Method and apparatus for classifying data objects, electronic device, and storage medium

Abstract

The application discloses a kind of classification methods of data object and its device, electronic equipment and storage medium, it is related to big data technology field or other related fields, wherein, the method comprises: obtaining the object feature set of data object;According to the initial selection feature weight, construct weighted initial selection feature vector, object sample in object feature set is carried out density clustering based on weighted initial selection feature vector, and initial clustering result is output;Based on fine-grained feature weight and initial clustering result, construct weighted fine selection feature vector, carry out secondary clustering on the basis of initial clustering result, obtain final clustering result;For each clustering cluster in final clustering result, the statistical characteristics of each clustering cluster are calculated, and the classification result of data object is constructed based on statistical characteristics and final clustering result.The application solves the technical problems that the classification result accuracy is low in the way of classifying data objects based on statistical indicators and static rule matching in related technologies.

Description

Technical Field

[0001] This invention relates to the field of big data technology or other related technical fields. Specifically, it relates to a method, apparatus, electronic device and storage medium for classifying data objects. Background Technology

[0002] With the rapid development of fintech and the deepening of digital transformation, financial institutions have accumulated massive amounts of data assets. In financial business scenarios, data objects are the basic units constituting business entities, mainly including financial accounts, financial transactions, and financial transaction entities. The ability to perform refined classification of massive amounts of data objects can improve the operational efficiency of financial institutions, optimize resource allocation, and reduce risk, which has significant positive implications for financial business.

[0003] In related technologies, data objects are classified based on statistical indicators and static rules. However, static rules are rigid and cannot adapt to the dynamic changes in the financial environment, which can easily lead to delayed classification results or misjudgments, resulting in inaccurate classification results.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and storage medium for classifying data objects, to at least address the technical problem in related technologies where the classification of data objects based on statistical indicators and static rule matching results in low accuracy.

[0006] According to one aspect of the present invention, a method for classifying data objects is provided, comprising: acquiring a multidimensional heterogeneous dataset of data objects to be classified, and performing data processing on the multidimensional heterogeneous dataset to obtain an object feature set; performing weighted calculation on each sample feature in the object feature set according to preliminary selected feature weights to construct a weighted preliminary selected feature vector; performing density clustering on the object samples in the object feature set based on the weighted preliminary selected feature vector to output an initial clustering result, wherein the preliminary selected feature weights are first preset weight values ​​assigned to the features of a first type of sample; performing weighted calculation on the sample features within each cluster of the initial clustering result based on fine-grained feature weights to construct a weighted fine-grained feature vector, and performing secondary clustering on the object samples based on the initial clustering result and the weighted fine-grained feature vector to obtain a final clustering result, wherein the fine-grained feature weights are second preset weight values ​​assigned to the features of a second type of sample; calculating statistical features for each cluster in the final clustering result, and constructing a classification result of the data objects based on the statistical features and the final clustering result.

[0007] Further, the step of processing the multidimensional heterogeneous dataset to obtain the object feature set includes: cleaning the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain the cleaned multidimensional heterogeneous dataset; standardizing the numerical data in the cleaned multidimensional heterogeneous dataset to obtain standardized numerical features; performing one-hot encoding on the categorical features in the cleaned multidimensional heterogeneous dataset to obtain encoded categorical features; and obtaining the object feature set based on the standardized numerical features and the encoded categorical features.

[0008] Further, the step of performing density clustering on object samples in the object feature set based on weighted initial feature vectors and outputting initial clustering results includes: calculating the weighted Euclidean distance between object samples based on the weighted initial feature vectors to obtain a distance matrix; performing connectivity analysis on the distance matrix using density reachability relations based on preset neighborhood radius and minimum number of points to form multiple density connected regions; identifying density connected regions as initial clusters and marking object samples distributed outside the density connected regions as noise points; removing the noise points and obtaining the initial clustering results based on the multiple initial clusters after removing noise points.

[0009] Furthermore, before performing weighted calculations on each sample feature in the object feature set based on the initial selected feature weights, the method further includes: determining a first type of sample feature in the object feature set, wherein the first type of sample feature includes, but is not limited to, asset-related features; obtaining the distribution standard deviation of the first type of sample feature in the object feature set; calculating a dynamic adjustment coefficient based on the distribution standard deviation, and determining a first preset weight value based on the benchmark weight value and the dynamic adjustment coefficient; assigning the first preset weight value to the first type of sample feature, and assigning a preset static weight value to other sample features in the object feature set besides the first type of sample feature; and obtaining the initial selected feature weights based on the first preset weight value and the static weight value.

[0010] Further, the step of constructing a weighted refined feature vector by weighting the sample features within each cluster of the initial clustering result based on fine-grained feature weights includes: traversing each cluster in the initial clustering result and extracting the initial feature vector of the object samples within the cluster; determining the second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to, life cycle type features; assigning a second preset weight value to the second type of sample features in the initial feature vector according to the preset fine-grained feature weights, and assigning a third preset weight value to other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; and using the second preset weight value and the third preset weight value to perform weighted calculations on the values ​​of each dimension of the initial feature vector to obtain the weighted refined feature vector of the object samples.

[0011] Further, the steps of performing secondary clustering on the object samples based on the initial clustering results and combining the weighted refinement feature vector to obtain the final clustering results include: Step 1, setting a cluster centroid for each cluster in the initial clustering results; Step 2, calculating the weighted distance from each object sample to the cluster centroid based on the weighted refinement feature vector, and assigning the object sample to the cluster centroid with the smallest weighted distance value based on the weighted distance, thereby obtaining a new clustering result; Step 3, updating the cluster centroid of each cluster according to the new clustering result, determining whether the updated cluster centroid meets the iteration condition, and repeating Steps 2 to 3 if the updated cluster centroid does not meet the iteration condition, until the updated cluster centroid meets the iteration condition, thereby obtaining the final clustering result.

[0012] Furthermore, the step of constructing a classification result for the data object based on the statistical features and the final clustering result includes: inputting the statistical features into a pre-constructed large language model, and having the large language model output classification explanation text according to the explanation text template; integrating the final clustering result and the classification explanation text to generate a classification result for the data object.

