Data classification method and device based on AD-GRU and medium
By introducing an improved reset gate and a new memory candidate state unit into the GRU model, combined with the preset activation function, the problem that traditional GRU algorithms fail to fully consider the text grammar structure, and more accurate semantic analysis and text data classification are achieved.
Patent Information
- Application Number
- CN202311322891.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-12
- Publication Date
- 2025-04-15
AI Technical Summary
Traditional GRU algorithms fail to fully consider the syntactic structure information of the text when processing text data, resulting in difficulty in performing accurate semantic analysis when sufficient context information is lacking.
A data classification method based on the improved gated cycle unit AD-GRU is proposed. By introducing an improved reset gate and a new memory candidate state unit in the AD-GRU classification model, combined with a preset activation function, different frequencies are updated at the level to which the input token belongs.
It improves the accuracy of low-context text data classification, and can more effectively capture the syntax structure information in text data, thereby improving the accuracy of semantic analysis.
Smart Images
Figure CN119830086A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a data classification method, device and readable storage medium based on AD-GRU. Background Art
[0002] GRU (Gated Recurrent Unit) is a variant of RNN (Recurrent Neural Network) used to process sequence data. Similar to traditional RNN and LSTM (Long shortterm memory), the GRU model also has memory capabilities and can capture dependencies and long-term dependencies in sequence data. The GRU model is designed to solve the gradient vanishing and gradient exploding problems in traditional RNN while reducing the complexity of the model.
[0003] Although the traditional GRU algorithm takes into account the contextual information of the text, it does not take into account the grammatical information of the text language, and the grammatical structure information of the text is of great significance to the semantics. For example, "Nanjing Yangtze River Bridge" can be understood as "Nanjing City" + "Yangtze River Bridge", or "Nanjing Mayor" + "River Bridge". Therefore, if there is not enough contextual information, the traditional GRU algorithm is difficult to perform accurate semantic analysis. Summary of the invention
[0004] The technical problem to be solved by the present application is to provide a data classification method, device and readable storage medium based on AD-GRU to solve the problems existing in the prior art in view of the above-mentioned deficiencies in the prior art.
[0005] In a first aspect, the present application provides a data classification method based on AD-GRU, the method comprising:
[0006] Obtaining data to be classified, and extracting sentence vectors corresponding to the data to be classified;
[0007] Inputting the sentence vector into the trained improved gated recurrent unit AD-GRU classification model to obtain confidences corresponding to multiple preset categories;
[0008] Determine the confidence level with the largest median confidence level as the target confidence level, and determine the preset category corresponding to the target confidence level as the target category of the data to be classified;
[0009] Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
[0010] In some embodiments, the preset activation function is:
[0011]
[0012] Among them, x i It represents the input of AD-GRU classification model at time i, τ represents the dimension coefficient, c is a constant, d represents the word segmentation in the sentence vector, and j represents the ordinal number of the word segmentation.
[0013] In some embodiments, the calculation formula of the improved reset gate is:
[0014] r ad =MonoSoftmax(W rad ·[h t-1 ,X t ])
[0015] Among them, r ad Represents the output data of the improved reset gate, W rad represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
[0016] In some embodiments, the output of the improved reset gate is monotonically increasing in the interval [0, 1].
[0017] In some embodiments, the calculation formula of the new memory candidate state unit is:
[0018]
[0019] in, Represents the output data of the new memory candidate state unit, W C and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node, r t Represents the output data of the reset gate, r ad Represents the output data of the improved reset gate.
[0020] In some embodiments, the calculation formula of the hidden state update unit is:
[0021]
[0022] Among them, h t represents the output data of the hidden state update unit, z t represents the output data of the update gate, h t-1 represents the output of the previous node, Represents the output data of the new memory candidate state unit.
