Business data classification method and device, terminal equipment and storage medium

Abstract

The invention discloses a business data classification method and apparatus, a terminal device and a storage medium. The method comprises the steps of obtaining to-be-classified business data; inputting the to-be-classified business data into a business data classification model, so that the business data classification model outputs a business category of the to-be-classified business data; wherein the service category comprises a primary service category and a secondary service category; the construction of the business data classification model comprises the following steps: obtaining a business data training set; the business data training set comprises a plurality of business data training subsets under a plurality of events; each business data training subset comprises a plurality of business data samples representing that the business category is a main business category and a plurality of business data samples representing that the business category is a secondary business category; constructing an initial business data classification model; and training the initial business data classification model by using the business data training set until the initial business data classification model reaches a preset convergence condition, and generating a business data classification model.

Description

Technical Field

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

[0002] Knowledge base is an intelligent data management system that aims to provide efficient services for tasks such as information extraction, knowledge reasoning and decision support by organizing, storing and managing multi-source heterogeneous data resources. In the process of building a knowledge base, the existing association analysis technology based on knowledge structure captures and represents the relationship between knowledge base information by quantitatively processing the information appearing in various information carriers. However, due to the difference in the amount of different business data in the knowledge base, the problem of classification imbalance is prone to occur when using existing classifiers for association analysis classification. This is because the existing classifiers are trained based on feature extraction of a large number of data samples to train the classification effect. The classifiers obtained by this training method pay less attention to the features of small sample data and have low accuracy in the classification of small sample data. Specifically, for different business data under the same event, the data volume of some businesses is large and the information density of association rules is high, while the data volume of other businesses is relatively small and the information of association rules is sparse. In this case, due to the difference in data volume, the representation ability of small sample data is poor, resulting in a bias effect. When classifying business data, it is easy to ignore the association rules of small sample data, resulting in inaccurate classification results. For example, a business data to be classified contains a large amount of query business data with the keyword "power outage" and a small amount of complaint business data with the keyword "power outage". Existing classifiers usually adopt a training method that focuses on the characteristics of data with a larger sample size during training. During training, the features represented by the data with a large sample size are mainly extracted, and the features of the small sample data are less learned. Therefore, when the classifier is generated, the classifier has a weak recognition ability for the features represented by the small sample data. In actual application, when classifying the business data to be classified, since the sample size of the complaint business is significantly lower than that of the query business, the association rules of the complaint business in the classifier are easily identified as boring rules, thereby ignoring the association rules of the complaint business, resulting in the complaint business being ultimately classified as a result of the query business, making the classification result inaccurate. Summary of the invention

[0003] The embodiments of the present invention provide a business data classification method, apparatus, terminal device and storage medium, which can effectively solve the problem of inaccurate classification of small sample business data when classifying in a knowledge base in the prior art, and improve the accuracy of business data classification.

[0004] An embodiment of the present invention provides a method for classifying business data, including:

[0005] Obtaining business data to be classified;

[0006] Inputting the business data to be classified into a business data classification model so that the business data classification model outputs business categories of the business data to be classified; wherein the business categories include primary business categories and secondary business categories;

[0007] The construction of the business data classification model includes:

[0008] Acquire a business data training set; the business data training set includes a number of business data training subsets under a number of events; each business data training subset includes a number of business data samples representing a business category as a primary business category and a number of business data samples representing a business category as a secondary business category;

[0009] Construct an initial business data classification model;

[0010] The initial business data classification model is trained with the business data training set until the initial business data classification model reaches a preset convergence condition, thereby generating a business data classification model; when training the initial business data classification model, several business data training subsets under each event are measured with a preset interest measurement algorithm to obtain interest measurement results for each business type under each event, and the business category of the business data sample is determined based on the interest measurement results.

[0011] Furthermore, the preset interest measurement algorithm is specifically:

[0012]

[0013] Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of business category C in event A; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

[0014] Furthermore, the initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer;

[0015] The training of the initial business data classification model with the business data training set includes:

[0016] In each iterative training process, the business data sample is input into the input layer, and then transmitted to the feature extraction layer through the input layer. The feature extraction layer extracts the feature data of the business data sample and transmits the feature data to the interest measurement layer.

