Object clustering method, device and medium

The data is preclustered and subject word distribution information determined through the Dilicre process mixing model, which solves the problems of large data volume and low clustering accuracy and efficiency of complex dimensional object information in the existing technology, and realizes efficient object category division .

CN112508059BActive Publication Date: 2024-11-22YUEKAI SECURITIES
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011298103.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-18
Publication Date
2024-11-22
Estimated Expiration
2040-11-18

AI Technical Summary

Technical Problem

When the prior art clusters object information with large data volume and complex dimensions, the clustering accuracy and efficiency are low, the calculation speed is slow, the termination conditions are unclear, and it is difficult to apply to data with large sample sizes and complex dimensions.

Method used

The Dirichlet process mixing model is adopted to obtain object index data, and determine preclustering and subject word distribution information. Combined with the prior distribution in the Dirichlet process mixing model, the objects are clustered to realize the object category information. Get it.

Benefits of technology

It improves the clustering accuracy and efficiency of large data volume and high-dimensional object data, and is suitable for large sample size and complex dimension data, realizing the effective division of object categories.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112508059B_ABST
    Figure CN112508059B_ABST
Patent Text Reader

Abstract

The present application discloses an object clustering method, device and medium. The present application obtains indicator data of multiple objects and performs pre-clustering to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets; determines a preset number of clustering indicators based on the subject word distribution information corresponding to the preset number of object indicator data sets; determines a target clustering indicator based on a preset object association indicator and a preset number of clustering indicators; determines distribution parameters of the indicator data of multiple objects based on the target clustering indicator; determines a priori distribution obeyed by the distribution parameters of the indicator data of multiple objects; clusters multiple objects based on the priori distribution obeyed by the distribution parameters and the target clustering indicator to obtain object category information of multiple objects, thereby realizing customer category division of customer data with huge data volume and complex dimensions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an object clustering method, device and medium. Background Art

[0002] With the continuous development of big data, through data mining, enterprises can discover the hidden information in object data, thereby effectively identifying different types of customer objects to achieve precision marketing. The research on object clustering is widely used in the financial industry and the communications industry. In the prior art, when training a model, it is difficult to determine the label of the object information in advance. Generally, clustering algorithms such as the K-means algorithm and the improvement of the K-means algorithm are used to cluster the object information. However, these clustering algorithms require the number of clusters to be set in advance, and the selection of this value will greatly affect the effect of clustering, limiting the accuracy of clustering. Alternatively, hierarchical clustering is used to cluster object information, and the number of object categories does not need to be set in advance. However, compared with non-hierarchical algorithms, the algorithm has a very slow calculation speed when the data volume is large, and the termination condition is unclear. It is not suitable for object information with large sample size and complex dimensions. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides an object clustering method, device and medium, which can realize clustering of object indicator data with large data volume and complex dimensions, greatly improving the accuracy and efficiency of object clustering.

[0004] In order to achieve the purpose of the above application, the present application provides an object clustering method, which may include:

[0005] Get indicator data for multiple objects;

[0006] Pre-clustering the indicator data of the multiple objects to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets;

[0007] Based on the subject word distribution information corresponding to the preset number of object indicator data sets, respectively determining the preset number of clustering indicators from the preset number of object indicator data sets;

[0008] Determining target clustering indicator data based on the preset object association indicator and the preset number of clustering indicators;

[0009] Determine prior distributions in Dirichlet process mixture models;

[0010] Based on the prior distribution in the Dirichlet process mixture model and the target clustering index, the multiple objects are clustered to obtain object category information of the multiple objects

[0011] On the other hand, the present application also provides an object clustering device, the device comprising:

[0012] A data acquisition module is used to obtain indicator data of multiple objects;

[0013] A pre-clustering module, used to pre-cluster the indicator data of the multiple objects to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets;

[0014] A clustering index determination module, configured to determine the preset number of clustering indexes from the preset number of object index data sets respectively based on the subject word distribution information corresponding to the preset number of object index data sets;

[0015] A target clustering index determination module, used to determine the target clustering index based on a preset object association index and the preset number of clustering indexes;

[0016] Prior distribution determination module, used to determine the prior distribution in the Dirichlet process mixture model;

[0017] A clustering module is used to cluster the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain object category information of the multiple objects.