[0013] According to another aspect of the present invention, a data object classification apparatus is also provided, comprising: an acquisition unit, configured to acquire a multidimensional heterogeneous dataset of data objects to be classified, and perform data processing on the multidimensional heterogeneous dataset to obtain an object feature set; and a first clustering unit, configured to perform weighted calculation on each sample feature in the object feature set according to preliminary selected feature weights, construct a weighted preliminary selected feature vector, perform density clustering on the object samples in the object feature set based on the weighted preliminary selected feature vector, and output an initial clustering result, wherein the preliminary selected feature weights are first preset weights assigned to the features of the first type of sample. The system comprises: a second clustering unit, configured to perform weighted calculations on the sample features within each cluster of the initial clustering result based on fine-grained feature weights, construct a weighted refined feature vector, and perform secondary clustering on the object samples based on the initial clustering result and the weighted refined feature vector to obtain the final clustering result, wherein the fine-grained feature weights are assigned a second preset weight value to the second type of sample features; and a classification unit, configured to calculate the statistical features of each cluster in the final clustering result, and construct a classification result for the data object based on the statistical features and the final clustering result.

[0014] Further, the acquisition unit includes: a first cleaning module, used to clean the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain the cleaned multidimensional heterogeneous dataset; a first processing module, used to standardize the numerical data in the cleaned multidimensional heterogeneous dataset to obtain standardized numerical features; a second processing module, used to perform one-hot encoding on the classification features in the cleaned multidimensional heterogeneous dataset to obtain encoded classification features; and a first acquisition module, used to obtain the object feature set based on the standardized numerical features and the encoded classification features.

[0015] Further, the first clustering unit includes: a first calculation module, used to calculate the weighted Euclidean distance between the object samples based on the weighted initial selection feature vector to obtain a distance matrix; a first analysis module, used to perform connectivity analysis on the distance matrix based on a preset neighborhood radius and minimum number of points parameter, using density reachability relationship to form multiple density connected regions; a first identification module, used to identify the density connected regions as initial clusters and mark the object samples distributed outside the density connected regions as noise points; and a first elimination module, used to eliminate the noise points and obtain the initial clustering result based on the multiple initial clusters after eliminating noise points.

[0016] Furthermore, the data object classification device further includes: a first determining module, configured to determine a first type of sample features in the object feature set, wherein the first type of sample features includes, but is not limited to, asset-related features; a second obtaining module, configured to obtain the distribution standard deviation of the first type of sample features in the object feature set; a second calculating module, configured to calculate a dynamic adjustment coefficient based on the distribution standard deviation, and determine a first preset weight value based on a benchmark weight value and the dynamic adjustment coefficient; a first assigning module, configured to assign the first preset weight value to the first type of sample features, and assign a preset static weight value to other sample features in the object feature set besides the first type of sample features; and a third obtaining module, configured to obtain the initial selected feature weights based on the first preset weight value and the static weight value.

[0017] Further, the second clustering unit includes: a first traversal module, used to traverse each cluster in the initial clustering result and extract the initial feature vector of the object samples within the cluster; a second determination module, used to determine the second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to, life cycle type features; a first assignment module, used to assign a second preset weight value to the second type of sample features in the initial feature vector according to the preset fine-grained feature weights, and assign a third preset weight value to other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; and a third calculation module, used to perform weighted calculation on the values ​​of each dimension of the initial feature vector using the second preset weight value and the third preset weight value to obtain the weighted fine-tuned feature vector of the object samples.

[0018] Furthermore, the second clustering unit further includes: a first setting module, used in step one, for setting a cluster centroid for each cluster in the initial clustering result; a fourth calculation module, used in step two, for calculating the weighted distance from each object sample to the cluster centroid based on the weighted refinement feature vector, and assigning the object sample to the cluster centroid with the smallest weighted distance value based on the weighted distance, to obtain a new clustering result; and a first iteration module, used in step three, for updating the cluster centroid of each cluster according to the new clustering result, determining whether the updated cluster centroid meets the iteration condition, and repeating steps two to three if the updated cluster centroid does not meet the iteration condition, until the updated cluster centroid meets the iteration condition, to obtain the final clustering result.

[0019] Furthermore, the classification unit includes: a first output module, used to input the statistical features into a pre-constructed large language model, and the large language model outputs classification explanation text according to the explanation text template; and a first generation module, used to integrate the final clustering result and the classification explanation text to generate the classification result of the data object.

[0020] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform any of the above-described data object classification methods.

[0021] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any of the above-described data object classification methods.

[0022] According to another aspect of the present invention, a computer program product is also provided, the computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements any of the above-described data object classification methods.

[0023] In this application, the following steps are performed: A multidimensional heterogeneous dataset of data objects to be classified is obtained, and the dataset is processed to obtain an object feature set. Weighted calculations are performed on the features of each sample in the object feature set according to the initial selected feature weights to construct a weighted initial selected feature vector. Density clustering is then performed on the object samples in the object feature set based on the weighted initial selected feature vector, and an initial clustering result is output. Here, the initial selected feature weights are assigned a first preset weight value to the features of the first type of samples. Weighted calculations are then performed on the features of samples within each cluster of the initial clustering result based on the fine-grained feature weights to construct a weighted fine-grained feature vector. Based on the initial clustering result, the object samples are then subjected to secondary clustering using the weighted fine-grained feature vector to obtain a final clustering result. Here, the fine-grained feature weights are assigned a second preset weight value to the features of the second type of samples. For each cluster in the final clustering result, the statistical features of each cluster are calculated, and a classification result for the data objects is constructed based on the statistical features and the final clustering result.

[0024] This application introduces a dynamic feature weighting mechanism, which weights different business features at different clustering stages, improving the sample discriminability in the feature space and enhancing the fit between the final clustering results and real business logic. Simultaneously, a dual-model coupled clustering algorithm is employed for initial and fine-grained clustering, thereby improving the accuracy of data object classification. This addresses the technical problem of low accuracy in classification results associated with methods based on statistical indicators and static rule matching used in related technologies. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a data object classification method is shown.

[0027] Figure 2 This is a flowchart of an optional data object classification method according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of an optional data object classification process according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of an optional data object classification device according to an embodiment of the present invention;

[0030] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs an optional method for classifying data objects according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] It should be noted that the data object classification method and apparatus in this application can be used in the field of big data technology for classifying data objects based on feature-weighted clustering, and can also be used in any field other than big data technology for classifying data objects based on feature-weighted classification. This application does not limit the application field of the data object classification method and apparatus.

[0034] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0035] The following embodiments of the present invention can be applied to classification systems / applications / devices for various data objects. The present invention employs a two-stage clustering framework: an initial clustering model quickly screens noise and focuses on the first type of features; a refined clustering model deeply identifies the second type of features and makes further subdivisions; and a large language model is introduced to interpret the clustering results, improving the accuracy and interpretability of the clustering results. This provides strong flexibility, allowing for differentiated customization to meet different business needs and enhancing adaptability to complex business environments.

