CN112203311B - Network element abnormity diagnosis method, device, equipment and computer storage medium - Google Patents

Abstract

The embodiment of the invention relates to the technical field of communication, and discloses a method, a device, equipment and a computer storage medium for diagnosing network element abnormity, wherein the method comprises the following steps: acquiring KPI data of a VNF layer target network element in a preset time window, KPI data of a correlation network element correlated with the VNF layer and the target network element and KPI data corresponding to an NFVI layer and the target network element to obtain test data; inputting the test data into a prediction model to obtain predicted KPI data of the target network element within a preset time period; and carrying out abnormity diagnosis on the target network element according to the predicted KPI data of the target network element and the real KPI data of the target network element. By the method, the network element abnormity is diagnosed according to the error between the real KPI data and the predicted KPI data of the network element, so that the abnormity of the network element can be diagnosed before the network element fails, and the network operation and maintenance efficiency is improved.

Description

Network element abnormal diagnosis method, device, equipment and computer storage medium

technical field

Embodiments of the present invention relate to the field of communication technologies, and in particular to a network element abnormality diagnosis method, device, equipment, and computer storage medium.

Background technique

With the advancement of mobile communication technology and the rapid development of data services, the structure of the mobile communication network is becoming more and more complex, and the number of network elements in the mobile communication network is also increasing, which puts forward higher requirements for the maintenance of the communication network.

One of the key tasks of the maintenance work of the communication network is network element fault diagnosis. In the process of implementing the embodiment of the present invention, the inventor found that: the existing network element fault diagnosis method mainly directly sets the threshold value for the network element KPI. The method can only diagnose the faults that have occurred, and cannot detect the abnormality of the network element before the network element fails.

Contents of the invention

In view of the above problems, embodiments of the present invention provide a network element abnormality diagnosis method, device, device, and computer storage medium, which overcome the above problems or at least partially solve the above problems.

According to an aspect of an embodiment of the present invention, a network element abnormality diagnosis method is provided, and the method includes:

Obtain the KPI data of the VNF layer target network element within the preset time window, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element, to obtain test data; inputting the test data into a predictive model to obtain predictive KPI data of the target network element within a preset period of time, wherein the predictive model is trained according to multiple sets of training data, each of the multiple sets of training data All include the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element; according to the predicted KPI of the target network element An error between the data and the real KPI data of the target network element is used to diagnose the abnormality of the target network element.

In an optional manner, performing abnormal diagnosis on the target network element according to the error between the predicted KPI data of the target network element and the real KPI data of the target network element includes: calculating the KPI of the target network element Predict the error between each item of KPI data in the KPI data and the corresponding KPI data in the real KPI data of the target network element; calculate the root mean square error of the error within the preset time period; when there is at least When the root mean square error of one KPI data exceeds its corresponding set threshold, it is determined that the target network element is abnormal.

In an optional manner, calculating the root mean square error of the error within the preset time period includes:

计算所述误差在所述预设时间段内的均方根误差，其中，

和

分别表示预测时间为t+a时刻的第i项预测KPI数据及真实KPI数据，m表示所述预设时间段包含的分钟数。According to the formula

calculating the root mean square error of the error within the preset time period, wherein,

and

Respectively represent the predicted KPI data and real KPI data of the i-th item whose predicted time is t+a, and m represents the number of minutes included in the preset time period.

In an optional manner, after obtaining the test data, the method further includes: performing normalization processing on the test data to obtain standard test data; inputting the test data into the prediction model to obtain the preset time The predicted KPI data of the target network element in the period includes: inputting the standard test data into a forecast model to obtain the predicted KPI data of the target network element in the preset time period.

In an optional manner, before obtaining the test data, the method further includes: constructing an LSTM neural network framework; performing normalization processing on multiple sets of training data acquired to obtain standard training data; The dimensions of each set of training data in the data are converted into three-dimensional training data; the LSTM neural network framework is trained according to the three-dimensional training data to obtain the prediction model.

In an optional manner, constructing the LSTM neural network framework includes: constructing an LSTM neural network framework comprising an output layer, sixteen hidden layers and an output layer, wherein the sixteen hidden layers include eight LSTM layers and eight dropout layers.

In an optional manner, training the LSTM neural network framework according to multiple sets of training data to obtain the prediction model includes: obtaining the weight of the LSTM neural network framework according to the multiple sets of training data; according to the The weight is used to calculate the value of the loss function; the weight is repeatedly updated according to the optimization algorithm until the value of the loss function is the smallest; and the prediction model is obtained according to the weight with the smallest value of the loss function.