[0018] In addition, the present application also provides a storage medium, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement any of the above-mentioned object clustering methods.

[0019] Implementing this application has the following beneficial effects:

[0020] The present application obtains indicator data of multiple objects; pre-clusters the indicator data of the multiple objects to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets; determines the preset number of clustering indicators from the preset number of object indicator data sets based on the subject word distribution information corresponding to the preset number of object indicator data sets; determines target clustering indicator data based on preset object association indicators and the preset number of clustering indicators; determines a priori distribution in a Dirichlet process mixture model; clusters the multiple objects based on the priori distribution in the Dirichlet process mixture model and the target clustering indicators to obtain object category information of the multiple objects, so as to realize object category division of object data with huge data volume and complex dimensions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 A schematic diagram of a process of an object clustering method provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of a process for obtaining indicator data of multiple objects provided in an embodiment of the present application;

[0024] Figure 3 A schematic diagram of a process of clustering using a Dirichlet mixture model provided in an embodiment of the present application;

[0025] Figure 4 A schematic diagram of a process for determining a priori distribution provided in an embodiment of the present application;

[0026] Figure 5 A schematic diagram of a process for obtaining indicator data of multiple objects provided by another embodiment of the present application;

[0027] Figure 6 A schematic diagram of a process of clustering using a Dirichlet mixture model provided in another embodiment of the present application;

[0028] Figure 7 A schematic diagram of a process of clustering using a Dirichlet mixture model provided in another embodiment of the present application;

[0029] Figure 8 A schematic diagram of the structure of an object clustering device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0032] In order to implement the technical solution of the present application and enable more engineering and technical workers to easily understand and apply the present application, the working principle of the present application will be further explained in combination with specific embodiments.

[0033] The present application can be applied to the field of data processing, and in particular can be applied to clustering customers in the financial industry or the communications industry.

[0034] The following describes an embodiment of an object clustering method of the present application. Figure 1 This is a flowchart of an object clustering method provided in an embodiment of the present application. This specification provides the method operation steps described in the embodiment or flowchart, but based on conventional or non-creative labor, more or fewer operation steps may be included. The order of steps listed in the embodiment is only one way of executing the steps among many, and does not represent the only execution order. Specifically, Figure 1 As shown, the method may include:

[0035] S101: Obtain indicator data of multiple objects.

[0036] Specifically, the indicator data of the multiple objects refers to dimensionless standardized data corresponding to each of the multiple objects.

[0037] In some embodiments, Figure 2 As shown, obtaining indicator data of multiple objects may include:

[0038] S1011: Acquire source data of multiple objects.

[0039] Specifically, the source data of the multiple objects may be original object information data collected from a big data center platform, and the source data of the multiple objects may have different dimensions and dimensional units.

[0040] S1013: using a deviation normalization function to normalize the source data of the multiple objects to obtain index data of the multiple objects.

[0041] Since the source data of multiple objects can have different dimensions and dimensional units, the source data of multiple objects can be normalized during data analysis to eliminate the impact of difference data on data analysis. The Min-MaxNormalization deviation normalization function can be used to process the source data of multiple objects into data of the same order of magnitude to obtain the indicator data of multiple objects. The specific formula of the deviation normalization function is as follows:

[0042]

[0043] x * is the index data of the object being sought, x is the source data of the object, min is the minimum source data, and max is the maximum source data.

[0044] S103: Pre-clustering the indicator data of multiple objects to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets.

