Slice SLA decomposition method, device, computing equipment and computer storage medium

By using a pre-trained slice SLA decomposition device, the SLA requirement indicators and sub-slicing capability parameters are processed based on the codec neural network model, the problem of low SLA decomposition efficiency in the prior art is solved, and more efficient and accurate SLA decomposition is achieved.

CN114051269BActive Publication Date: 2024-11-01CHINA MOBILE GROUP ZHEJIANG +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010721805.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-07-24
Publication Date
2024-11-01
Estimated Expiration
2040-07-24

AI Technical Summary

Technical Problem

In the prior art, SLA decomposition efficiency is low, mainly relying on manual division of labor experience, resulting in low efficiency.

Method used

By obtaining the target SLA requirement indicators and target capability parameters of subslices, processing and encoding, and decomposing them using a pre-trained slice SLA decomposing device. The decomposing device uses multiple sets of training data to learn decomposition rules.

Benefits of technology

The automation of SLA decomposition is realized, which improves the decomposition efficiency and accuracy, reduces manual intervention and improves the decomposition efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114051269B_ABST
    Figure CN114051269B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention relates to the field of communication technology, and discloses a slice SLA decomposition method, device, computing equipment and computer storage medium, the method comprising: obtaining a target SLA demand indicator and a target capability parameter of each sub-slice; processing the target SLA demand indicator and the target capability parameter of each sub-slice to obtain a target input sequence, the target input sequence including target index values ​​corresponding to each target SLA demand indicator and each target capability parameter of each sub-slice; encoding each target index value respectively to obtain a target word vector corresponding to each target index value; inputting the target word vector corresponding to each target index value into a pre-trained slice SLA decomposer to obtain a target decomposition result; the target decomposition result is used to characterize the component of the target SLA demand indicator processed by each sub-slice. In the above manner, the embodiment of the present invention realizes SLA demand indicator decomposition.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of communication technology, and in particular to a slice SLA decomposition method, apparatus, computing device, and computer storage medium. Background Art

[0002] 5G slice SLA decomposition can determine the weight of each SLA demand indicator in each sub-slice based on the SLA demand indicators proposed by users and the actual processing capacity of each sub-slice in the existing network, thereby improving the working efficiency of the existing network.

[0003] In the prior art, SLA decomposition is mainly performed based on the experience of division of labor, which is inefficient. Summary of the invention

[0004] In view of the above problems, embodiments of the present invention provide a slice SLA decomposition method, apparatus, computing device and computer storage medium, which are used to solve the problem of low efficiency of manual SLA decomposition in the prior art.

[0005] According to one aspect of an embodiment of the present invention, a slice SLA decomposition method is provided, the method comprising:

[0006] Obtain the target SLA requirement indicators and target capacity parameters of each sub-slice;

[0007] The target SLA requirement indicator and the target capability parameters of each sub-slice are processed to obtain a target input sequence, wherein the target input sequence includes target index values ​​corresponding to each target SLA requirement indicator and each target capability parameter of each sub-slice;

[0008] Encode each target index value respectively to obtain the target word vector corresponding to each target index value;

[0009] The target word vectors corresponding to the target index values ​​are input into a pre-trained slice SLA decomposer to obtain a target decomposition result; the target decomposition result is used to characterize the components of the target SLA demand indicators processed by each sub-slice; the pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple groups of training data, and each group of training data in the multiple groups of training data includes a target word vector and a target decomposition result corresponding to the target word vector.

[0010] In an optional manner, the processing of the target SLA requirement indicator and the target capability parameters of each sub-slice to obtain a target input sequence includes:

[0011] Performing text cleaning on the target SLA requirement indicator and the target capacity parameters of each sub-slice to obtain a first SLA requirement indicator and a first actual capacity parameter of each sub-slice;

[0012] The first SLA requirement indicator and the first actual capability parameter are serialized to obtain the target input sequence.

[0013] In an optional manner, encoding each target index value separately to obtain a target word vector corresponding to each target index value includes:

[0014] Each target index value is encoded separately through word embedding to obtain the target word vector corresponding to each target index value.

[0015] In an optional manner, before obtaining the target SLA requirement indicator and the target capability parameter of each sub-slice, the method further includes:

[0016] Obtain multiple groups of historical SLA demand indicators, actual capacity parameters of each sub-slice corresponding to each group of historical SLA demand indicators, and decomposition results corresponding to each group of historical SLA demand indicators;

[0017] Processing each set of historical SLA demand indicators and the corresponding actual capacity parameters of each sub-slice to obtain multiple sets of training input sequences; each set of training input sequences includes a set of index values ​​corresponding to the SLA demand indicators and the capacity parameters of the sub-slice;

[0018] Encode the index values ​​corresponding to each group of training input sequences respectively to obtain the corresponding training word vectors;

[0019] Encoding the decomposition results corresponding to the groups of historical SLA demand indicators to obtain reference output sequences corresponding to the groups of historical SLA demand indicators;

[0020] A set of training word vectors and a corresponding set of reference output sequences are used as a set of training data to obtain multiple sets of training data;

[0021] The encoding and decoding neural network model is trained according to the multiple sets of training data to obtain a trained slice SLA decomposer.