[0017] The interest measurement layer performs measurement based on the feature data and the preset interest measurement algorithm to obtain the interest measurement result;

[0018] The business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

[0019] Furthermore, in the business data training subset, the total number of business data samples representing business categories as primary business categories is greater than the total number of business data samples representing business categories as secondary business categories.

[0020] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0021] An embodiment of the present invention provides a business data classification device, including: a data acquisition module, a classification module and a classification model construction module;

[0022] The data acquisition module is used to acquire the business data to be classified;

[0023] The classification module is used to input the business data to be classified into the business data classification model, so that the business data classification model outputs the business category of the business data to be classified; wherein the business category includes a primary business category and a secondary business category;

[0024] The classification model construction module is used to obtain a business data training set; the business data training set includes several business data training subsets under several events; each business data training subset includes several business data samples representing business categories as main business categories and several business data samples representing business categories as secondary business categories; construct an initial business data classification model; train the initial business data classification model with the business data training set until the initial business data classification model reaches a preset convergence condition to generate a business data classification model; when training the initial business data classification model, measure several business data training subsets under each event with a preset interest measurement algorithm to obtain interest measurement results for each business type under each event, and determine the business category of the business data sample based on the interest measurement results.

[0025] Furthermore, the preset interest measurement algorithm is specifically:

[0026]

[0027] Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of the business data sample; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

[0028] Furthermore, the initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer;

[0029] The training of the initial business data classification model with the business data training set includes:

[0030] In each iterative training process, the business data sample is input into the input layer, and then transmitted to the feature extraction layer through the input layer. The feature extraction layer extracts the feature data of the business data sample and transmits the feature data to the interest measurement layer.

[0031] The interest measurement layer measures the business data samples according to the feature data and the preset interest measurement algorithm to obtain the interest measurement results;

[0032] The business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

[0033] Furthermore, in the business data training subset, the total number of business data samples representing business categories as primary business categories is greater than the total number of business data samples representing business categories as secondary business categories.

[0034] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a business data classification method described in the above-mentioned embodiment of the invention is implemented.

[0035] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a business data classification method described in the above-mentioned embodiment of the invention.

[0036] The following beneficial effects are achieved by implementing the present invention:

[0037] The present invention provides a business data classification method, device, terminal equipment and storage medium. The method trains an initial business data classification model through a business data training set; wherein the business data training set includes a plurality of business data training subsets under a plurality of events; each business data training subset includes a plurality of business data samples representing a business category as a main business category and a plurality of business data samples representing a business category as a secondary business category; and during the training process, a preset interest measurement algorithm is used to measure the plurality of business data training subsets under each event to obtain interest measurement results of each business type under each event, and the business category of the business data sample is determined according to the interest measurement results; by training the initial business classification with a plurality of business data training subsets under a plurality of events; each business data training subset includes a plurality of business data samples representing a business category as a main business category and a plurality of business data samples representing a business category as a secondary business category, and the preset interest measurement algorithm is combined to perform a comprehensive measurement on the main business category and the secondary business category, so that the generated classification model will not ignore the features of the secondary business category when measuring and classifying according to the features after extracting the features, thereby improving the accuracy of the classification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flowchart of a business data classification method provided by an embodiment of the present invention.

[0039] Figure 2 It is a structural diagram of a business data classification device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] In order to better understand the technical solution of the present invention, it is necessary to explain in advance that the technical solution of the present invention is mainly applied to the classification of business data of the power system knowledge base to improve the accuracy of business data classification. The power system knowledge base is a collection of knowledge and information covering various fields of the power system. In the power system, it mainly includes power business data generated by power plants, substations, power transmission and distribution devices, power networks, and user power equipment. These business data are classified and stored in the power system knowledge base. In the process of constructing the power system knowledge base, due to the difference in the amount of data of different businesses, there is a problem of classification imbalance in data classification, that is, the amount of data of some businesses is large and the information density of association rules is high, while the amount of data of other businesses is relatively small and the information of association rules is sparse. This classification imbalance is a common problem in the prior art. The imbalanced distribution of data of different businesses may cause the characteristics of business types with a small sample size to be ignored when classifying the associated business under the same event, thereby causing inaccurate classification when classifying the features with a small sample size. The technical solution of the present invention is to solve this classification problem.