[0036] The present invention will now be described in detail with reference to various embodiments.

[0037] Example 1

[0038] According to an embodiment of the present invention, an embodiment of a data object classification method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0039] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a data object classification method is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0040] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the data object classification method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned data object classification method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0043] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0044] Under the aforementioned operating environment, this application provides the following: Figure 2 The data object classification method shown is implemented by a data object classification system.

[0045] Figure 2 This is a flowchart of an optional data object classification method according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes the following steps:

[0046] Step S201: Obtain a multidimensional heterogeneous dataset of data objects to be classified, and perform data processing on the multidimensional heterogeneous dataset to obtain an object feature set.

[0047] In step S201 above, the data objects to be classified include, but are not limited to, financial accounts, financial transactions, and financial transaction entities. For the data objects to be classified, multidimensional heterogeneous datasets are collected from databases and data lakes of financial business systems, and the data is processed to construct a high-quality feature space. The data processing specifically includes data cleaning, standardization, and encoding to solve the problems of noise sensitivity and dimensional imbalance caused by directly inputting raw heterogeneous data into the clustering algorithm, thereby improving the stability of the subsequent clustering process and the effectiveness of feature representation, and laying a solid data foundation for high-precision data object partitioning.

[0048] In one optional embodiment, the step of processing a multidimensional heterogeneous dataset to obtain an object feature set includes: cleaning the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain a cleaned multidimensional heterogeneous dataset; standardizing the numerical data in the cleaned multidimensional heterogeneous dataset to obtain standardized numerical features; performing one-hot encoding on the categorical features in the cleaned multidimensional heterogeneous dataset to obtain encoded categorical features; and obtaining an object feature set based on the standardized numerical features and the encoded categorical features.

[0049] Furthermore, when processing the data, the multidimensional heterogeneous dataset is first traversed to identify and locate missing values ​​and outliers. For example, for numerical features such as total assets and transaction frequency, a median imputation strategy is used to resist the interference of extreme values ​​on statistical characteristics. For categorical features such as account level and business type, mode imputation is used to preserve the main distribution pattern, thereby generating a cleaned multidimensional heterogeneous dataset with complete structure and the lowest noise.

[0050] Subsequently, the numerical data in the cleaned multidimensional heterogeneous dataset were standardized. By calculating the mean and standard deviation, the data were mapped to a unified dimension, eliminating distance metric bias caused by differences in feature magnitudes and obtaining standardized numerical features. This ensures that each numerical variable has an equivalent influence in clustering calculations. Simultaneously, the categorical features in the cleaned multidimensional heterogeneous dataset were one-hot encoded, transforming discrete categorical variables into sparse vectors. This avoids spurious ordering relationships between categories, resulting in encoded categorical features that preserve the essential differences in category information.

[0051] Finally, the standardized numerical features and the coded classification features are concatenated and fused along the feature dimension to form a unified high-dimensional vector representation that includes information on numerical continuity and class discreteness. Based on the standardized numerical features and the coded classification features, an object feature set is obtained, providing a high-quality, dimensionless and structurally unified input data foundation for the subsequent construction of weighted preliminary feature vectors.

[0052] The above steps, through a refined data preprocessing workflow, effectively overcome the problems of dimensional imbalance, noise interference, and category encoding distortion that exist when raw heterogeneous data is directly used for cluster analysis. This significantly improves the accuracy and robustness of feature representation and provides stable and reliable data support for subsequent density-based clustering algorithms.

[0053] Step S202: Calculate the weighted features of each sample in the object feature set according to the initial selected feature weights, construct a weighted initial selected feature vector, perform density clustering on the object samples in the object feature set based on the weighted initial selected feature vector, and output the initial clustering result.

[0054] Step S202 aims to achieve preliminary stratification and noise filtering of data objects by combining a business-oriented dynamic weighting mechanism with a density clustering algorithm. By applying high weights to the features of the first type of samples, the clustering process prioritizes features that are most discriminative to the business objectives, such as asset features and transaction financial features. This effectively overcomes the business logic deviation problem caused by the equal weighting of features in traditional clustering algorithms. At the same time, the characteristics of density clustering are used to automatically identify and remove abnormal noise, improving the robustness and business relevance of the initial selection stage. This provides high-quality, denoised input data that focuses on key features for subsequent fine-grained clustering.

[0055] In an optional embodiment, before performing weighted calculations on each sample feature in the object feature set according to the initially selected feature weights, the method further includes: determining a first type of sample features in the object feature set, wherein the first type of sample features includes, but is not limited to, asset-related features; obtaining the distribution standard deviation of the first type of sample features in the object feature set; calculating a dynamic adjustment coefficient based on the distribution standard deviation, and determining a first preset weight value based on the benchmark weight value and the dynamic adjustment coefficient; assigning the first preset weight value to the first type of sample features, and assigning a preset static weight value to other sample features in the object feature set besides the first type of sample features; and obtaining the initially selected feature weights based on the first preset weight value and the static weight value.

[0056] Furthermore, the initial selection feature weights assign a first preset weight value to the features of the first type of samples. Specifically, the system first performs feature recognition and dynamic weight generation to ensure that the clustering in the initial selection stage can accurately focus on the features of the first type of samples. The system traverses the object feature set and clearly defines the features of the first type of samples. These features of the first type of samples include, but are not limited to: asset features and transaction amount features.

[0057] Subsequently, the system calculates the standard deviation of the distribution of the first type of sample features in the object feature set. This standard deviation is used to measure the dispersion and distinguishability of asset data in the overall data object group. The larger the standard deviation, the more significant the difference between different data objects, and the more valuable it is as a key classification basis.

[0058] Based on the calculated standard deviation of the distribution, the system constructs a dynamic adjustment coefficient, which is positively correlated with the standard deviation of the distribution and is used to quantify the importance level of the features. Subsequently, the system multiplies the baseline weight value by the dynamic adjustment coefficient to determine the first preset weight value. This process allows the weight allocation to automatically adapt to the data distribution characteristics, ensuring that the first-class sample features with high discriminative power receive higher weight incentives.

[0059] After determining the first preset weight value, the system assigns it to the first type of sample features. At the same time, it assigns preset static weight values ​​to other sample features in the object feature set besides the first type of sample features to maintain the baseline influence of other auxiliary features. Finally, by combining the first preset weight value and the static weight value, a complete initial feature weight vector is generated.