[0045] Specifically, the latent Dirichlet distribution algorithm can be used to calculate the keyword distribution matrix β, β i,v represents the probability of generating from the topic to the vocabulary v, β i,v The calculation formula is as follows:

[0046]

[0047] Where M is the number of texts expected in the indicator data training of multiple objects, N d is the number of words in the dth text, Represents the nth word w in the dth document n The probability of being generated by the i-th topic, 1(v=w n ) is an indicative function, indicating that when the vocabulary w n When it is v, it returns 1, otherwise it returns 0. Each object indicator data set may include at least one object and indicator data corresponding to the at least one object.

[0048] S105: Based on the topic word distribution information corresponding to the preset number of object indicator data sets, respectively determine a preset number of clustering indicators from the preset number of object indicator data sets.

[0049] Specifically, the preset number of object indicator data sets refers to the number of different topics of texts corresponding to different object category information. When the preset number is T, that is, the initial number of clusters of the topic word distribution matrix β calculated using the latent Dirichlet distribution algorithm is T, then based on the topic word distribution information corresponding to the T object indicator data sets, the words with the highest probability of appearing under each topic can be determined from the T object indicator data sets, and these T words are selected as clustering indicators.

[0050] S107: Determine a target clustering index based on a preset object association index and a preset number of clustering indexes.

[0051] Specifically, the preset object association index refers to the object association index related to the application scenario in the specific implementation and deployment process. For example, when dividing the customer categories in the financial industry, the object association index can be an object association index extracted from the customer segmentation model. The preset object association index in different application scenarios can be different. Based on the preset object association index and the preset number of clustering indicators, the target clustering index is determined, which can be to compare the preset number of clustering indicators with the preset object association indicators to obtain some features that do not intersect the preset number of clustering indicators and the preset object association indicators, and add the partial features to the preset number of clustering indicators as the target clustering indicators. If the number of objects in the object indicator data set is N, then the number of target clustering indicators is [1, N].

[0052] S109: Determining prior distributions in Dirichlet process mixture models.

[0053] Specifically, Figure 3 As shown, determining the prior distribution in the Dirichlet process mixture model can include:

[0054] S1091: Construct a priori distribution obeyed by distribution parameters of indicator data of multiple objects.

[0055] Specifically, the prior distribution obeyed by the distribution parameters is a Dirichlet process.

[0056] S1093: Determine the prior distribution that the latent variable obeys.

[0057] S111: Based on the prior distribution and the target clustering index in the Dirichlet process mixture model, multiple objects are clustered to obtain object category information of the multiple objects.

[0058] Specifically, suppose a dataset of object indicators to be clustered is x n ={a1, a2, ...a m}, where a1, a2, ..., a m For object x n For example, the target clustering indexes of customers 1, 2, ..., N are a1, a2, ..., a n , a1, a2, ..., a m It can be customer age, customer gender, customer education and other characteristics. In the Dirichlet process mixed model, x n Satisfies the following distribution:

[0059] G|{α, G0}~DP(α, G0)

[0060] η n |G~G

[0061] x n |η n ~p(x n |η n )

[0062] Among them, η n For x n The distribution parameter, η n The prior distribution obeyed is a Dirichlet process with a base distribution G0 and a scale parameter α. In the formula, the scale parameter α and the parameters of the base distribution G0 are known. Based on the prior distribution and target clustering index in the Dirichlet process mixture model, multiple objects are clustered to obtain object category information of multiple objects.

[0063] Specifically, the prior distribution in the Dirichlet process mixture model includes the prior distribution obeyed by the distribution parameters of the index data of multiple objects and the prior distribution obeyed by the latent variable, where the latent variable is Z n , V i , Among them, Z n is the quantity to be solved, V i This independent variable follows a Beta distribution, that is, V i ~Beta(1,α), Subordinate to the base distribution G0, i = {1, 2, ...}, Z n , V i , The prior distribution of is known. Figure 4 As shown, based on the prior distribution and target clustering index in the Dirichlet process mixture model, multiple objects are clustered, and the object category information of the multiple objects obtained may include:

[0064] S1111: Construct the variational distribution obeyed by the latent variables.