[0023] In some embodiments, the calculation formulas of the reset gate and the update gate are:
[0024] z t =σ(W z ·[h t-1 ,X t ])
[0025] r t =σ(W r ·[h t-1 ,X t ])
[0026] Among them, z t Represents the output data of the update gate, r t Represents the output data of the reset gate, W z and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
[0027] In a second aspect, the present application provides a data classification device based on AD-GRU, the device comprising:
[0028] A data acquisition module, which is configured to acquire data to be classified and extract sentence vectors corresponding to the data to be classified;
[0029] A data processing module, which is configured to input the sentence vector into a trained improved gated recurrent unit AD-GRU classification model to obtain confidences corresponding to a plurality of preset categories;
[0030] a category determination module, configured to determine the confidence level with the largest median confidence level as a target confidence level, and to determine the preset category corresponding to the target confidence level as a target category of the data to be classified;
[0031] Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
[0032] In a third aspect, the present application provides a data classification device based on AD-GRU, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the data classification method based on AD-GRU described in the first aspect above.
[0033] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the AD-GRU-based data classification method described in the first aspect is implemented.
[0034] The data classification method, device and readable storage medium based on AD-GRU provided by the present application are specifically, to obtain the data to be classified, and extract the sentence vector corresponding to the data to be classified; the sentence vector is input into the trained improved gated recurrent unit AD-GRU classification model to obtain the confidences corresponding to multiple preset categories; the confidence with the largest median value of the confidence is determined as the target confidence, and the preset category corresponding to the target confidence is determined as the target category of the data to be classified; wherein, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function. The present application adds an improved gate structure on the basis of the existing GRU gate structure, which not only takes into account the context understanding mechanism of the traditional GRU, but also combines the "grammatical structure" mechanism proposed in the present application. The AD-GRU classification model will update the level to which the input token belongs at different frequencies to improve the accuracy of low-context text data classification. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0036] Figure 1 It is a structural diagram of the traditional GRU model;
[0037] Figure 2 A schematic diagram of an AD-GRU classification model provided in an embodiment of the present application;
[0038] Figure 3 A schematic diagram of a data classification method based on AD-GRU provided in an embodiment of the present application;
[0039] Figure 4 A schematic diagram of a data classification device based on AD-GRU provided in an embodiment of the present application;
[0040] Figure 5 Another schematic diagram of the AD-GRU-based data classification device provided in an embodiment of the present application.
[0041] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0042] In order to enable those skilled in the art to better understand the technical solution of the present application, the implementation mode of the present application will be further described in detail below with reference to the accompanying drawings.
[0043] It should be understood that the specific embodiments and drawings described herein are only used to explain the present application, rather than to limit the present application.
[0044] It can be understood that, in the absence of conflict, the various embodiments in the present application and the various features in the embodiments can be combined with each other.
[0045] It can be understood that, for the convenience of description, the drawings of the present application only show the parts related to the present application, while the parts unrelated to the present application are not shown in the drawings.
[0046] It can be understood that each unit and module involved in the embodiments of the present application may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.
[0047] It can be understood that the terms "first", "second", etc. in the embodiments of the present application are used to distinguish different objects, or to distinguish different processing of the same object, rather than to describe a specific order of objects.
[0048] It is understandable that, in the absence of conflict, the functions and steps marked in the flowcharts and block diagrams of the present application may occur in an order different from that marked in the drawings.
[0049] It is understood that the flowcharts and block diagrams of the present application illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present application. Among them, each box in the flowchart or block diagram may represent a unit, module, program segment, code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or by a combination of hardware and computer instructions.
[0050] It can be understood that the units and modules involved in the embodiments of the present application can be implemented by software or hardware, for example, the units and modules can be located in a processor.
[0051] First, the network model involved in this application is explained:
[0052] Figure 1 This is a schematic diagram of the structure of the traditional GRU model, such as Figure 1 As shown, the traditional GRU model includes the following gating units:
[0053] 1. Update Gate (z in the figure) t It is used to control whether the memory of the previous moment needs to be updated. The update gate is based on the current input and the hidden state h of the previous moment. t-1 , generating a weight between 0 and 1, which determines how much memory of the previous moment is retained.
[0054] 2. Reset Gate (r in the figure) t Representation): It is used to control the influence of the memory of the previous moment on the input of the current moment. The reset gate generates a weight between 0 and 1 according to the current input and the hidden state of the previous moment, which determines how much memory of the previous moment is reset.
[0055] 3. New Memory Candidate (in the figure Representation): Generate a new memory candidate state based on the current input and the hidden state at the previous moment. This memory candidate state is the memory that has not passed through the update gate and reset gate, and is used to update the hidden state at the current moment.