[0060] The above steps, by introducing a dynamic weight adjustment mechanism based on the standard deviation of the distribution, can adapt to changes in data distribution, enabling the clustering algorithm to adaptively strengthen the distinguishing ability of key features in the initial selection stage, thereby more accurately identifying specific groups of data objects and improving the business sensitivity and classification accuracy of the initial clustering.

[0061] In one optional embodiment, the step of performing density clustering on object samples in the object feature set based on weighted initial feature vectors and outputting initial clustering results includes: calculating the weighted Euclidean distance between object samples based on the weighted initial feature vectors to obtain a distance matrix; performing connectivity analysis on the distance matrix using density reachability relations based on preset neighborhood radius and minimum number of points parameters to form multiple density connected regions; identifying density connected regions as initial clusters and marking object samples distributed outside the density connected regions as noise points; removing noise points and obtaining initial clustering results based on the multiple initial clusters after removing noise points.

[0062] Furthermore, firstly, the system calculates the weighted Euclidean distance between object samples based on the weighted initial selection feature vectors. This distance metric incorporates the high weights assigned to key features during the initial selection stage, thereby generating a distance matrix that reflects the true business similarity of the samples.

[0063] Subsequently, the system introduces preset neighborhood radius and minimum number of points parameters, and performs connectivity analysis based on the distance matrix: for each sample in the distance matrix, a neighborhood is defined with that sample as the center and the neighborhood radius as the range. The number of samples falling into this neighborhood is counted. If the number is not less than the minimum number of points, the sample is marked as a core sample. Next, the system traverses the core samples using density reachability relationships. Starting from any unvisited core sample, it aggregates all core samples and their boundary samples that are directly and indirectly density reachable, forming densely connected regions. In this process, the system identifies each independent density connected region as an initial cluster, which represents a potential group of objects with high density characteristics.

[0064] Simultaneously, object samples that do not belong to any density-connected region, i.e., those located outside of density-connected regions, are marked as noise points. These noise points typically represent abnormal transactions, data errors, or atypical edge data objects. Finally, the system removes the marked noise points to ensure the data purity for the next stage and generates initial clustering results based on the multiple initial clusters after removing noise points.

[0065] The above steps, through density connectivity analysis, effectively overcome the shortcomings of traditional clustering algorithms, such as sensitivity to initial centroids and difficulty in handling non-convex cluster structures. At the same time, the noise removal mechanism improves the robustness of the data, providing a high-quality, denoised, and structurally clear input foundation for subsequent fine-grained clustering based on fine-grained features.

[0066] Specifically, the formula for calculating the weighted Euclidean distance between object samples is as follows: ,in, This represents the weight value of object sample k. This represents the weighted initial feature vector of object sample i. This represents the weighted initial feature vector of object sample k.

[0067] Step S203: Based on the fine-grained feature weights, the sample features within each cluster of the initial clustering result are weighted and calculated to construct a weighted fine-selection feature vector. Then, based on the initial clustering result, the object samples are clustered a second time using the weighted fine-selection feature vector to obtain the final clustering result.

[0068] Step S203 aims to achieve more refined clustering operations based on the initial clustering by introducing fine-grained feature weighting and a secondary clustering mechanism, in order to capture subtle differences within different clusters based on lifecycle or behavioral characteristics. Based on the initial clustering results, the system introduces fine-grained feature weights, assigning a second preset weight value to the features of the second type of samples. These second-type sample features typically include indicators used to distinguish the lifecycle of data objects, such as occupational stage or consumption preferences. By assigning them higher second preset weight values, the secondary clustering process pays more attention to the contribution of these fine-grained features to sample similarity.

[0069] The above steps, through a cluster-independent, fine-grained weighting strategy, significantly improve the granularity and business adaptability of the clustering results, providing a more detailed and accurate basis for subsequent precise sales, personalized services, and risk control.

[0070] In one optional embodiment, the step of constructing a weighted refined feature vector by weighting the sample features within each cluster of the initial clustering result based on fine-grained feature weights includes: traversing each cluster in the initial clustering result and extracting the initial feature vector of the object samples within the cluster; determining a second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to, life cycle type features; assigning a second preset weight value to the second type of sample features in the initial feature vector according to preset fine-grained feature weights, and assigning a third preset weight value to other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; and using the second preset weight value and the third preset weight value to perform a weighted calculation on the values ​​of each dimension of the initial feature vector to obtain the weighted refined feature vector of the object samples.

[0071] Furthermore, the step of constructing a weighted fine-selection feature vector by weighting the sample features within each cluster of the initial clustering results based on fine-grained feature weights aims to strengthen the dominant position of life cycle class features in secondary clustering through a cluster-independent fine-weighting strategy, thereby achieving a more refined group division.

[0072] Specifically, the system first traverses each cluster in the initial clustering results, extracting an initial feature vector from each cluster containing all object samples within that cluster. This ensures that subsequent weighting operations are performed only within the homogeneous groups already formed in the initial selection phase. Subsequently, the system clearly defines a second type of sample feature in the object feature set. This feature primarily covers indicators reflecting lifecycle attributes, such as career stage and account level. Based on preset fine-grained feature weighting rules, the system assigns higher second preset weight values ​​to the second type of sample features in the initial feature vector, while assigning lower third preset weight values ​​to other sample features in the object feature set. This weighting mechanism aims to highlight the crucial role of lifecycle features in distinguishing differences within groups.

[0073] Finally, the system uses the second and third preset weight values ​​to perform weighted calculations on the corresponding dimensions of the initial feature vector, generating a weighted refined feature vector of object samples that combines numerical standardization and differences in feature importance.

[0074] The above steps, by independently performing differentiated weighting within each cluster, avoid feature interference that may be caused by global weighting. This allows secondary clustering to more sensitively capture subtle feature differences based on the life cycle, thereby constructing a weighted and refined feature vector that can accurately reflect the life cycle stage of the data object, laying a data foundation for the refinement of the final clustering results.

[0075] In one optional embodiment, the steps of performing secondary clustering on object samples based on the initial clustering results and combining weighted refinement feature vectors to obtain the final clustering results include: Step 1, setting a cluster centroid for each cluster in the initial clustering results; Step 2, calculating the weighted distance from each object sample to the cluster centroid based on the weighted refinement feature vectors, and assigning the object samples to the cluster centroid with the smallest weighted distance value based on the weighted distance, thereby obtaining a new clustering result; Step 3, updating the cluster centroids of each cluster according to the new clustering results, determining whether the updated cluster centroids meet the iteration conditions, and repeating Steps 2 to 3 if the updated cluster centroids do not meet the iteration conditions, until the updated cluster centroids meet the iteration conditions, thereby obtaining the final clustering result.