[0065] Specifically, the latent variables include object category information of multiple objects and independent variables, and the variational distribution includes variational parameters.

[0066] S1113: Update variational parameters.

[0067] Specifically, when the variational parameters are updated, since the variational distribution obeyed by the latent variables contains the variational parameters, the variational distribution obeyed by the latent variables is also updated.

[0068] S1115: Determine whether the relative entropy between the variational distribution obeyed by the latent variable and the posterior distribution obeyed by the latent variable meets the first preset condition.

[0069] Specifically, the posterior distribution obeyed by the latent variable is the posterior distribution corresponding to the prior distribution obeyed by the latent variable

[0070] If not, return to step S1113.

[0071] S1117: Based on the variational distribution obeyed by the updated latent variables and the target clustering index, an approximate solution is obtained for the posterior distribution obeyed by the object category information to obtain the object category information of the multiple objects.

[0072] Specifically, the latent variable Z that needs to be constructed for the variational distribution is n , V i , Among them, Z n The prior distribution of follows a multinomial distribution. It is an object category variable, that is, the quantity that needs to be solved. V i The prior distribution of this independent variable follows the Beta distribution, that is, V i ~Beta(1,α), The prior distribution of follows the base distribution G0, i = {1, 2, ...}, Z n , V i , The prior distribution of is known, and the posterior distribution is solved The object category information Z can be determined n In this embodiment, the variational inference method is used to solve the approximate solution of the posterior distribution. Based on the prior distribution of the latent variable, the variational distribution obeyed by the variational parameter c is constructed. The expression of the variational distribution obeyed by the variational parameter c is:

[0073]

[0074] in, is a Beta distribution, is an exponential family distribution, is a multinomial distribution with variational parameter Let w = {Z, V, η *}, update the variational parameter c, and determine whether the relative entropy of the posterior distribution obeyed by the variational distribution including the variational parameter c and the object category information meets the first preset condition. The first preset condition is that the relative entropy (Kullback-Leibler divergence) of the posterior distribution obeyed by the variational distribution including the variational parameter c and the object category information is minimized. The expression of the relative entropy is:

[0075] D(q c (w)||p(w|x,θ))=E q [log q c (w)]-E q[log p(w,x|θ)]+log p(x|θ)

[0076] In some embodiments, Figure 5 As shown, determining the prior distribution in the Dirichlet process mixture model can include:

[0077] S1092: Using a first preset Dirichlet process, determine a prior distribution obeyed by distribution parameters of indicator data of multiple objects.

[0078] S1094: Using the second preset Dirichlet process, determine the prior distribution obeyed by the latent variable.

[0079] Specifically, the base distribution in the first preset Dirichlet process is the preset base distribution, and the scale parameter is the preset scale parameter, that is, η n For x n The distribution parameter, η n The prior distribution obeyed is a Dirichlet process with a base distribution of G0 and a scale parameter of α. The Dirichlet process can be used to determine the prior distribution obeyed by the distribution parameters of the index data of multiple objects. In addition, the prior distribution obeyed by the hidden variable can be determined by the second preset Dirichlet process.

[0080] Specifically, the variational distribution obeyed by the latent variable includes a variational distribution including independent variables obeying a Beta distribution, independent variables obeying a preset base distribution, and object category variables obeying a multinomial distribution.

[0081] In some embodiments, Figure 6 As shown, based on the prior distribution and the target clustering index in the Dirichlet process mixture model, after clustering multiple objects and obtaining object category information of the multiple objects, the method may further include:

[0082] S113: Analyze the object category information of each object using the silhouette coefficient.