[0042] like Figure 1 As shown, a business data classification method provided by an embodiment of the present invention includes:

[0043] Step S1: Obtaining business data to be classified;

[0044] Step S2: inputting the business data to be classified into a business data classification model, so that the business data classification model outputs business categories of the business data to be classified; wherein the business categories include primary business categories and secondary business categories;

[0045] For step S1, obtaining the business data to be classified in the power system knowledge base;

[0046] For step S2, the business data to be classified in the power system knowledge base is input into the business data classification model, and the business data classification model outputs the business category of the business data to be classified according to the learned feature data. Exemplary: the business data classification model has learned that in the event of a power outage, the complaint business is a secondary business category, and the query business is a primary business category; the business data to be classified is a text containing "power outage complaint business", which is input into the business data classification model, and the business data classification model determines that the business data to be classified is a secondary business category according to "complaint business" and the learned feature data (that is, in the event of a power outage, the complaint business is a secondary business category, and the query business is a primary business category).

[0047] The reason why the business data classification model can accurately classify secondary business categories mainly depends on the training process of the model.

[0048] In the process of training the business data classification model, it is first necessary to obtain several business data training subsets under each event in the power system knowledge base, each of which includes several business data samples of the main business categories and several business data samples of the secondary business categories. Exemplarily, several business data training subsets under the power outage event in the power system knowledge base are obtained, each of which includes several business samples of query services and several business samples of complaint services; wherein the query service corresponds to the main business category, and the complaint service corresponds to the secondary business category.

[0049] In a preferred embodiment, the total number of service data samples representing service categories as primary service categories in the service data training subset is greater than the total number of service data samples representing service categories as secondary service categories.

[0050] Specifically and exemplarily, in several business data training subsets under a power outage event, the total number of business samples of the query business is greater than, or much greater than, the total number of business samples of the complaint business; that is, under a power outage event, the query business is multi-sample business data, and the complaint business is small-sample business data.

[0051] Construct an initial business data classification model. In a preferred embodiment, the initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer; the initial business data classification model is trained with a business data training set, including: in each iterative training process, a business data sample is input into the input layer, transmitted to the feature extraction layer via the input layer, the feature extraction layer extracts feature data of the business data sample, and transmits the feature data to the interest measurement layer; the interest measurement layer measures according to the feature data and a preset interest measurement algorithm to obtain an interest measurement result; the business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

[0052] Specifically, when training the initial business data classification model, each business data sample is input into the input layer, and the input layer transmits the business data sample to the feature extraction layer, and the feature extraction layer extracts features of the business data sample; for data samples whose business category is the main business category, the features associated with the main business category are extracted, and for data samples whose business category is the secondary category, the features associated with the secondary business category are extracted, and the extracted features are identified, so that the features of the main business category and the features of the secondary business category can be learned during the training process, and then the extracted feature data is transmitted to the interest measurement layer. In the interest measurement layer, an interest measurement algorithm is pre-stored, and the business category to which the business data sample belongs is measured by the preset interest measurement algorithm, and then the interest measurement result is obtained.

[0053] It should be noted that in the training of traditional classifiers, it is usually assumed that the number of data to be classified in the data set to be classified is balanced, and the minority class is not considered. Therefore, the traditional classifier can easily discover the characteristics of the main category, but it is difficult to discover the characteristics of the secondary category. The traditional classifier generates the same specific interest metric for the main category with a large sample size and the secondary category with a small sample size. Usually, the Laplace metric is generated. This metric is only applicable to the case of a balanced data set, not to the case of an unbalanced data set. Then, in the case of a classification requirement for a few sample categories, the present invention proposes an interest metric algorithm based on the combination of the Lift metric and the Laplace metric that can extract the features of both the main business category and the secondary business category.

[0054] Lift is the best metric for predicting secondary business categories, and Laplace is the best metric for predicting primary business categories. However, if only one of them is selected for interest measurement, there will be deviation in the feature extraction of one of the categories. Therefore, combining them has better results. Based on this idea, the present invention proposes a new interest measurement algorithm.