[0076] Furthermore, the step of performing secondary clustering on the object samples based on the initial clustering results and combining weighted refined feature vectors to obtain the final clustering results aims to achieve refined partitioning based on lifecycle features within the initial clusters through an iterative optimization mechanism. Specifically, for each cluster in the initial clustering results, the system first sets an initial cluster centroid, which is determined by the mean position of all samples in the cluster in the weighted refined feature vector space, serving as the initial reference point for secondary clustering.

[0077] Subsequently, the system calculates the weighted distance from each object sample to the centroid of the current cluster based on the weighted refined feature vector. This distance fully considers the high weight of the second type of sample features, ensuring that sample allocation is more inclined towards data objects with similar lifecycles. Based on the calculated weighted distance, each object sample is assigned to the cluster to which the centroid of the cluster with the smallest weighted distance value belongs, thus obtaining a new clustering result for preliminary reorganization.

[0078] Subsequently, the system recalculates and updates the centroids of each cluster based on the new clustering results, ensuring they reflect the distribution centers of the latest samples within each cluster. The system then determines whether the updated centroids meet the iteration conditions, which typically include a centroid position change of less than a preset threshold or reaching the maximum number of iterations over multiple iterations. If the updated centroids do not meet the iteration conditions, the system repeats the steps of allocating samples based on weighted distance and updating centroids until the updated centroids meet the iteration conditions, i.e., the centroids stabilize. At this point, iteration stops and the final clustering result is obtained.

[0079] Specifically, the secondary clustering is performed on each initial cluster. Perform a weighted clustering algorithm independently to minimize the sum of squares within each cluster: ,in For the first Clusters, Let be the centroid of the cluster. The weighted clustering algorithm iteratively updates the cluster assignments and centroid positions until the centroids stabilize or the maximum number of iterations is reached, yielding the final clustering result.

[0080] The above steps, through local iterative optimization, effectively overcome the shortcomings of static partitioning in adapting to complex distributions within clusters, ensuring the internal consistency and separation of the secondary clustering results in the lifecycle feature dimension, thereby generating a final clustering result with stable structure and clear business meaning.

[0081] Step S204: For each cluster in the final clustering result, calculate the statistical characteristics of each cluster, and construct the classification result of the data object based on the statistical characteristics and the final clustering result.

[0082] In step S204 above, by quantitatively analyzing the internal statistical characteristics of each cluster in the final clustering result, the abstract clusters are transformed into classification labels with clear business meanings, thereby completing the mapping from data grouping to business classification. Specifically, the system first traverses each cluster in the final clustering result, and extracts the statistical characteristics of all object samples contained in each cluster on key feature dimensions. The statistical characteristics include, but are not limited to, the mean, median, mode, standard deviation, and quantiles of each feature value, in order to comprehensively characterize the central tendency and dispersion of the sample group within the cluster in terms of asset size, life cycle structure, and behavioral patterns.

[0083] Subsequently, based on the extracted set of statistical features and the identification information of the clusters, the system constructs the classification results of the data objects. The classification results not only record the cluster number to which each data object belongs, but also associate the classification explanation text of the cluster, thus making the originally black-box cluster output transparent and interpretable, providing an intuitive basis for subsequent business decisions.

[0084] In one optional embodiment, the step of constructing a classification result of a data object based on statistical features and the final clustering result includes: inputting the statistical features into a pre-built large language model, and having the large language model output classification explanation text according to the explanation text template; integrating the final clustering result and the classification explanation text to generate a classification result of the data object.

[0085] Furthermore, after obtaining the final clustering results, the system first extracts the statistical features of each cluster in the final clustering results and sends these statistical features as input data to a pre-built large language model. Upon receiving the input, the large language model performs semantic analysis and logical reasoning on the statistical features based on internally stored explanatory text templates, generating a classification explanation text corresponding to each cluster. This text details the typical characteristics of the data objects within that cluster, such as transaction preferences, transaction risks, and the business implications of account risk levels. Subsequently, the system integrates the original final clustering results with the generated classification explanation texts to generate a comprehensive classification result containing information such as clustering results, cluster labels, and classification explanation texts.

[0086] The above steps, through the parsing capabilities of the large language model, automatically complete the conversion from data features to business language, improving the operability and application efficiency of classification results in actual business scenarios, such as product recommendation and business risk control, and realizing the effective connection between technical analysis and business decision-making.

[0087] Through the above steps, a multidimensional heterogeneous dataset of data objects to be classified is obtained, and the multidimensional heterogeneous dataset is processed to obtain an object feature set. Based on the initial selected feature weights, the features of each sample in the object feature set are weighted to construct a weighted initial selected feature vector. Based on the weighted initial selected feature vector, density clustering is performed on the object samples in the object feature set to output the initial clustering result. Here, the initial selected feature weights are assigned a first preset weight value to the features of the first type of samples. Based on the fine-grained feature weights, the features of samples within each cluster of the initial clustering result are weighted to construct a weighted fine-grained feature vector. Based on the initial clustering result, the object samples are then subjected to secondary clustering using the weighted fine-grained feature vector to obtain the final clustering result. Here, the fine-grained feature weights are assigned a second preset weight value to the features of the second type of samples. For each cluster in the final clustering result, the statistical features of each cluster are calculated, and the classification result of the data objects is constructed based on the statistical features and the final clustering result.

[0088] This embodiment introduces a dynamic feature weighting mechanism, which weights different business features at different clustering stages, improving the sample discriminability in the feature space and enhancing the alignment between the final clustering results and real business logic. Simultaneously, a dual-model coupled clustering algorithm is employed for initial and refined clustering, thereby improving the accuracy of data object classification. This addresses the technical problem in related technologies where data object classification based on statistical indicators and static rule matching results in low accuracy.

[0089] The following describes in detail another optional implementation method.