[0022] In an optional manner, the step of training the encoding and decoding neural network model according to the multiple sets of training data to obtain a trained slice SLA decomposer includes:

[0023] Input the training word vector into the encoding and decoding neural network model to obtain a training output sequence corresponding to the training word vector, wherein the training output sequence is used to characterize the component of the SLA demand indicator corresponding to each sub-slice;

[0024] Calculating a loss function value according to the training output sequence and the reference output sequence;

[0025] Adjusting the weight of the encoding and decoding neural network model according to the loss function value until a preset number of iterations is reached;

[0026] When the preset number of iterations is reached, the weight that minimizes the loss function value is used as the weight of the encoding and decoding neural network model to obtain a trained slice SLA decomposer.

[0027] In an optional manner, the encoding and decoding neural network model includes an encoder layer, a decoder layer and a fully connected layer;

[0028] The encoder layer includes at least one encoding layer, each encoding layer includes a plurality of first long short-term memory neurons connected in sequence, and the number of the first long short-term memory neurons is the same as the dimension of the training word vector; the encoder layer is used to extract features from the training word vector to obtain multiple groups of feature vectors corresponding to the training word vector;

[0029] The decoder layer includes at least one decoding layer, each decoding layer includes a plurality of second long short-term memory neurons connected in sequence, the number of the second long short-term memory neurons is the same as the number of the first long short-term memory neurons, each second long short-term memory neuron is respectively connected to a first long short-term memory neuron in the last encoding layer of the encoder layer, and the decoder is used to obtain the training output result corresponding to the training word vector according to the multiple groups of feature vectors;

[0030] The fully connected layer includes multiple output neurons, each of which is respectively connected to each second long short-term memory neuron in the last decoding layer of the decoder layer, so as to obtain the training output sequence according to the training output result; the number of the multiple output neurons is the same as the dimension of the output sequence; the output of each output neuron is used to characterize the component of an SLA demand indicator on a sub-slice.

[0031] In an optional manner, adjusting the weight of the encoding and decoding neural network model according to the loss function value until a preset number of iterations is reached includes:

[0032] The weight of the encoding and decoding neural network model is adjusted by the gradient descent method according to the loss function value until a preset number of iterations is reached.

[0033] According to another aspect of an embodiment of the present invention, a slice SLA decomposition device is provided, the device comprising:

[0034] The acquisition module is used to obtain the target SLA requirement indicators and the target capacity parameters of each sub-slice;

[0035] A processing module, configured to process the target SLA requirement indicator and the target capability parameters of each sub-slice to obtain a target input sequence, wherein the target input sequence includes target index values ​​corresponding to each target SLA requirement indicator and each target capability parameter of each sub-slice;

[0036] An encoding module, used to encode each target index value respectively to obtain a target word vector corresponding to each target index value;

[0037] An input module is used to input the target word vectors corresponding to each target index value into a pre-trained slice SLA decomposer to obtain a training decomposition result; the training decomposition result is used to characterize the components of the target SLA demand indicators processed by each sub-slice; the pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple sets of training data, and each set of training data in the multiple sets of training data includes a training word vector and a training decomposition result corresponding to the training word vector.

[0038] According to another aspect of an embodiment of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0039] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned slice SLA decomposition method.

[0040] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a computing device / apparatus to perform operations corresponding to the above-mentioned slice SLA decomposition method.

[0041] The embodiment of the present invention processes the acquired target SLA demand indicators and the target capacity parameters of each sub-slice to obtain a target input sequence, and encodes the target input sequence to obtain a corresponding target word vector. The target word vector can be applied to the slice SLA decomposer for decomposition, so the slice SLA decomposition automation can be realized, thereby improving the efficiency of slice SLA decomposition. The slice SLA decomposer is obtained by training the encoding and decoding neural network model based on a large amount of training data. The decomposition rules of the target word vector of the slice SLA decomposer are obtained by learning a large amount of training data. Therefore, decomposing the SLA demand indicators by the slice SLA decomposer can make the slice SLA decomposition more accurate.

[0042] The above description is only an overview of the technical solution of the embodiment of the present invention. In order to more clearly understand the technical means of the embodiment of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiment of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings:

[0044] Figure 1 A schematic diagram of a process of a slice SLA decomposition method provided by an embodiment of the present invention is shown;

[0045] Figure 2 A schematic diagram of the structure of a slice SLA decomposer provided by an embodiment of the present invention is shown;

[0046] Figure 3 A schematic diagram of the structure of a long short-term memory neuron provided by an embodiment of the present invention is shown;

[0047] Figure 4 A schematic flow chart of a slice SLA decomposition method provided by another embodiment of the present invention is shown;

[0048] Figure 5 A functional block diagram of a slice SLA decomposition device provided by an embodiment of the present invention is shown;

[0049] Figure 6 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0051] In order to facilitate the understanding of the embodiments of the present application, the application scenarios of the embodiments of the present invention are first introduced. The application scenario of the embodiments of the present invention is the decomposition of service level agreement (SLA) indicators in sub-slices in 5G networks and other networks that emerge with the development of networks. After the user proposes an SLA demand indicator, the SLA demand indicator proposed by the user is allocated to each network slice according to the SLA demand indicator proposed by the user and the actual capacity parameters of each sub-slice in the existing network, thereby achieving efficient operation of the network.

