0.5, a cycle ending condition A is less than or equal to 0.001 or the cycle frequency reaches 50 times, the target optimization function is solved, and M is obtained through calculation ⁽ⁱ⁾ And after H, iteratively solving the low-dimensional representation and the view weight of the view data until a loop stopping condition is met, wherein the obtained Y is a first feature extraction result,

extracting the matrix, alpha, for the first feature _i Is the view weight. If newly added service data to be processed can be counted and/or judged>

For the labeled service data, the following detailed description describes a method for processing the service data provided by the present disclosure, which may include:

s1, analyzing, splitting and classifying the to-be-processed service data based on the realistic significance of each dimension characteristic to form multi-view data.

S2, constructing an indication matrix based on the overall view data and the label,

the indicating matrix for representing the kth class is constructed as follows:

and S3, constructing a supervised target optimization function for the generated multi-view data. The target optimization function is composed of an information discrimination term, a regression term and a regular term respectively, and the view weight is adjusted by a self-adaptive method, and the specific target optimization function is as follows:

in the formula, v is the number of views, and gamma is a hyper-parameter and is used for adjusting the influence of a regularization term on the whole optimization problem;

for the information discrimination item, the characteristic extraction effect can be improved in a mode of aggregating similar sample points as closely as possible; />

In order to maximize the divergence of the variables after the feature extraction, the correlation between the variables after the feature extraction can be reduced, and the learning accuracy of a subsequent model can be effectively improved; alpha is alpha _i ，Y，P ⁽ⁱ⁾ For the optimization goal of the optimization problem, where α _i Weighting factors for adaptive viewsA seed; p _(i) Extracting a dimension reduction matrix for the features under each view; and Y is the feature extraction result of the current data.

S4, setting the hyper-parameter gamma to a proper value and initializing the parameter alpha _i ＝1/v。

S5, solving an objective optimization function and calculating intermediate data

Optimizing the extraction result of the objective optimization function, and setting the cycle end condition: a is less than or equal to sigma or the cycle number reaches k times. And (3) circularly and iteratively calculating the result by using the following formula:

(2) Solving for

/>

S6, further, obtaining a characteristic extraction matrix based on the solving result

In the formula, Y is a feature extraction result corresponding to the service data.

And S7, carrying out label reconstruction on the target view data, and searching K view data closest to the target view data, wherein the reconstructed label of the target view data is the label with the largest proportion in the K adjacent view data. If a plurality of labels in the target view data are the same, the label corresponding to the view data closest to the target view data in the view data corresponding to the labels can be selected as the reconstruction label of the target view data.

S8, constructing an optimized indication matrix based on the whole original view data and the reconstructed labels

The steps S3 to S8 are repeated, and the data-corrected feature matrix is obtained>

And the result of the feature extraction->

S9, for the new service data to be processed, calculation can be carried out

And quickly obtaining a corresponding feature extraction result.

In addition, for the service data without a tag, the following describes in detail a method for processing the service data provided by the present disclosure, and the method may include:

S2, constructing a global local structure retention matrix based on the overall view data:

if v is _i And v _j Under all views are neighbors.

And S3, constructing an unsupervised target optimization function for the generated multi-view data. The target optimization function is composed of a global local structure retention term, a regression term and a regular term respectively, and the view weight is adjusted through a self-adaptive method, and the specific target optimization function is as follows:

maintaining items for global local structure, and measuring local neighbor relation between view data from the perspective of the overall view data; />

In order to maximize the divergence of the variables after the feature extraction, the correlation among the variables after the feature extraction can be reduced, and the learning accuracy of a subsequent model can be effectively improved; alpha is alpha _i ，Y，P ⁽ⁱ⁾ For the optimization goal of the optimization problem, where α _i Weighting the balance factor for the adaptive view; p is _(i) Extracting a dimension reduction matrix for the features under each view; and Y is the feature extraction result of the current data.

S4, setting the hyper-parameter gamma to be a proper value and initializing the parameter alpha _i ＝1/v。

L is a Laplace matrix based on a global local structure maintaining matrix; optimizing the extraction result of the objective optimization function, and setting the cycle end condition: a is less than or equal to sigma or the cycle number reaches k times. And (3) circularly and iteratively calculating the result by using the following formula: solving by using generalized eigenvalue solving method

Solving Y into matrix formed by eigenvectors corresponding to the first M minimum eigenvalues, and solving

S6, obtaining a feature extraction matrix based on the solving result

In the formula, Y represents a feature extraction result.

S7, for new data to be processed, calculation can be carried out

And quickly obtaining a corresponding feature extraction result.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the present application further provides a service data processing apparatus for implementing the service data processing method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so that specific limitations in the following embodiments of one or more service data processing devices may refer to the limitations in the above service data processing method, and details are not described herein.

In one embodiment, as shown in fig. 7, there is provided a service data processing apparatus, including: a dividing module 10, a first constructing module 11 and a first obtaining module 12, wherein:

the dividing module 10 is configured to divide the service data to be processed into multiple types of view data according to a preset classification dimension.