[0090] Figure 3 This is a schematic diagram of an optional data object classification process according to an embodiment of the present invention, such as... Figure 3 As shown, the data object classification process includes:

[0091] Step 1: Multi-dimensional data collection;

[0092] Step 2: Data preprocessing and feature standardization;

[0093] Step 3: Construct a weighted initial feature vector and perform feature weighting;

[0094] Step four, first-stage clustering;

[0095] Step 5: Anomaly detection and removal;

[0096] Step 6: Generate initial clustering results;

[0097] Step 7: Construct a weighted, refined feature vector and perform feature weighting;

[0098] Step 8, second-stage clustering;

[0099] Step nine: Generate the final clustering results;

[0100] Step 10: Determine the classification results of the data objects;

[0101] Step 11: The large language model outputs explanatory text;

[0102] Step 12: Generate interpretable classification results.

[0103] The embodiments of this invention employ a two-stage clustering framework. An initial clustering model is used to quickly screen for noise and focus on the first type of features. A refined clustering model is then used to deeply identify the second type of features and further subdivide them. At the same time, a large language model is introduced to interpret the clustering results, which improves the accuracy and interpretability of the clustering results. This approach is highly flexible and can be customized to meet different business needs, thereby enhancing adaptability to complex business environments.

[0104] The following is a detailed description with reference to another embodiment.

[0105] Example 2

[0106] The data object classification device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in the above embodiment one. The specific implementation method and beneficial effects can be referred to the foregoing method embodiment, and will not be repeated here.

[0107] Figure 4 This is a schematic diagram of an optional data object classification device according to an embodiment of the present invention, such as... Figure 4 As shown, the data object classification device may include: an acquisition unit 41, a first clustering unit 42, a second clustering unit 43, and a classification unit 44, wherein,

[0108] The acquisition unit 41 is used to acquire a multidimensional heterogeneous dataset of data objects to be classified, and to process the multidimensional heterogeneous dataset to obtain an object feature set.

[0109] The first clustering unit 42 is used to perform weighted calculation on the features of each sample in the object feature set according to the initial selected feature weights, construct a weighted initial selected feature vector, perform density clustering on the object samples in the object feature set based on the weighted initial selected feature vector, and output the initial clustering result. The initial selected feature weights are the first preset weight values ​​assigned to the features of the first type of sample.

[0110] The second clustering unit 43 is used to perform weighted calculation on the sample features within each cluster of the initial clustering result based on fine-grained feature weights, construct a weighted fine-selection feature vector, and perform secondary clustering on the object samples based on the initial clustering result and the weighted fine-selection feature vector to obtain the final clustering result. The fine-grained feature weights are assigned a second preset weight value to the features of the second type of sample.

[0111] Classification unit 44 is used to calculate the statistical characteristics of each cluster in the final clustering result, and construct the classification result of the data object based on the statistical characteristics and the final clustering result.

[0112] The aforementioned data object classification device acquires a multidimensional heterogeneous dataset of data objects to be classified through acquisition unit 41, and processes the multidimensional heterogeneous dataset to obtain an object feature set. A first clustering unit 42 performs weighted calculations on the features of each sample in the object feature set according to the initial selected feature weights, constructing a weighted initial selected feature vector. Based on the weighted initial selected feature vector, density clustering is performed on the object samples in the object feature set, outputting an initial clustering result. The initial selected feature weights are assigned a first preset weight value to the features of the first type of samples. A second clustering unit 43 performs weighted calculations on the features of samples within each cluster of the initial clustering result based on fine-grained feature weights, constructing a weighted refined feature vector. Based on the initial clustering result, the weighted refined feature vector is combined with the object samples to perform secondary clustering, obtaining a final clustering result. The fine-grained feature weights are assigned a second preset weight value to the features of the second type of samples. A classification unit 44 calculates the statistical features of each cluster in the final clustering result, and constructs a classification result for the data objects based on the statistical features and the final clustering result.

[0113] This embodiment introduces a dynamic feature weighting mechanism, which weights different business features at different clustering stages, improving the sample discriminability in the feature space and enhancing the alignment between the final clustering results and real business logic. Simultaneously, a dual-model coupled clustering algorithm is employed for initial and refined clustering, thereby improving the accuracy of data object classification. This addresses the technical problem in related technologies where data object classification based on statistical indicators and static rule matching results in low accuracy.

[0114] Furthermore, the acquisition unit includes: a first cleaning module, used to clean the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain a cleaned multidimensional heterogeneous dataset; a first processing module, used to standardize the numerical data in the cleaned multidimensional heterogeneous dataset to obtain standardized numerical features; a second processing module, used to perform one-hot encoding on the classification features in the cleaned multidimensional heterogeneous dataset to obtain encoded classification features; and a first acquisition module, used to obtain an object feature set based on the standardized numerical features and the encoded classification features.

[0115] Further, the first clustering unit includes: a first calculation module, used to calculate the weighted Euclidean distance between object samples based on the weighted initial selected feature vectors to obtain a distance matrix; a first analysis module, used to perform connectivity analysis on the distance matrix based on a preset neighborhood radius and minimum number of points, using density reachability relations to form multiple density connected regions; a first identification module, used to identify density connected regions as initial clusters and mark object samples distributed outside the density connected regions as noise points; and a first removal module, used to remove noise points and obtain initial clustering results based on the multiple initial clusters after removing noise points.

[0116] Furthermore, the data object classification device further includes: a first determining module, used to determine a first type of sample features in the object feature set, wherein the first type of sample features includes, but is not limited to, asset features; a second obtaining module, used to obtain the distribution standard deviation of the first type of sample features in the object feature set; a second calculating module, used to calculate a dynamic adjustment coefficient based on the distribution standard deviation, and determine a first preset weight value based on the benchmark weight value and the dynamic adjustment coefficient; a first assigning module, used to assign the first preset weight value to the first type of sample features, and assign a preset static weight value to other sample features in the object feature set besides the first type of sample features; and a third obtaining module, used to obtain the initial selection feature weights based on the first preset weight value and the static weight value.

[0117] Further, the second clustering unit includes: a first traversal module, used to traverse each cluster in the initial clustering result and extract the initial feature vector of the object samples within the cluster; a second determination module, used to determine the second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to, life cycle type features; a first assignment module, used to assign a second preset weight value to the second type of sample features in the initial feature vector according to preset fine-grained feature weights, and assign a third preset weight value to other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; and a third calculation module, used to perform weighted calculation on the values ​​of each dimension of the initial feature vector using the second preset weight value and the third preset weight value to obtain the weighted fine-tuned feature vector of the object samples.