[0052] In order to facilitate understanding of the application scenarios of the embodiments of the present invention, various special terms involved in the application scenarios are explained.

[0053] A network slice is an end-to-end logical function and a collection of physical or virtual resources required for it, including access network slices, core network slices, and transport network slices. Network slices can be considered as virtualized private networks in 5G networks, which are used to achieve low-cost and efficient operations. Network slicing technology can achieve logical isolation of communication networks, allowing network elements and functions to be configured and reused in each network slice.

[0054] SLA indicators are used to characterize the requirements that a slice needs to meet, that is, the requirements that the tenant of a slice wants the slice to meet. For example, SLA indicators may include at least one of the following indicators: latency, throughput, packet loss rate, call drop rate, reliability, service scope, user scale, isolation, security, network capability, etc.

[0055] The actual capacity parameters of each sub-slice in the existing network are parameters that characterize the network processing capabilities of each sub-slice in the existing network. For example, the actual capacity parameters of each sub-slice in the existing network include at least one of the following indicators: the current computing resource utilization rate of each sub-slice in the existing network, the current storage resource utilization rate, the current network resource utilization rate, the cell real-time service success rate, the success rate of each domain service, the remaining service capacity, the cell remaining resource block resource utilization rate, etc.

[0056] In order to facilitate understanding of the overall concept of the present invention, specific implementation methods of the present invention are described below.

[0057] Figure 1 FIG. 4 is a flow chart showing a slice SLA decomposition method according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0058] Step 110: Obtain the target SLA requirement index and the target capacity parameters of each sub-slice.

[0059] In this step, the target SLA demand indicator is a demand indicator that needs to be decomposed by SLA. The target capacity parameter of each sub-slice is the network processing capacity parameter of each sub-slice in the communication network where the user sending the target SLA demand indicator is located. In the communication network, the target SLA demand indicator can be obtained from the network element that receives the user demand, and the target capacity parameter of each sub-slice can be obtained from the network slice management function network element. For example, in a 5G communication network, the communication service management function (CSMF) network element is responsible for completing the demand ordering and processing of user business communication services, where the user business communication service demand includes SLA demand indicators. The NSMF network element is used to manage each network slice, and the target capacity parameter of each sub-slice can be obtained from the NSMF network element. After obtaining the target decomposition result corresponding to the target SLA demand indicator, the device / equipment executing the embodiment of the present invention sends the components of the target SLA demand indicator corresponding to each sub-slice to the NSMF according to the target decomposition result, so that the NSMF allocates resources according to the received target SLA demand indicator components that need to be processed by each sub-slice.

[0060] Step 120: Process the target SLA requirement indicator and the target capacity parameters of each sub-slice to obtain a target input sequence.

[0061] In this step, the method of processing the target SLA requirement indicator and the target capacity parameter of each sub-slice can be determined according to the data type of the target SLA requirement indicator and the target capacity parameter of each sub-slice. For example, if the target SLA requirement indicator and the target capacity parameter of each sub-slice are text sequences or numeric sequences, the text sequence or numeric sequence can be directly used as the target input sequence.

[0062] In some embodiments, the target SLA requirement indicator and the target capability parameter of each sub-slice include both numerical data and text data. Then, the target SLA requirement indicator and the target capability parameter of each sub-slice are subjected to text cleaning to obtain the first SLA requirement indicator and the first capability parameter of each sub-slice after text cleaning. Text cleaning includes: removing punctuation marks in the text data, and, for the text data containing English text, converting all English texts into lowercase letters, and performing word segmentation on the Chinese text in the text data, etc. The obtained first SLA requirement indicator and the first capability parameter of each sub-slice are serialized by indexing to obtain a target input sequence. The present invention does not limit the specific implementation method of indexing. For example, in a specific implementation method, the first SLA requirement indicator and the first capability parameter of each sub-slice are serialized by the tokenize module in Python.

[0063] After serializing the target SLA requirement indicators and the target capability parameters of each sub-slice, the target input sequence obtained includes the target index values ​​corresponding to each target SLA requirement indicator and each target capability parameter of each sub-slice. The target index value is a randomly generated value for each target SLA requirement indicator and each target capability parameter of each sub-slice during serialization. The target index value is used to identify the target SLA requirement indicator of a specific field or the target capability parameter of a specific field. For the same target SLA requirement indicator or the same target capability parameter, the corresponding target index value is the same.

[0064] Step 130: Encode each target index value separately to obtain a target word vector corresponding to each target index value.

[0065] In this step, the dimension of the target word vector corresponding to each target index value is the same as the number of neurons that receive the target word vector in the pre-trained slice SLA decomposer. For example, in a specific embodiment, the number of neurons that receive the target word vector is 128, then each target index value is encoded as a 128-dimensional target word vector.