[0083] Specifically, the silhouette coefficient is an analysis method of clustering effect, which can be used to evaluate the impact of different algorithms or different operation modes of algorithms on clustering effect based on the same multiple object index data. For the index data of a single object, let a be the average distance between the index data of the object and the index data of other objects in the same category. The average distance represents the average value of the dissimilarity between the index data vector of the object and the index data of other objects in the same category. Let b be the average distance between the index data of the object and all points in the nearest adjacent cluster, that is, the minimum value of the average dissimilarity between the index data of the object and other clusters. Then the silhouette coefficient of the index data i of an object is:

[0084]

[0085] From the above formula, we can see that the value of S(i) is between [-1, 1].

[0086] S115: Determine whether the analysis result of each object meets a second preset condition.

[0087] Specifically, the second preset condition is a preset S(i) threshold. When S(i) is higher than the preset S(i) threshold, it indicates that the object category analysis result is good.

[0088] S117: When the analysis results of the multiple objects all meet the second preset condition, determine the object category information corresponding to each of the multiple objects as the target object category information corresponding to each of the multiple objects.

[0089] In some embodiments, since the calculation of relative entropy requires the result of the posterior distribution of the hidden variable, the relative entropy cannot be calculated. The variational parameters can be updated by maximizing the lower bound of evidence to minimize the relative entropy. Figure 7 As shown, based on the prior distribution and target clustering index in the Dirichlet process mixture model, multiple objects are clustered to obtain object category information of the multiple objects including:

[0090] S1112: Construct the variational distribution obeyed by the latent variables.

[0091] Specifically, the latent variables include object category information of multiple objects and independent variables, and the variational distribution includes variational parameters.

[0092] S1114: Determine the lower bound of the evidence based on the variational distribution obeyed by the latent variable and the posterior distribution obeyed by the latent variable.

[0093] Specifically, the posterior distribution obeyed by the latent variable is the posterior distribution corresponding to the prior distribution obeyed by the latent variable. ELBO (Evidence Lower Bound) refers to the probability density of object category information. The expression of ELBO is E q [log p(w,x|θ)]-E q [log q c (w)].

[0094] S1116: Update variational parameters.

[0095] Specifically, when the variational parameters are updated, since the variational distribution obeyed by the latent variables contains the variational parameters, the variational distribution obeyed by the latent variables is also updated.

[0096] S1118: Determine whether the lower bound of the evidence satisfies the preset third preset condition.

[0097] Specifically, after calculation, for each variational parameter c iThe update can be obtained by calculating the following expected value:

[0098] c i =E q [g i (w -i , x, θ)]

[0099] Wherein, g is a natural parameter in the exponential family distribution. The third preset condition refers to the convergence of the variational parameter c to a preset threshold.

[0100] If not, return to step S1116.

[0101] S1120: Based on the variational distribution obeyed by the updated latent variables and the target clustering index, the posterior distribution obeyed by the latent variables is solved to obtain object category information of multiple objects.

[0102] The present application also provides an embodiment of an object clustering device, Figure 8 A schematic diagram of the structure of an object clustering device provided in an embodiment, such as Figure 8 As shown, the device may include:

[0103] The data acquisition module 201 is used to acquire indicator data of multiple objects.

[0104] The pre-clustering module 203 is used to pre-cluster the index data of multiple objects to obtain a preset number of object index data sets and the subject word distribution information corresponding to each of the preset number of object index data sets.

[0105] The clustering index determination module 205 is used to determine a preset number of clustering indexes from a preset number of object index data sets based on the subject word distribution information corresponding to the preset number of object index data sets.

[0106] The target clustering index determination module 207 is used to determine the target clustering index based on the preset object association index and a preset number of clustering indexes.

[0107] The prior distribution determination module 209 is used to determine the prior distribution in the Dirichlet process mixture model.

[0108] The clustering module 211 is used to cluster multiple objects based on the prior distribution and target clustering index in the Dirichlet process mixture model to obtain object category information of the multiple objects.

[0109] The present application further provides an object clustering device, characterized in that the device includes a processor and a memory, the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method in any of the above embodiments.

[0110] The present application further provides an embodiment of a computer-readable storage medium, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the method in any of the above embodiments.