[0055] In a preferred embodiment, the preset interest measurement algorithm is specifically:

[0056]

[0057] Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of business category C in event A; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

[0058] Specifically, using the above example, in a power outage event, the complaint service is a secondary service category, and the query service is a primary service category. Then, in the power outage event, there are two association rules: power outage → query service and power outage → complaint service. In actual service data, the total number of complaint service samples is significantly lower than the total number of query service samples, resulting in the value of the total number of complaint service samples divided by the total number of query service samples being much greater than 1. In the existing Laplace metric, the metric result is extremely sensitive to support, which makes the complaint service rules usually considered to be boring rules. For this reason, the present invention replaces the fixed value part of the numerator of the Laplace metric with a variable To make corrections, the measurement result depends not only on the support of the current business category (i.e., the main business category), but also on the support of other business categories (i.e., the secondary business category). When the data distribution is balanced, the value of the MLaplace metric is approximately equal to the value of the Laplace metric. In fact, if we take a binary imbalanced data set as an example, A→C 1 The MLaplace value is equal to:

[0059]

[0060] Where N(A) represents the total number of samples of event A, that is, the total number of samples of power outage events; C 1 Indicates the main business category, namely, query business; C 2 Indicates secondary business category, i.e. complaint business; N(AC 1 ) indicates that in event A, it belongs to business category C 1 The total number of samples; N(C 1 ) indicates business category C 1 The total sample size; N(C 2 ) indicates that in event A, it belongs to business category C 2 The total number of samples.

[0061] When the data distribution is unbalanced, the MLaplace metric will greatly reduce the penalty for the secondary business category, so this metric is beneficial. In fact, if we take the binary imbalanced data set as an example, A→C 2 The MLaplace value is equal to:

[0062]

[0063] Then, combined with the above preset interest measurement algorithm, when the training subsets are unbalanced under each event in the business data training set, The value of will be much larger than 1, making the value of MLaplace larger, so it can be shown that the MLaplace metric is very effective in extracting C 2 That is, in the interest measurement layer, when measuring the interest of the secondary business category, the characteristics of the secondary business category and the main business category can be well extracted, thereby reducing the problem of ignoring the characteristics of the secondary business category due to the sample size.

[0064] Preferably, each business data sample is sorted based on a multi-criteria decision algorithm such as ELECTRE TRI and interest measurement results, and by setting preference and non-preference thresholds, a partial order relationship of each business data sample in each business category under each event is established to obtain the sorting result of each business data sample, and then the business category of each business data sample is output according to the sorting result.

[0065] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0066] like Figure 2 As shown, an embodiment of the present invention provides a business data classification device, including: a data acquisition module, a classification module and a classification model construction module;

[0067] The data acquisition module is used to acquire the business data to be classified;

[0068] The classification module is used to input the business data to be classified into the business data classification model, so that the business data classification model outputs the business category of the business data to be classified; wherein the business category includes a primary business category and a secondary business category;

[0069] The classification model construction module is used to obtain a business data training set; the business data training set includes several business data training subsets under several events; each business data training subset includes several business data samples representing business categories as main business categories and several business data samples representing business categories as secondary business categories; construct an initial business data classification model; train the initial business data classification model with the business data training set until the initial business data classification model reaches a preset convergence condition to generate a business data classification model; when training the initial business data classification model, measure several business data training subsets under each event with a preset interest measurement algorithm to obtain interest measurement results for each business type under each event, and determine the business category of the business data sample based on the interest measurement results.

[0070] Furthermore, the preset interest measurement algorithm is specifically:

[0071]

[0072] Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of the business data sample; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

[0073] Furthermore, the initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer;

[0074] The training of the initial business data classification model with the business data training set includes:

[0075] In each iterative training process, the business data sample is input into the input layer, and then transmitted to the feature extraction layer through the input layer. The feature extraction layer extracts the feature data of the business data sample and transmits the feature data to the interest measurement layer.

[0076] The interest measurement layer measures the business data samples according to the feature data and the preset interest measurement algorithm to obtain the interest measurement results;

[0077] The business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

[0078] Furthermore, in the business data training subset, the total number of business data samples representing business categories as primary business categories is greater than the total number of business data samples representing business categories as secondary business categories.

[0079] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0080] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0081] Based on the above method item embodiments, the present invention provides corresponding terminal device item embodiments.