[0066] Step 140: Input the target word vector corresponding to each target index value into the pre-trained slice SLA decomposer to obtain the target decomposition result.

[0067] In this step, the target word vector corresponding to each target index value is input into a pre-trained slice SLA decomposer, and the slice SLA decomposer extracts the features of the target word vector, and decomposes the word vector according to the features to obtain a target decomposition result. The target decomposition result is used to characterize the components of the target SLA demand indicators processed by each sub-slice. The pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple sets of training data. Each set of training data in the multiple sets of training data includes a training word vector and a training decomposition result corresponding to the training word vector. The specific training process is described in the next embodiment, and please refer to the specific description of the next embodiment.

[0068] Figure 2 FIG. 4 shows a schematic diagram of the structure of a slice SLA decomposer according to an embodiment of the present invention. Figure 2As shown, the pre-trained slice SLA decomposer includes an encoder layer, a decoder layer and a fully connected layer. The encoder layer includes at least one encoding layer, each encoding layer includes a plurality of first long short-term memory neurons connected in sequence, and the number of first long short-term memory neurons is the same as the dimension of the target word vector. The target word vector corresponding to each target index value is input into each first long short-term memory neuron according to the dimension, so that each first long short-term memory neuron extracts features of the target word vector corresponding to each target index value. The decoder layer includes at least one decoding layer, each decoding layer includes a plurality of second long short-term memory neurons connected in sequence, and the number of second long short-term memory neurons is the same as the number of first long short-term memory neurons. The decoder is used to establish a mapping relationship between the output result of the first long short-term based neuron and the corresponding SLA decomposition indicator, so as to obtain the slice SLA decomposition result.

[0069] Figure 2 It shows that the encoder layer includes an encoding layer, the decoder layer includes a decoding layer, the encoding layer includes 128 first long short-term memory neurons connected in sequence, for receiving a 128-dimensional target word vector, there are n target word vectors, respectively represented by a1~an, and each target word vector is 128 dimensions. The decoding layer includes 128 second long short-term memory neurons, and the 128 second long short-term memory neurons of the encoding layer are respectively connected to the 128 first long short-term memory neurons of the decoding layer. The fully connected layer includes multiple output neurons, and the number of output neurons is the same as the dimension of the demand indicator after the SLA decomposition is finally obtained. Each output neuron is connected to each second long short-term memory neuron of the decoding layer.

[0070] In a specific implementation, there are 10 target index values, and the target word vector corresponding to each target index value is 128-dimensional. Assume that the encoder layer includes an encoding layer, and the number of first long short-term memory neurons in the encoding layer is 128. Each first long short-term memory neuron receives data of one dimension in the target word vector, that is, each first long short-term memory neuron receives 10 data at a time, and these 10 data correspond to the data of the same dimension in the target word vector corresponding to the 10 target index values. For example, with a1=[a11,a12,a13...a1 128 ]、a2=[a21,a22,a23...a2 128 ]...a10=[a 10 1,a 10 2,a 10 3...a 10 128 ] represent the 128-dimensional target word vectors corresponding to the 10 target index values, and the data received by the first long short-term memory neuron are a11, a21...a10 1. The data received by the second first long short-term memory neuron is a21, a22...a 10 2. At the same time, the output of the first first long short-term memory neuron also serves as the input of the second first long short-term memory neuron.

[0071] Take one of the first long short-term memory neurons as an example. The structure of this neuron is as follows: Figure 3 As shown, the first long short-term memory neuron includes a forget gate, an input gate and an output gate. The output of each first long short-term memory neuron will be stored in the cache. When the neuron has data input at the next time step, the data in the cache will be regarded as part of the input. At each time step, the output of the first long short-term memory neuron will be placed in the cache. At the next time step, the value in the cache will be overwritten. In this embodiment of the present invention, a11, a21...a 10 The time processing interval between 1 data is one time step. The forget gate determines the information h stored in the previous time step. t-1 The information discarded and retained in the calculation formula is: t =σ(W f ·[h t-1 ,x t ]+b f ), where x t Indicates input information, b f is the input layer bias vector, σ represents the sigmoid function, W f Represents the weight of the forget gate. The output of the forget gate is a number between 0 and 1, where 1 means "totally retain the information" and 0 means "totally discard the information". The input gate determines the neuron state C at the previous moment. t-1 The information that needs to be updated is calculated as follows: t =σ(W i ·[h t-1 ,x t ]+b i ), Among them, C t Represents the state of the neuron at the current moment, W i , W c Represent the weight of the input gate, b i and b c are the input gate bias vectors respectively. The output of the input gate consists of two parts, one is the total output, which is used to output to the second long short-term memory neuron of the decoder connected to it; the other is the output of the next first long short-term memory neuron, which is used to input the next first long short-term memory neuron connected to it. The calculation formula for the total output is: t =σ(W o [ht-1 ,x t ]+b o ), the calculation formula for inputting the output of the next first long short-term memory neuron is: t =o t *tanh(C t ), where W o is the output gate weight, b o is the output gate bias vector. t After that, t The output result of the first long short-term memory neuron can be obtained by denormalization, and the output result represents the encoding of the target word vector of this dimension by the encoder.