[0111] It can be seen from the above embodiments that the present application obtains indicator data of multiple objects, pre-clusters the indicator data of multiple objects, obtains a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets, determines a preset number of clustering indicators from the preset number of object indicator data sets based on the subject word distribution information corresponding to the preset number of object indicator data sets, determines a target clustering indicator based on the preset object association indicator and the preset number of clustering indicators, determines a prior distribution in a Dirichlet process mixture model, clusters multiple objects based on the prior distribution and the target clustering indicator in the Dirichlet process mixture model, and obtains object category information of multiple objects.

[0112] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0113] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the following intention: that the claimed invention requires more features than the features explicitly recited in each claim. More specifically, as reflected in the claims of the present invention, the inventive aspects lie in less than all the features of the single embodiment disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0114] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0115] In addition, those skilled in the art will appreciate that, although the embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims of the present invention, any one of the claimed embodiments may be used in any combination.

[0116] The present invention may also be implemented as a device or system program (such as a computer program and a computer program product) for performing part or all of the methods described herein. Such a program implementing the present invention may be stored on a computer readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or may be provided on a carrier signal, or in any other form.

[0117] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps, etc., that are not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present invention may be implemented by means of hardware comprising several different elements and by means of appropriately programmed computers. In a unit claim that lists several systems, several of these systems may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order, and these words may be interpreted as names.

Claims

1. An object clustering method, characterized in that: The method comprises: Get indicator data for multiple objects; Pre-clustering the indicator data of the multiple objects to obtain a preset number of object indicator data sets and subject word distribution information corresponding to each of the preset number of object indicator data sets; the pre-clustering is based on the implicit Dirichlet distribution algorithm to calculate the subject word distribution matrix β for processing; β i,v represents the probability of generating from the topic to the vocabulary v, β i,v The calculation formula is as follows: Where M is the number of texts expected in the indicator data training of multiple objects, N d is the number of words in the dth text, Represents the nth word w in the dth document n The probability of being generated by the i-th topic, 1(v=w n ) is an indicative function, indicating that when the vocabulary w n When it is v, it returns 1, otherwise it returns 0; each object indicator data set includes at least one object and indicator data corresponding to the at least one object; Based on the subject word distribution information corresponding to the preset number of object indicator data sets, respectively determining the preset number of clustering indicators from the preset number of object indicator data sets; Determining target clustering index data based on a preset object association index and the preset number of clustering indexes; the preset object association index refers to an object association index related to the application scenario; Determine prior distributions in Dirichlet process mixture models; Based on the prior distribution in the Dirichlet process mixture model and the target clustering index, clustering the multiple objects to obtain object category information of the multiple objects; The clustering of the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain the object category information of the multiple objects includes: Based on the variational distribution obeyed by the updated latent variables and the target clustering index, the posterior distribution obeyed by the object category information is solved to obtain the object category information of multiple objects; the expression of the variational distribution is: Among them, q γt (v t ) is the Beta distribution, is an exponential family distribution, is a polynomial distribution; the variational distribution includes a variational parameter; the variational parameter The posterior distribution obeyed by the latent variable is the posterior distribution corresponding to the prior distribution obeyed by the latent variable.

2. The method according to claim 1, characterized in that Determining the prior distribution in the Dirichlet process mixture model includes: Constructing a priori distribution obeyed by distribution parameters of the indicator data of the multiple objects, wherein the priori distribution obeyed by the distribution parameters is a Dirichlet process; Determine the prior distribution that the latent variables follow.

3. The method according to claim 2, characterized in that The clustering of the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain the object category information of the multiple objects includes: Constructing a variational distribution obeyed by the latent variables, wherein the latent variables include object category information of the plurality of objects and independent variables; Updating the variational parameters until the relative entropy between the variational distribution obeyed by the latent variable and the posterior distribution obeyed by the latent variable satisfies a first preset condition, and the posterior distribution obeyed by the latent variable is a posterior distribution corresponding to the prior distribution obeyed by the latent variable; Based on the variational distribution obeyed by the updated latent variable and the target clustering index, an approximate solution is obtained for the posterior distribution obeyed by the latent variable to obtain object category information of the multiple objects.