[0056] 4. Hidden state update unit (h in the figure) t Representation): According to the update gate, reset gate and new memory candidate state, calculate and update the hidden state at the current moment. This updated hidden state will be used as the output y at the current moment t , and participate in the calculation of the next moment.
[0057] Compared with the LSTM model, the GRU model simplifies the structure, removes the cell state in the LSTM, and retains only one hidden state. This reduces the number of model parameters and computational complexity, while still being able to effectively handle long-term dependencies in the sequence.
[0058] The above-mentioned gating units determine the information flow and memory update mode in the GRU model by adjusting the weight vector. Through these gating units, the GRU model can effectively handle the long-term dependencies and gradient vanishing problems in sequence data, thereby achieving better performance in tasks such as data classification and language modeling.
[0059] In the traditional GRU model, the calculation formula of each gate control unit is:
[0060] z t=σ(W z ·[h t-1 ,X t ]) (1)
[0061] r t =σ(W r ·[h t-1 ,X t ]) (2)
[0062]
[0063]
[0064] According to the above calculation formula, The overall "memory" of the current control is determined by r t Control how much historical information and current information are input t Combined with (Formula 3), h t To output information, z t Controls how much current information is output and how much historical information is output (Formula 4), z t and r t The gate vector for the control information input is trained by the parameter W z and W r They are obtained through linear transformation and activation function, as shown in (Formula 1, 2).
[0065] In the GRU model, a basic unit or element in a text sequence is called a "token". In natural language processing (NLP) tasks, text can be segmented into different tokens, such as words, characters, or subwords. When processing text data, the text is segmented according to certain rules, and each segmented unit is a token. In this way, the text data can be converted into a form that can be understood and processed by the machine, and then tasks such as text classification, sentiment analysis, and machine translation can be performed. In the GRU model, a "token" represents a basic unit in a text sequence, which can be a word, character, or subword, etc., and is used to represent and process text data.
[0066] In the GRU model, tokens are usually used as part of the input sequence, and each token is processed by the GRU unit to capture the contextual information in the sequence. The representation of each token (such as a word embedding vector) is used as the input of the GRU model, processed according to the contextual information, and the corresponding result is output.
[0067] Taking English sentences as an example, a sentence can be segmented into multiple words, each word is a token. For example, the sentence "I love natural language processing" can be segmented into the following sequence of tokens: ["I", "love", "natural", "language", "processing"].
[0068] Taking Chinese sentences as an example, a sentence can be segmented into multiple words, and each word can be segmented into multiple characters. For example, the sentence "Nanjing Yangtze River Bridge" can be segmented into "Nanjing City" + "Yangtze River Bridge", or "Nanjing Mayor" + "River Bridge".
[0069] From human grammar, it can be clearly analyzed that the highest level "sentences" should have more information in grammatical semantics, and their span should be larger, so their "memory" should retain more. Similarly, the lowest level single "word" has a narrow span of information, so its "memory" should be updated faster.
[0070] The memory unit of the traditional GRU updates each token at the same frequency, i.e., r t It is of the same frequency and has no monotonicity, and cannot reflect the hierarchy of the "grammatical structure". This means that if there is not enough contextual information, the traditional GRU algorithm is difficult to perform accurate semantic analysis.
[0071] In view of the technical problems existing in traditional solutions, "memory" means more t Monotonic functions should be added according to the levels. Based on this, the present application proposes an improved gated recurrent unit AD-GRU classification model and a data classification method based on AD-GRU. In the present application, the AD-GRU classification model will update the level of the input token at different frequencies to improve the accuracy of low-context text data classification.
[0072] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.
[0073] The present application provides a data classification method based on AD-GRU. The working process of the method can be implemented by electronic devices, such as computers, handheld smart terminals, etc. For the convenience of explanation, the method execution subject is described as a computer in each embodiment of the present application.
[0074] Figure 2A schematic diagram of the AD-GRU classification model provided in the embodiment of the present application, Figure 3 A schematic diagram of a data classification method based on AD-GRU provided in an embodiment of the present application, such as Figure 2 as well as Figure 3 As shown, the present application provides a data classification method based on AD-GRU, the method comprising:
[0075] S1. Obtain data to be classified, and extract sentence vectors corresponding to the data to be classified;
[0076] The data to be classified may be enterprise data assets, or other data that needs to be classified, which is not specifically limited here.