[0118] Furthermore, the second clustering unit also includes: a first setting module, used in step one, for setting the centroid of each cluster in the initial clustering result; a fourth calculation module, used in step two, for calculating the weighted distance from each object sample to the centroid of the cluster based on the weighted refinement feature vector, and assigning the object sample to the centroid of the cluster with the smallest weighted distance value based on the weighted distance, to obtain a new clustering result; and a first iteration module, used in step three, for updating the centroid of each cluster according to the new clustering result, determining whether the updated centroid of the cluster meets the iteration condition, and repeating steps two to three if the updated centroid of the cluster does not meet the iteration condition, until the updated centroid of the cluster meets the iteration condition, to obtain the final clustering result.

[0119] Furthermore, the classification unit includes: a first output module, used to input statistical features into a pre-built large language model, and the large language model outputs classification explanation text according to the explanation text template; and a first generation module, used to integrate the final clustering results and the classification explanation text to generate the classification results of the data objects.

[0120] It should be noted that the acquisition unit 41, the first clustering unit 42, the second clustering unit 43, and the classification unit 44 mentioned above correspond to steps S201 to S204 in Embodiment 1. The instances and application scenarios implemented by the above units and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules or units can also be part of a device and can run in the computer terminal 10 provided in Embodiment 1.

[0121] The invention will now be described in conjunction with another alternative embodiment.

[0122] Example 3

[0123] The present invention can also provide an electronic device. Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs an optional data object classification method according to an embodiment of the present invention, such as... Figure 5 As shown, the electronic device may include: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0124] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0125] The processor can access information and applications stored in memory via a transmission device to execute the following steps: acquire a multidimensional heterogeneous dataset of data objects to be classified, and process the multidimensional heterogeneous dataset to obtain an object feature set; perform weighted calculations on the features of each sample in the object feature set according to the initial selected feature weights to construct a weighted initial selected feature vector; perform density clustering on the object samples in the object feature set based on the weighted initial selected feature vectors to output the initial clustering result, wherein the initial selected feature weights are assigned a first preset weight value to the features of the first type of samples; perform weighted calculations on the features of samples within each cluster of the initial clustering result based on the fine-grained feature weights to construct a weighted fine-grained feature vector, and perform secondary clustering on the object samples based on the initial clustering result and the weighted fine-grained feature vectors to obtain the final clustering result, wherein the fine-grained feature weights are assigned a second preset weight value to the features of the second type of samples; for each cluster in the final clustering result, calculate the statistical features of each cluster, and construct the classification result of the data objects based on the statistical features and the final clustering result.

[0126] The processor can call the information and application program stored in the memory through the transmission device to perform the following steps: clean the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain a cleaned multidimensional heterogeneous dataset; standardize the numerical data in the cleaned multidimensional heterogeneous dataset to obtain standardized numerical features; perform one-hot encoding on the classification features in the cleaned multidimensional heterogeneous dataset to obtain encoded classification features; and obtain an object feature set based on the standardized numerical features and encoded classification features.

[0127] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: calculate the weighted Euclidean distance between object samples based on the weighted initial feature vectors to obtain a distance matrix; perform connectivity analysis on the distance matrix using density reachability relations based on preset neighborhood radius and minimum number of points parameters to form multiple density connected regions; identify the density connected regions as initial clusters and mark object samples distributed outside the density connected regions as noise points; remove the noise points and obtain the initial clustering result based on the multiple initial clusters after removing the noise points.

[0128] The processor can invoke information and application programs stored in the memory through the transmission device to perform the following steps: determining a first type of sample features in the object feature set, wherein the first type of sample features includes, but is not limited to, asset features; obtaining the distribution standard deviation of the first type of sample features in the object feature set; calculating a dynamic adjustment coefficient based on the distribution standard deviation, and determining a first preset weight value based on the benchmark weight value and the dynamic adjustment coefficient; assigning the first preset weight value to the first type of sample features, and assigning a preset static weight value to other sample features in the object feature set other than the first type of sample features; obtaining the initial selection feature weights based on the first preset weight value and the static weight value.

[0129] The processor can invoke information and applications stored in the memory via a transmission device to perform the following steps: traverse each cluster in the initial clustering result and extract the initial feature vector of the object samples within the cluster; determine the second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to, life cycle features; assign a second preset weight value to the second type of sample features in the initial feature vector according to preset fine-grained feature weights, and assign a third preset weight value to other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; use the second preset weight value and the third preset weight value to perform weighted calculation on the values ​​of each dimension of the initial feature vector to obtain the weighted fine-tuned feature vector of the object samples.

[0130] The processor can access the information and application programs stored in the memory via the transmission device to execute the following steps: Step 1, for each cluster in the initial clustering result, set the cluster centroid; Step 2, calculate the weighted distance from each object sample to the cluster centroid based on the weighted refinement feature vector, and assign the object samples to the cluster centroid with the smallest weighted distance value based on the weighted distance, to obtain a new clustering result; Step 3, update the cluster centroid of each cluster according to the new clustering result, determine whether the updated cluster centroid meets the iteration condition, and if the updated cluster centroid does not meet the iteration condition, repeat steps 2 to 3 until the updated cluster centroid meets the iteration condition, to obtain the final clustering result.

[0131] The processor can access the information and application stored in the memory via the transmission device to perform the following steps: inputting statistical features into a pre-built large language model, which outputs classification explanation text according to the explanation text template; and integrating the final clustering results and the classification explanation text to generate the classification results of the data objects.

[0132] This invention provides a data object classification scheme. A dynamic feature weighting mechanism is introduced, weighting different business features at different clustering stages. This improves the sample discriminative power in the feature space and enhances the alignment between the final clustering results and actual business logic. Simultaneously, a dual-model coupled clustering algorithm is employed for initial and refined clustering, thereby improving the accuracy of data object classification. This solves the technical problem of low accuracy in related technologies that classify data objects based on statistical indicators and static rule matching.

[0133] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0134] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0135] The invention will now be described in conjunction with another alternative embodiment.

[0136] Example 4

[0137] This invention also provides a computer-readable storage medium. Optionally, in this invention, the computer-readable storage medium can be used to store the program code executed by the data object classification method provided in Embodiment 1.