[0072] The structure of the second long short-term memory neurons included in the decoder layer is the same as that of the first long short-term memory neurons. The calculation process of each second long short-term memory neuron is the same as that of the first long short-term memory neuron. The only difference lies in the weights of the forget gate, input gate and output gate. Please refer to the specific description of the first long short-term based neuron, which will not be repeated here.

[0073] The embodiment of the present invention processes the acquired target SLA demand indicators and the target capacity parameters of each sub-slice to obtain a target input sequence, and encodes the target input sequence to obtain a corresponding target word vector. The target word vector can be applied to the slice SLA decomposer for decomposition, so the slice SLA decomposition automation can be realized, thereby improving the efficiency of slice SLA decomposition. The slice SLA decomposer is obtained by training the encoding and decoding neural network model based on a large amount of training data. The decomposition rules of the target word vector of the slice SLA decomposer are obtained by learning a large amount of training data. Therefore, decomposing the SLA demand indicators by the slice SLA decomposer can make the slice SLA decomposition more accurate.

[0074] Figure 4 FIG. 1 is a flowchart of a slice SLA decomposition method according to another embodiment of the present invention. Compared with the previous embodiment, the present embodiment includes steps 110 to 140 and, before step 110, further includes: Figure 4 The following steps are shown:

[0075] Step 210: Acquire multiple groups of historical SLA demand indicators, actual capacity parameters of each sub-slice corresponding to each group of historical SLA demand indicators, and decomposition results corresponding to each group of historical SLA demand indicators.

[0076] In this step, multiple groups of historical SLA demand indicators and actual capacity parameters of each sub-slice corresponding to each group of historical SLA demand indicators are obtained based on historical user needs. The decomposition results corresponding to each group of historical SLA demand indicators are obtained by manually analyzing each group of historical SLA demand indicators and the actual capacity parameters of each corresponding sub-slice based on expert experience.

[0077] Step 220: Process each group of historical SLA demand indicators and the corresponding actual capacity parameters of each sub-slice to obtain multiple groups of training input sequences.

[0078] In this step, each set of input sequences includes a set of SLA demand indicators and index values ​​corresponding to the capacity parameters of the sub-slices. The processing process of each set of historical SLA demand indicators and the corresponding actual capacity parameters of each sub-slice is the same as Figure 1 The processing process of the target SLA requirement indicator and the target capacity parameters of each sub-slice in step 120 is the same. Please refer to the specific instructions of step 120 and will not be repeated here.

[0079] Step 230: Encode the index values ​​corresponding to each group of training input sequences respectively to obtain corresponding training word vectors.

[0080] In this step, the process of encoding the index values ​​corresponding to each group of training input sequences is the same as the process of encoding the index values ​​corresponding to the target input sequence in step 130. Please refer to the instructions for encoding the target input sequence in step 130. For the sake of simplicity, they are not repeated here.

[0081] Step 240: Encode the decomposition results corresponding to each group of historical SLA demand indicators to obtain a reference output sequence corresponding to each group of historical SLA demand indicators.

[0082] In this step, the decomposition results corresponding to each group of historical SLA demand indicators are used to characterize the components of each SLA demand indicator on each sub-slice. For example, in a specific implementation, a group of historical SLA demand indicators includes delay s1, and the sub-slices include wireless network sub-slices, transmission network sub-slices, and core network sub-slices. The components of delay s1 in the three sub-slices are s11, s12, and s13, respectively, wherein the sum of s11, s12, and s13 is delay s1. The embodiment of the present invention does not limit the specific encoding method of the decomposition results corresponding to each group of historical SLA demand indicators. For example, in a specific implementation, the method for encoding the decomposition results is the same as the method for encoding the input sequence. After encoding, the obtained reference output sequence is a vector, and each three elements in the vector represent the components of an SLA demand indicator on three sub-slices. Alternatively, the encoding is performed according to the proportion of the components of an SLA demand indicator on three sub-slices. For example, if the components of an SLA demand indicator on three sub-slices are all 1 / 3 of the SLA demand indicator, the corresponding encoding is 1 / 3, 1 / 3, and 1 / 3.

[0083] Step 250: Take a set of training word vectors and a corresponding set of reference output sequences as a set of training data to obtain multiple sets of training data.

[0084] Step 260: Train the encoding and decoding neural network model according to the multiple sets of training data to obtain a trained slice SLA decomposer.

[0085] In this step, the encoding and decoding neural network model includes an encoder layer, a decoder layer and a fully connected layer. The encoder layer includes at least one encoding layer, each encoding layer includes a plurality of first long short-term memory neurons connected in sequence, and the number of first long short-term memory neurons is the same as the dimension of the word vector. For example, if the dimension of the word vector is 128 dimensions, the number of first long short-term memory neurons is 128. The first long short-term memory neurons of the encoding layer perform feature extraction on each dimension in the word vector respectively to obtain multiple groups of feature vectors corresponding to multiple groups of word vectors. The decoder layer includes at least one decoding layer, each decoding layer includes a plurality of second long short-term memory neurons connected in sequence, the number of second long short-term memory neurons is the same as the number of first long short-term memory neurons, and each second long short-term memory neuron is respectively connected to a first long short-term memory neuron in the last encoding layer of the encoder layer, and the decoder is used to obtain the output result corresponding to the word vector according to the multiple groups of feature vectors. The fully connected layer includes multiple output neurons, each of which is connected to each second long short-term memory neuron in the last decoding layer of the decoder layer, so as to obtain an output sequence according to the output result. The number of the multiple neurons is the same as the dimension of the output sequence. The output of each output neuron is used to characterize the component of an SLA demand indicator on a sub-slice.