[0077] In addition, the sentence vector corresponding to the data to be classified can be extracted by the trained field feature extractor.
[0078] S2. Input the sentence vector into the trained improved gated recurrent unit AD-GRU classification model to obtain the confidence corresponding to multiple preset categories;
[0079] Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
[0080] Among them, the preset category can be a level category, or other classification categories, which are not specifically limited here.
[0081] Optionally, the preset activation function is:
[0082]
[0083] Among them, x i It represents the input of AD-GRU classification model at time i, τ represents the dimension coefficient, c is a constant, d represents the word segmentation in the sentence vector, and j represents the ordinal number of the word segmentation.
[0084] In this step, compared with the traditional softmax function, the dimension coefficient τ can play the role of "smoothing" the function, making the effect of grammatical stratification more realistic.
[0085] In addition, the constant c can make the function more stable and prevent overflow.
[0086] After using the above-mentioned activation function proposed in this application, a set of functions with probability distribution properties and monotonically increasing can be obtained, which not only meets the requirements of hierarchical updating in the "grammatical structure", but also has good interpretability and can be back-propagated in deep learning training.
[0087] In some embodiments, the calculation formula of the improved reset gate is:
[0088] r ad =MonoSoftmax(W rad ·[h t-1 ,X t ]) (6)
[0089] Among them, r ad Represents the output data of the improved reset gate, W rad represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
[0090] Optionally, the output of the improved reset gate is monotonically increasing in the interval [0, 1].
[0091] In some embodiments, the calculation formula of the new memory candidate state unit is:
[0092]
[0093] in, Represents the output data of the new memory candidate state unit, W C and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node, r t Represents the output data of the reset gate, r ad Represents the output data of the improved reset gate.
[0094] In some embodiments, the calculation formula of the hidden state update unit is:
[0095]
[0096] Among them, h t represents the output data of the hidden state update unit, z t represents the output data of the update gate, h t-1 represents the output of the previous node, Represents the output data of the new memory candidate state unit.
[0097] In some embodiments, the calculation formulas of the reset gate and the update gate are:
[0098] z t =σ(W z ·[h t-1 ,X t ]) (9)
[0099] r t =σ(W r ·[h t-1 ,X t ]) (10)
[0100] Among them, z t Represents the output data of the update gate, r t Represents the output data of the reset gate, W z and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
[0101] In the above formula (7), Based on the improved door structure in this application ad And traditional door structure t It is found that the context understanding mechanism of the traditional GRU is taken into consideration, and the “grammatical structure” mechanism proposed in this application is combined to improve the accuracy of low-context text data classification.
[0102] S3. Determine the confidence level with the largest median confidence level as the target confidence level, and determine the preset category corresponding to the target confidence level as the target category of the data to be classified;
[0103] Specifically, after obtaining the confidences corresponding to multiple preset categories through the AD-GRU classification model, the confidence with the largest value is determined as the target confidence. Since each confidence corresponds to a preset category, this embodiment determines the preset category corresponding to the target confidence as the target category of the data to be classified.
[0104] This application proposes an AD-GRU classification model and a data classification method based on AD-GRU. An improved gate structure is added on the basis of the existing GRU gate structure. It not only takes into account the context understanding mechanism of the traditional GRU, but also combines the "grammatical structure" mechanism proposed in this application. The AD-GRU classification model will update the level to which the input token belongs at different frequencies to improve the accuracy of low-context text data classification.
[0105] It should be understood that, although the various steps in the flowcharts in the above-described embodiments are sequentially displayed according to the indications of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least a portion of the steps in the figure may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily to be carried out sequentially, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
[0106] Figure 4 A schematic diagram of a data classification device based on AD-GRU provided in an embodiment of the present application, such as Figure 4 As shown, the present application provides a data classification device based on AD-GRU, the device comprising:
[0107] A data acquisition module 11 is configured to acquire data to be classified and extract sentence vectors corresponding to the data to be classified;
[0108] A data processing module 12 is configured to input the sentence vector into a trained improved gated recurrent unit AD-GRU classification model to obtain confidences corresponding to a plurality of preset categories;
[0109] A category determination module 13, which is configured to determine the confidence level with the largest median confidence level as a target confidence level, and determine the preset category corresponding to the target confidence level as the target category of the data to be classified;
[0110] Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
[0111] Regarding the limitation of the data classification device based on AD-GRU, reference may be made to the limitation of the data classification method based on AD-GRU in the above-mentioned embodiments of the present application, which will not be repeated in this embodiment.