4. The method according to claim 3, characterized in that The variational distribution obeyed by the latent variable includes a variational distribution including independent variables obeying a Beta distribution, independent variables obeying the preset base distribution, and object category variables obeying a multinomial distribution.

5. The method according to claim 1, characterized in that After clustering the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain object category information of the multiple objects, the method further includes: analyzing the object category information of the plurality of objects using a silhouette coefficient; Determining whether the analysis result of each object meets a second preset condition; When the analysis results of the multiple objects all meet the second preset condition, the object category information corresponding to each of the multiple objects is determined as the target object category information corresponding to each of the multiple objects.

6. The method according to claim 2, characterized in that The clustering of the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain the object category information of the multiple objects includes: Constructing a variational distribution obeyed by the latent variable, wherein the variational distribution includes a variational parameter, the latent variable includes object category information of the plurality of objects and an independent variable, and the posterior distribution obeyed by the latent variable is a posterior distribution corresponding to the prior distribution obeyed by the latent variable; Determining an evidence lower bound based on the variational distribution obeyed by the latent variable and the posterior distribution obeyed by the latent variable; Updating the variational parameter until the evidence lower bound satisfies a preset third preset condition; Based on the variational distribution obeyed by the updated latent variables and the target clustering index, the posterior distribution obeyed by the object category information is solved to obtain the object category information of multiple objects.

7. The method according to claim 1, characterized in that The obtaining of indicator data of multiple objects includes: Get source data for multiple objects; The source data of the multiple objects are normalized using a deviation normalization function to obtain index data of the multiple objects.

8. An object clustering device, characterized in that: The device comprises: A data acquisition module is used to obtain indicator data of multiple objects; A pre-clustering module is used to pre-cluster the indicator data of the multiple objects to obtain a preset number of object indicator data sets and the subject word distribution information corresponding to each of the preset number of object indicator data sets; the pre-clustering is based on the implicit Dirichlet distribution algorithm to calculate the subject word distribution matrix β for processing; β i,v represents the probability of generating from the topic to the vocabulary v, β i,v The calculation formula is as follows: Where M is the number of texts expected in the indicator data training of multiple objects, N d is the number of words in the dth text, Represents the nth word w in the dth document n The probability of being generated by the i-th topic, 1(v=w n ) is an indicative function, indicating that when the vocabulary w n When it is v, it returns 1, otherwise it returns 0; each object indicator data set includes at least one object and indicator data corresponding to the at least one object; A clustering index determination module, configured to determine the preset number of clustering indexes from the preset number of object index data sets respectively based on the subject word distribution information corresponding to the preset number of object index data sets; A target clustering index determination module, used to determine the target clustering index based on a preset object association index and the preset number of clustering indexes; the preset object association index refers to an object association index related to the application scenario; Prior distribution determination module, used to determine the prior distribution in the Dirichlet process mixture model; A clustering module is used to cluster the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain object category information of the multiple objects; clustering the multiple objects based on the prior distribution in the Dirichlet process mixture model and the target clustering index to obtain object category information of the multiple objects includes: Based on the variational distribution obeyed by the updated latent variables and the target clustering index, the posterior distribution obeyed by the object category information is solved to obtain the object category information of multiple objects; the expression of the variational distribution is: Among them, q γt (v t ) is the Beta distribution, is an exponential family distribution, is a polynomial distribution; the variational distribution includes a variational parameter; the variational parameter The posterior distribution obeyed by the latent variable is the posterior distribution corresponding to the prior distribution obeyed by the latent variable.

9. An object clustering device, characterized in that: The device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the object clustering method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the object clustering method according to any one of claims 1 to 7.