[0086] Each group of word vectors in multiple groups of training data is input into the codec neural network model in turn, and the loss function value is calculated according to the output sequence corresponding to the word vectors of the previous group input and the corresponding reference output sequence, and the weight of the codec neural network model is adjusted according to the loss function value. The latter group of word vectors is input into the codec neural network model after the weight adjustment to obtain the corresponding output sequence. The weight of the codec neural network model is adjusted cyclically, and the word vectors in multiple groups of training data are repeatedly input until the preset number of iterations is reached. When the number of iterations is reached, the weight that minimizes the loss function value is used as the weight of the codec neural network model to obtain a trained slice SLA decomposer. The output result obtained after each group of word vectors is input into the codec neural network model is the component of each SLA demand indicator in each sub-slice. The embodiment of the present invention does not limit the specific form of the loss function. For example, in a specific embodiment, the loss function is a multi-class cross entropy loss function. The loss function value is: Among them, loss represents the loss function value, Represents the i-th element in the reference output sequence, y i represents the i-th element in the output sequence, and m represents the total number of elements in the output sequence.

[0087] In some embodiments, the weights of the codec neural network model are adjusted according to a gradient descent method until a preset number of iterations is reached.

[0088] The embodiment of the present invention obtains multiple groups of training data to train the encoding and decoding neural network model to obtain a trained slice SLA decomposer. The weights in the trained slice SLA decomposer are obtained based on the historical SLA demand indicators in the multiple groups of training data and the actual capacity parameters of each sub-slice and the actual decomposition results. Therefore, the decomposition result of the trained slice SLA decomposer is more accurate.

[0089] Figure 5 FIG. 2 is a schematic diagram showing the structure of a slicing SLA decomposition device according to an embodiment of the present invention. Figure 5As shown, the device includes: an acquisition module 310, a processing module 320, an encoding module 330 and an input module 340. The acquisition module 310 is used to acquire the target SLA demand index and the target capability parameters of each sub-slice. The processing module 320 is used to process the target SLA demand index and the target capability parameters of each sub-slice to obtain a target input sequence, wherein the target input sequence includes target index values ​​corresponding to each target SLA demand index and each target capability parameter of each sub-slice. The encoding module 330 is used to encode each target index value respectively to obtain a target word vector corresponding to each target index value. The input module 340 is used to input the target word vector corresponding to each target index value into a pre-trained slice SLA decomposer to obtain a target decomposition result; the target decomposition result is used to characterize the components of the target SLA demand index processed by each sub-slice; the pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple sets of training data, and each set of training data in the multiple sets of training data includes a training word vector and a training decomposition result corresponding to the training word vector.

[0090] In an optional manner, the processing module 320 is further configured to:

[0091] Performing text cleaning on the target SLA requirement indicator and the target capacity parameters of each sub-slice to obtain a first SLA requirement indicator and a first actual capacity parameter of each sub-slice;

[0092] The first SLA requirement indicator and the first actual capability parameter are serialized to obtain the target input sequence.

[0093] In an optional manner, the encoding module 330 is further configured to:

[0094] Each target index value is encoded separately through word embedding to obtain the target word vector corresponding to each target index value.

[0095] In an optional manner, the device further includes:

[0096] The first acquisition module 350 is used to obtain multiple groups of historical SLA demand indicators, actual capacity parameters of each sub-slice corresponding to each group of historical SLA demand indicators, and decomposition results corresponding to each group of historical SLA demand indicators.

[0097] The first processing module 360 ​​is used to process each group of historical SLA demand indicators and the corresponding actual capacity parameters of each sub-slice to obtain multiple groups of training input sequences; each group of training input sequences includes a group of SLA demand indicators and index values ​​corresponding to the capacity parameters of the sub-slice.

[0098] The first encoding module 370 is used to encode the index values ​​corresponding to each group of training input sequences to obtain corresponding training word vectors.

[0099] The second encoding module 380 is used to encode the decomposition results corresponding to the groups of historical SLA demand indicators to obtain reference output sequences corresponding to the groups of historical SLA demand indicators.

[0100] The determination module 390 is used to use a set of training word vectors and a corresponding set of reference output sequences as a set of training data to obtain multiple sets of training data.

[0101] The training module 300 is used to train the encoding and decoding neural network model according to the multiple sets of training data to obtain a trained slice SLA decomposer.