[0112] Figure 5 Another schematic diagram of a data classification device based on AD-GRU provided in an embodiment of the present application is as follows Figure 5 As shown, in some embodiments, the present application provides a data classification device based on AD-GRU, including a memory 22 and a processor 21, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the data classification method based on AD-GRU in the above-mentioned embodiments of the present application.
[0113] The memory is connected to the processor, the memory may be a flash memory or a read-only memory or other memory, and the processor may be a central processing unit or a single-chip microcomputer.
[0114] In some embodiments, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the AD-GRU-based data classification method in the above-mentioned embodiments of the present application is implemented.
[0115] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.
[0116] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present application, but the present application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of the present application, and these modifications and improvements are also considered to be within the scope of protection of the present application.
Claims
1. A data classification method based on AD-GRU, characterized in that: The method comprises: Obtaining data to be classified, and extracting sentence vectors corresponding to the data to be classified; Inputting the sentence vector into the trained improved gated recurrent unit AD-GRU classification model to obtain confidences corresponding to multiple preset categories; Determine the confidence level with the largest median confidence level as the target confidence level, and determine the preset category corresponding to the target confidence level as the target category of the data to be classified; Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
2. The data classification method based on AD-GRU according to claim 1 is characterized in that: The preset activation function is: Among them, x i It represents the input of AD-GRU classification model at time i, τ represents the dimension coefficient, c is a constant, d represents the word segmentation in the sentence vector, and j represents the ordinal number of the word segmentation.
3. The data classification method based on AD-GRU according to claim 2 is characterized in that: The calculation formula of the improved reset gate is: r ad =MonoSoftmax(W rad ·[h t-1 ,X t ]) Among them, r ad Represents the output data of the improved reset gate, W rad represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
4. The data classification method based on AD-GRU according to claim 3 is characterized in that: The output of the improved reset gate increases monotonically in the interval [0, 1].
5. The data classification method based on AD-GRU according to claim 3 is characterized in that: The calculation formula of the new memory candidate state unit is: in, Represents the output data of the new memory candidate state unit, W C and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node, r t Represents the output data of the reset gate, r ad Represents the output data of the improved reset gate.
6. The data classification method based on AD-GRU according to claim 5 is characterized in that: The calculation formula of the hidden state update unit is: Among them, h t represents the output data of the hidden state update unit, z t represents the output data of the update gate, h t-1 represents the output of the previous node, Represents the output data of the new memory candidate state unit.
7. The data classification method based on AD-GRU according to claim 6 is characterized in that: The calculation formulas for the reset gate and the update gate are: z t =σ(W z ·[h t-1 ,X t ]) r t =σ(W r ·[h t-1 ,X t ]) Among them, z t Represents the output data of the update gate, r t Represents the output data of the reset gate, W z and W r represents the learning parameters, X t represents the input of the AD-GRU classification model at time t, h t-1 Represents the output of the previous node.
8. A data classification device based on AD-GRU, characterized in that: The device comprises: A data acquisition module, which is configured to acquire data to be classified and extract sentence vectors corresponding to the data to be classified; A data processing module, which is configured to input the sentence vector into a trained improved gated recurrent unit AD-GRU classification model to obtain confidences corresponding to a plurality of preset categories; a category determination module, configured to determine the confidence level with the largest median confidence level as a target confidence level, and to determine the preset category corresponding to the target confidence level as a target category of the data to be classified; Among them, the structure in the AD-GRU classification model includes a reset gate, an update gate, an improved reset gate, a new memory candidate state unit and a hidden state update unit, and the improved reset gate is implemented based on a preset activation function.
9. A data classification device based on AD-GRU, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the AD-GRU-based data classification method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the AD-GRU-based data classification method as described in any one of claims 1 to 7 is implemented.