[0082] An embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a business data classification method described in any one of the present invention is implemented.

[0083] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0084] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0085] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0086] Based on the above method item embodiments, the present invention provides a corresponding storage medium item embodiment.

[0087] An embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a business data classification method described in any one of the present inventions.

[0088] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0089] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for classifying business data, characterized in that: include: Obtaining business data to be classified; Inputting the business data to be classified into a business data classification model so that the business data classification model outputs business categories of the business data to be classified; wherein the business categories include primary business categories and secondary business categories; The construction of the business data classification model includes: Acquire a business data training set; the business data training set includes a number of business data training subsets under a number of events; each business data training subset includes a number of business data samples representing a business category as a primary business category and a number of business data samples representing a business category as a secondary business category; Construct an initial business data classification model; The initial business data classification model is trained with the business data training set until the initial business data classification model reaches a preset convergence condition, thereby generating a business data classification model; when training the initial business data classification model, several business data training subsets under each event are measured with a preset interest measurement algorithm to obtain interest measurement results for each business type under each event, and the business category of the business data sample is determined based on the interest measurement results.

2. A method for classifying business data according to claim 1, characterized in that: The preset interest measurement algorithm is specifically: Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of business category C in event A; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

3. A method for classifying business data according to claim 1, characterized in that: The initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer; The training of the initial business data classification model with the business data training set includes: In each iterative training process, the business data sample is input into the input layer, and then transmitted to the feature extraction layer through the input layer. The feature extraction layer extracts the feature data of the business data sample and transmits the feature data to the interest measurement layer. The interest measurement layer performs measurement based on the feature data and the preset interest measurement algorithm to obtain the interest measurement result; The business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

4. A business data classification method as claimed in claim 1, characterized in that: The total number of business data samples representing business categories as primary business categories in the business data training subset is greater than the total number of business data samples representing business categories as secondary business categories.

5. A business data classification device, characterized in that: include: Data acquisition module, classification module and classification model building module; The data acquisition module is used to acquire the business data to be classified; The classification module is used to input the business data to be classified into the business data classification model, so that the business data classification model outputs the business category of the business data to be classified; wherein the business category includes a primary business category and a secondary business category; The classification model construction module is used to obtain a business data training set; the business data training set includes several business data training subsets under several events; each business data training subset includes several business data samples representing business categories as main business categories and several business data samples representing business categories as secondary business categories; construct an initial business data classification model; train the initial business data classification model with the business data training set until the initial business data classification model reaches a preset convergence condition to generate a business data classification model; when training the initial business data classification model, measure several business data training subsets under each event with a preset interest measurement algorithm to obtain interest measurement results for each business type under each event, and determine the business category of the business data sample based on the interest measurement results.

6. A service data classification device as claimed in claim 5, characterized in that: The preset interest measurement algorithm is specifically: Where A represents the event to which the business data sample belongs; C represents the business category corresponding to the business data sample; MLaplace(A→C) represents the interest measurement result of the business data sample; N(A) represents the total number of samples of event A; N(AC) represents the number of samples belonging to business category C in event A; Indicates business category N(C) represents the total number of samples of business category C; when C is the main business category, is a secondary business category; when C is a secondary business category, The main business category.

7. A service data classification device as claimed in claim 5, characterized in that: The initial business data classification model includes: an input layer, a feature extraction layer, an interest measurement layer and an output layer; The training of the initial business data classification model with the business data training set includes: In each iterative training process, the business data sample is input into the input layer, and then transmitted to the feature extraction layer through the input layer. The feature extraction layer extracts the feature data of the business data sample and transmits the feature data to the interest measurement layer. The interest measurement layer measures the business data samples according to the feature data and the preset interest measurement algorithm to obtain the interest measurement results; The business data samples in the business data training set are sorted according to the interest measurement result to obtain the sorting result of each business data sample under each event, and the business category of each business data sample is output according to the sorting result.

8. A service data classification device as claimed in claim 5, characterized in that: The total number of business data samples representing business categories as primary business categories in the business data training subset is greater than the total number of business data samples representing business categories as secondary business categories.

9. A terminal device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a business data classification method as described in any one of claims 1 to 4 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a business data classification method as described in any one of claims 1 to 4.

Images