[0102] In an optional manner, the training module 300 is further used to:

[0103] Inputting the training word vector into the encoding and decoding neural network model to obtain a training output sequence corresponding to the word vector, wherein the training output sequence is used to characterize the component of the SLA demand indicator corresponding to each sub-slice;

[0104] Calculating a loss function value according to the training output sequence and the reference output sequence;

[0105] Adjusting the weight of the encoding and decoding neural network model according to the loss function value until a preset number of iterations is reached;

[0106] When the preset number of iterations is reached, the weight that minimizes the loss function value is used as the weight of the encoding and decoding neural network model to obtain a trained slice SLA decomposer.

[0107] In an optional manner, the encoding and decoding neural network model includes an encoder layer, a decoder layer and a fully connected layer;

[0108] The encoder layer includes at least one encoding layer, each encoding layer includes a plurality of first long short-term memory neurons connected in sequence, and the number of the first long short-term memory neurons is the same as the dimension of the training word vector; the encoder layer is used to perform feature extraction on the training word vector to obtain multiple groups of feature vectors corresponding to the training word vector;

[0109] The decoder layer includes at least one decoding layer, each decoding layer includes a plurality of second long short-term memory neurons connected in sequence, the number of the second long short-term memory neurons is the same as the number of the first long short-term memory neurons, each second long short-term memory neuron is respectively connected to a first long short-term memory neuron in the last encoding layer of the encoder layer, and the decoder is used to obtain the training output result corresponding to the training word vector according to the multiple groups of feature vectors;

[0110] The fully connected layer includes multiple output neurons, each of which is respectively connected to each second long short-term memory neuron in the last decoding layer of the decoder layer, so as to obtain the training output sequence according to the training output result; the number of the multiple output neurons is the same as the dimension of the output sequence; the output of each output neuron is used to characterize the component of an SLA demand indicator on a sub-slice.

[0111] In an optional manner, the training module 300 is further used to:

[0112] The weight of the encoding and decoding neural network model is adjusted by the gradient descent method according to the loss function value until a preset number of iterations is reached.

[0113] The embodiment of the present invention processes the acquired target SLA demand indicators and the target capacity parameters of each sub-slice to obtain a target input sequence, and encodes the target input sequence to obtain a corresponding target word vector, which can be applied to a slice SLA decomposer for decomposition, thereby realizing the automation of slice SLA decomposition. The slice SLA decomposer is obtained by training an encoding and decoding neural network model based on a large amount of training data, and the decomposition rules of the slice SLA decomposer for the target word vector are obtained by learning a large amount of training data. Therefore, decomposing the SLA demand indicators by the slice SLA decomposer can make the slice SLA decomposition more accurate.

[0114] Figure 6 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0115] like Figure 6 As shown, the computing device may include: a processor (processor) 402 , a communications interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0116] The processor 402, the communication interface 404, and the memory 406 communicate with each other via the communication bus 408. The communication interface 404 is used to communicate with other devices such as a client or other server network elements. The processor 402 is used to execute the program 410, which can specifically execute the relevant steps in the above-mentioned embodiment of the method for slicing SLA decomposition.

[0117] Specifically, the program 410 may include program code including computer executable instructions.

[0118] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0119] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0120] Program 410 can be specifically called by processor 402 to cause the computing device to execute Figure 1 Steps 110 to 140 in Figure 4 Steps 210 to 260 in, or, for implementing Figure 5 The functions of modules 310 to 300 in the embodiment.

[0121] An embodiment of the present invention provides a computer-readable storage medium, which stores at least one executable instruction. When the executable instruction is executed on a computing device / apparatus, the computing device / apparatus executes a slice SLA decomposition method in any of the above method embodiments.

[0122] An embodiment of the present invention provides a computer program, which can be called by a processor to enable a computing device to perform operations corresponding to a slice SLA decomposition method in any of the above method embodiments.

[0123] An embodiment of the present invention provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed on a computer, the computer performs operations corresponding to a slice SLA decomposition method in any of the above method embodiments.

[0124] The algorithm or display provided herein is not inherently related to any particular computer, virtual system or other equipment. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious to construct the structure required for this type of system. In addition, the embodiment of the present invention is not directed to any specific programming language yet. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description made to specific languages ​​above is for disclosing the best mode of the present invention.

[0125] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0126] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than those expressly recited in each claim.

[0127] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0128] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A slice SLA decomposition method, It is characterized in that The method comprises: Obtain the target SLA requirement indicators and target capacity parameters of each sub-slice; The target SLA requirement indicator and the target capability parameters of each sub-slice are processed to obtain a target input sequence, wherein the target input sequence includes target index values ​​corresponding to each target SLA requirement indicator and each target capability parameter of each sub-slice; Encode each target index value respectively to obtain the target word vector corresponding to each target index value; The target word vectors corresponding to the target index values ​​are input into a pre-trained slice SLA decomposer to obtain a target decomposition result; the target decomposition result is used to characterize the components of the target SLA demand indicators processed by each sub-slice; the pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple groups of training data, and each group of training data in the multiple groups of training data includes a training word vector and a training decomposition result corresponding to the training word vector.