[0138] Optionally, in this embodiment of the invention, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0139] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of a data object classification method: acquiring a multidimensional heterogeneous dataset of data objects to be classified, and processing the multidimensional heterogeneous dataset to obtain an object feature set; calculating the weighted weights of each sample feature in the object feature set according to the initial selected feature weights to construct a weighted initial selected feature vector; performing density clustering on the object samples in the object feature set based on the weighted initial selected feature vector to output an initial clustering result, wherein the initial selected feature weights are assigned a first preset weight value to the first type of sample features; calculating the weighted weights of the sample features within each cluster of the initial clustering result based on the fine-grained feature weights to construct a weighted fine-grained feature vector, and performing secondary clustering on the object samples based on the initial clustering result and the weighted fine-grained feature vector to obtain a final clustering result, wherein the fine-grained feature weights are assigned a second preset weight value to the second type of sample features; calculating the statistical features of each cluster in the final clustering result, and constructing a classification result of the data objects based on the statistical features and the final clustering result.

[0140] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0141] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0142] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0143] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0144] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0146] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for classifying data objects, characterized in that, include: Obtain a multidimensional heterogeneous dataset of data objects to be classified, and perform data processing on the multidimensional heterogeneous dataset to obtain an object feature set; The features of each sample in the object feature set are weighted according to the initial selected feature weights to construct a weighted initial selected feature vector. Based on the weighted initial selected feature vector, the object samples in the object feature set are density-clustered to output the initial clustering result. The initial selected feature weights are the first preset weight values ​​assigned to the features of the first type of sample. Based on the fine-grained feature weights, the sample features within each cluster of the initial clustering result are weighted and calculated to construct a weighted fine-selection feature vector. Then, based on the initial clustering result, the object samples are clustered a second time in combination with the weighted fine-selection feature vector to obtain the final clustering result. The fine-grained feature weights are assigned a second preset weight value to the features of the second type of sample. For each cluster in the final clustering result, calculate the statistical characteristics of each cluster, and construct the classification result of the data object based on the statistical characteristics and the final clustering result.

2. The method according to claim 1, characterized in that, The steps for processing the multidimensional heterogeneous dataset to obtain the object feature set include: Data cleaning is performed on the multidimensional heterogeneous data in the multidimensional heterogeneous dataset to obtain the cleaned multidimensional heterogeneous dataset. The numerical data in the cleaned multidimensional heterogeneous dataset are standardized to obtain standardized numerical features. One-hot encoding is performed on the classification features in the cleaned multidimensional heterogeneous dataset to obtain encoded classification features. The object feature set is obtained based on the standardized numerical features and the coded classification features.

3. The method according to claim 1, characterized in that, The steps of performing density clustering on object samples in the object feature set based on the weighted initial feature vector and outputting the initial clustering result include: Calculate the weighted Euclidean distance between the object samples based on the weighted preliminary feature vectors to obtain the distance matrix; Based on preset neighborhood radius and minimum number of points parameters, connectivity analysis is performed on the distance matrix using density reachability relationship to form multiple density connected regions; Density connected regions are identified as initial clusters, and object samples distributed outside the density connected regions are marked as noise points. The noise points are removed, and the initial clustering result is obtained based on the multiple initial clusters after the noise points are removed.

4. The method according to claim 1, characterized in that, Before performing weighted calculations on the features of each sample in the object feature set based on the initially selected feature weights, the method further includes: Determine a first type of sample features in the object feature set, wherein the first type of sample features includes, but is not limited to: asset features; Obtain the standard deviation of the distribution of the features of the first type of samples in the object feature set; The dynamic adjustment coefficient is calculated based on the distribution standard deviation, and the first preset weight value is determined based on the benchmark weight value and the dynamic adjustment coefficient. The first preset weight value is assigned to the first type of sample feature, and the preset static weight value is assigned to other sample features in the object feature set other than the first type of sample feature; The initial feature weights are obtained based on the first preset weight value and the static weight value.

5. The method according to claim 1, characterized in that, The steps for constructing a weighted refined feature vector by weighting the sample features within each cluster of the initial clustering result based on fine-grained feature weights include: Traverse each cluster in the initial clustering result and extract the initial feature vector of the object sample within the cluster; Determine a second type of sample features in the object feature set, wherein the second type of sample features includes, but is not limited to: life cycle features; According to the preset fine-grained feature weights, a second preset weight value is assigned to the second type of sample features in the initial feature vector, and a third preset weight value is assigned to the other sample features in the object feature set, wherein the second preset weight value is greater than the third preset weight value; The weighted selection feature vector of the object sample is obtained by weighting the values ​​of each dimension of the initial feature vector using the second preset weight value and the third preset weight value.

6. The method according to claim 1, characterized in that, The steps for performing secondary clustering on the object samples based on the initial clustering results and combining the weighted refined feature vectors to obtain the final clustering results include: Step 1: For each cluster in the initial clustering result, set the cluster centroid; Step 2: Calculate the weighted distance from each object sample to the centroid of the cluster based on the weighted refined feature vector, and assign the object sample to the centroid of the cluster with the smallest weighted distance value based on the weighted distance to obtain a new clustering result; Step 3: Update the centroids of each cluster based on the new clustering results, and determine whether the updated centroids meet the iteration conditions. If the updated centroids do not meet the iteration conditions, repeat steps 2 to 3 until the updated centroids meet the iteration conditions to obtain the final clustering results.

7. The method according to claim 1, characterized in that, The steps for constructing a classification result for the data object based on the statistical features and the final clustering result include: The statistical features are input into a pre-built large language model, which then outputs categorized explanatory texts according to the explanatory text template. The final clustering result and the classification explanation text are integrated to generate the classification result of the data object.

8. A data object classification device, characterized in that, include: The acquisition unit is used to acquire a multidimensional heterogeneous dataset of data objects to be classified, and to process the multidimensional heterogeneous dataset to obtain an object feature set. The first clustering unit is used to perform weighted calculation on the features of each sample in the object feature set according to the initial selected feature weights, construct a weighted initial selected feature vector, perform density clustering on the object samples in the object feature set based on the weighted initial selected feature vectors, and output the initial clustering result, wherein the initial selected feature weights are the first preset weight values ​​assigned to the features of the first type of sample; The second clustering unit is used to perform weighted calculations on the sample features within each cluster of the initial clustering result based on fine-grained feature weights, construct a weighted fine-grained feature vector, and perform secondary clustering on the object samples based on the initial clustering result and the weighted fine-grained feature vector to obtain the final clustering result. The fine-grained feature weights are assigned a second preset weight value to the second type of sample features. The classification unit is used to calculate the statistical characteristics of each cluster in the final clustering result, and to construct the classification result of the data object based on the statistical characteristics and the final clustering result.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the data object classification method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the data object classification method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes a computer program, wherein the computer program, when executed by a processor, implements the data object classification method according to any one of claims 1 to 7.

Images