The first construction module 11 is configured to construct a feature matrix by using the view data for each type of view data, and construct an objective optimization function based on the feature matrix and the view data.

The first obtaining module 12 is configured to solve the objective optimization function and obtain a feature extraction result of the service data; the feature extraction result is used for representing feature information of the service data.

The processing apparatus for service data provided in this embodiment may execute the foregoing method embodiment, and the implementation principle and technical effect are similar, which are not described herein again.

In an embodiment, the feature matrix is a first feature matrix or a second feature matrix, as shown in fig. 8, the first building module 11 includes: a determination unit 111, a first construction unit 112 and a second construction unit 113, wherein:

a determining unit 111, configured to determine whether the service data has a tag; the label is used for representing the type information of the service data.

The first constructing unit 112 is configured to construct a first feature matrix by using the view data and the label corresponding to the view data if the service data has the label.

A second constructing unit 113, configured to construct a second feature matrix using the view data if the service data does not have a label.

In an embodiment, the first constructing unit 112 is configured to sort the view data to obtain sorted view data; and constructing a first characteristic matrix according to the label corresponding to the sorted view data and the position information of the sorted view data.

In one embodiment, as shown in fig. 9, the first building block 11 further includes: a third building element 114, wherein:

a third construction unit 114, configured to construct an objective optimization function based on the first feature matrix and the view data.

In one embodiment, as shown in fig. 10, the above apparatus further comprises: a second acquisition module 13, a reconstruction module 14, and a second construction module 15 and a third acquisition module 16, wherein:

the second obtaining module 13 is configured to determine the target tag as a reconstruction tag of the view data, and construct an optimized first feature matrix according to the view data and the reconstruction tag corresponding to the view data.

And the reconstruction module 14 is configured to determine the target tag as a reconstruction tag of the view data, and construct an optimized first feature matrix according to the view data and the reconstruction tag corresponding to the view data.

And a second constructing module 15, configured to construct an adjusted objective optimization function based on the optimized first feature matrix and the view data.

And a third obtaining module 16, configured to solve the adjusted target optimization function, and obtain a feature extraction result.

In an embodiment, the second constructing unit 113 is configured to construct the second feature matrix based on the neighbor relation information of the view data.

In an embodiment, the third constructing unit 114 is configured to construct the objective optimization function based on the second feature matrix and the view data.

In one embodiment, as shown in fig. 11, the first obtaining module 12 includes: a solving unit 121 and an obtaining unit 122, wherein:

and the solving unit 121 is configured to solve the objective optimization function to obtain a feature extraction matrix of the service data.

An obtaining unit 122, configured to obtain a feature extraction result based on the feature extraction matrix.

In one embodiment, as shown in fig. 12, the above apparatus further comprises: a fourth obtaining module 17, wherein:

and a fourth obtaining module 17, configured to obtain a feature extraction result of the new service data to be processed by using the feature extraction matrix and the new service data to be processed.

The modules in the processing device of the service data may be implemented wholly or partially by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 13. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing business data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of processing business data.

It will be appreciated by those skilled in the art that the configuration shown in fig. 13 is a block diagram of only a portion of the configuration associated with the present application, and is not intended to limit the computing device to which the present application may be applied, and that a particular computing device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:

and dividing the service data to be processed into a plurality of types of view data according to a preset classification dimension.

And aiming at various types of view data, constructing a feature matrix by using the view data, and constructing an objective optimization function based on the feature matrix and the view data.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, performs the steps of:

It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by hardware instructions of a computer program, which may be stored in a non-volatile computer-readable storage medium, and when executed, may include the processes of the embodiments of the methods described above. Any reference to memory, databases, or other media used in the embodiments provided herein can include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), magnetic Random Access Memory (MRAM), ferroelectric Random Access Memory (FRAM), phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases involved in the embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for processing service data, the method comprising:

2. The method of claim 1, wherein the feature matrix is a first feature matrix or a second feature matrix; the constructing a feature matrix by using the view data for each type of the view data comprises:

and if not, constructing the second feature matrix by using the view data.

3. The method according to claim 2, wherein the constructing the first feature matrix using the view data and the label corresponding to the view data comprises:

sequencing the view data to obtain sequenced view data;

4. The method of claim 3, wherein constructing an objective optimization function based on the feature matrix and the view data comprises:

5. The method according to claim 3 or 4, characterized in that the method further comprises:

6. The method of claim 2, wherein the constructing the feature matrix using the view data comprises:

7. The method of claim 6, wherein constructing an objective optimization function based on the feature matrix and the view data comprises:

8. The method according to claim 1, wherein solving the objective optimization function to obtain the feature extraction result of the service data comprises:

9. The method of claim 8, further comprising:

10. An apparatus for processing service data, the apparatus comprising:

the dividing module is used for dividing the service data to be processed into multi-class view data according to the preset classification dimensionality;

the first acquisition module is used for solving the target optimization function and acquiring a feature extraction result of the service data; and the feature extraction result is used for representing feature information of the service data.

11. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the steps of the method of any one of claims 1 to 9 when executing the computer program.

12. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 9.

13. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method of any one of claims 1 to 9 when executed by a processor.