2. The method according to claim 1, It is characterized in that The processing of the target SLA requirement indicator and the target capability parameters of each sub-slice to obtain a target input sequence includes: Performing text cleaning on the target SLA requirement indicator and the target capacity parameters of each sub-slice to obtain a first SLA requirement indicator and a first actual capacity parameter of each sub-slice; The first SLA requirement indicator and the first actual capability parameter are serialized to obtain the target input sequence.

3. The method according to claim 1, It is characterized in that The encoding of each target index value to obtain a target word vector corresponding to each target index value includes: Each target index value is encoded separately through word embedding to obtain the target word vector corresponding to each target index value.

4. The method according to claim 1, It is characterized in that Before obtaining the target SLA requirement indicator and the target capability parameter of each sub-slice, the method further includes: Obtain multiple groups of historical SLA demand indicators, actual capacity parameters of each sub-slice corresponding to each group of historical SLA demand indicators, and decomposition results corresponding to each group of historical SLA demand indicators; Processing each set of historical SLA demand indicators and the corresponding actual capacity parameters of each sub-slice to obtain multiple sets of training input sequences; each set of training input sequences includes a set of index values ​​corresponding to the SLA demand indicators and the capacity parameters of the sub-slice; Encode the index values ​​corresponding to each group of training input sequences respectively to obtain the corresponding training word vectors; Encoding the decomposition results corresponding to the groups of historical SLA demand indicators to obtain reference output sequences corresponding to the groups of historical SLA demand indicators; A set of training word vectors and a corresponding set of reference output sequences are used as a set of training data to obtain multiple sets of training data; The encoding and decoding neural network model is trained according to the multiple sets of training data to obtain a trained slice SLA decomposer.

5. The method according to claim 4, It is characterized in that The step of training the encoding and decoding neural network model according to the plurality of sets of training data to obtain a trained slice SLA decomposer includes: Input the training word vector into the encoding and decoding neural network model to obtain a training output sequence corresponding to the training word vector, wherein the training output sequence is used to characterize the component of the SLA demand indicator corresponding to each sub-slice; Calculate a loss function value according to the training output sequence and the reference output sequence; Adjusting the weight of the encoding and decoding neural network model according to the loss function value until a preset number of iterations is reached; When the preset number of iterations is reached, the weight that minimizes the loss function value is used as the weight of the encoding and decoding neural network model to obtain a trained slice SLA decomposer.

6. The method according to claim 5, It is characterized in that The encoding and decoding neural network model includes an encoder layer, a decoder layer and a fully connected layer; The encoder layer includes at least one encoding layer, each encoding layer includes a plurality of first long short-term memory neurons connected in sequence, and the number of the first long short-term memory neurons is the same as the dimension of the training word vector; the encoder layer is used to extract features from the training word vector to obtain multiple groups of feature vectors corresponding to the training word vector; The decoder layer includes at least one decoding layer, each decoding layer includes a plurality of second long short-term memory neurons connected in sequence, the number of the second long short-term memory neurons is the same as the number of the first long short-term memory neurons, each second long short-term memory neuron is respectively connected to a first long short-term memory neuron in the last encoding layer of the encoder layer, and the decoder is used to obtain the training output result corresponding to the training word vector according to the multiple groups of feature vectors; The fully connected layer includes multiple output neurons, each of which is respectively connected to each second long short-term memory neuron in the last decoding layer of the decoder layer, so as to obtain the training output sequence according to the training output result; the number of the multiple output neurons is the same as the dimension of the output sequence; the output of each output neuron is used to characterize the component of an SLA demand indicator on a sub-slice.

7. The method according to claim 5, It is characterized in that The step of adjusting the weight of the encoding and decoding neural network model according to the loss function value until a preset number of iterations is reached includes: The weight of the encoding and decoding neural network model is adjusted by the gradient descent method according to the loss function value until a preset number of iterations is reached.

8. A slicing SLA decomposition device, It is characterized in that The device comprises: The acquisition module is used to obtain the target SLA requirement indicators and the target capacity parameters of each sub-slice; A processing module, configured to process the target SLA requirement indicator and the target capability parameters of each sub-slice to obtain a target input sequence, wherein the target input sequence includes target index values ​​corresponding to each target SLA requirement indicator and each target capability parameter of each sub-slice; An encoding module, used to encode each target index value respectively to obtain a target word vector corresponding to each target index value; An input module is used to input the target word vectors corresponding to each target index value into a pre-trained slice SLA decomposer to obtain a target decomposition result; the target decomposition result is used to characterize the components of the target SLA demand indicators processed by each sub-slice; the pre-trained slice SLA decomposer is obtained by training the encoding and decoding neural network model based on multiple sets of training data, and each set of training data in the multiple sets of training data includes a training word vector and a training decomposition result corresponding to the training word vector.

9. A computing device, It is characterized in that include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform an operation corresponding to a slice SLA decomposition method as described in any one of claims 1-7.

10. A computer-readable storage medium, It is characterized in that The storage medium stores at least one executable instruction. When the executable instruction is executed on a computing device / apparatus, the computing device / apparatus performs an operation corresponding to a slice SLA decomposition method as described in any one of claims 1 to 7.