According to another aspect of the embodiments of the present invention, a network element abnormality diagnosis device is provided, including: an acquisition module, an input module, and an abnormality diagnosis module, wherein the acquisition module is used to acquire VNF layer target network elements within a preset time window The KPI data of the KPI data, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element, to obtain test data; the input module is used to input the test data into the prediction A model to obtain the predicted KPI data of the target network element within a preset time period, wherein the prediction model is obtained according to multiple sets of training data training, and each set of the multiple sets of training data includes a VNF layer target network element The KPI data of the KPI data, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element; the abnormal diagnosis module is used to predict the KPI data according to the target network element An error with the real KPI data of the target network element is used to diagnose the abnormality of the target network element.

In an optional manner, the abnormality diagnosis module is further used to calculate the difference between each item of KPI data in the predicted KPI data of the target network element and the corresponding item of KPI data in the real KPI data of the target network element. Error; calculate the root mean square error of the error within the preset time period; when the root mean square error of at least one KPI data exceeds its corresponding set threshold, determine that the target network element is abnormal.

计算所述误差在所述预设时间段内的均方根误差，其中，

和

分别表示预测时间为t+a时刻的第i项预测KPI数据及真实KPI数据，m表示所述预设时间段包含的分钟数。In an optional manner, calculating the root mean square error of the error within the preset time period includes: according to the formula

calculating the root mean square error of the error within the preset time period, wherein,

and

Respectively represent the predicted KPI data and real KPI data of the i-th item whose predicted time is t+a, and m represents the number of minutes included in the preset time period.

In an optional manner, the device further includes: a construction module, configured to construct an LSTM neural network framework; a normalization module, configured to normalize multiple sets of acquired training data to obtain standard training data; The conversion module is used to convert the dimensions of each group of training data in the standard training data into three-dimensional training data; the training module is used to train the LSTM neural network framework according to the three-dimensional training data to obtain the described predictive model.

In an optional manner, the building block is further used to construct an LSTM neural network framework comprising an output layer, sixteen hidden layers and an output layer, wherein the sixteen hidden layers include eight LSTM layers and eight A dropout layer.

In an optional manner, the training module is further used to obtain the weight of the LSTM neural network framework according to the multiple sets of training data; calculate the loss function value according to the weight; repeatedly update the weight according to the optimization algorithm, Until the loss function value is minimum; according to the weight with the minimum loss function value, a prediction model is obtained.

In an optional manner, the device further includes: a verification module, configured to verify the prediction model according to multiple sets of verification data. The training module is further used to retrain the forecasting model when the accuracy of multiple sets of verification data is lower than a preset threshold.

According to another aspect of the embodiments of the present invention, a network element abnormality diagnosis device is provided, including: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface communicate through the The bus completes the communication with each other;

The memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform an operation corresponding to the above-mentioned network element abnormality diagnosis method.

According to still another aspect of the embodiments of the present invention, a computer storage medium is provided, and at least one executable instruction is stored in the storage medium, and the executable instruction causes the processor to execute the above-mentioned network element abnormality diagnosis The operation corresponding to the method.

In the embodiment of the present invention, the predicted KPI data of the target network element within the preset time period is obtained by inputting the acquired test data into the prediction model, and the target network element is calculated according to the error between the predicted KPI data and the real KPI data within the preset time period. Abnormal diagnosis is performed, so that the abnormal diagnosis of the network element can be performed before the network element fails, and the efficiency of network operation and maintenance is improved.

The above description is only an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and The advantages can be more obvious and understandable, and the specific embodiments of the present invention are enumerated below.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to designate the same parts. In the attached picture:

FIG. 1 shows a flow chart of a network element abnormality diagnosis method provided by the first embodiment of the present invention;

FIG. 2 shows a flow chart of a network element abnormality diagnosis method provided by the second embodiment of the present invention;

FIG. 3 shows a flow chart of a network element abnormality diagnosis method provided by a third embodiment of the present invention;

FIG. 4 shows a functional block diagram of a device for diagnosing network element abnormality provided by a fourth embodiment of the present invention;

Fig. 5 shows a schematic structural diagram of a network element abnormality diagnosis device provided by a fifth embodiment of the present invention.

detailed description

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 invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

Fig. 1 shows a flow chart of a network element abnormality diagnosis method according to the first embodiment of the present invention. As shown in Fig. 1, the method includes the following steps:

Step 110: Obtain the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element within the preset time window to obtain test data.

Among them, VNF (Virtualized Network Function) is a virtual network function corresponding to the traditional telecommunication service network. The physical network element in the traditional telecommunication service network is mapped to a virtual network element, and the function of the physical network element can be realized through pure software. Virtual network elements run on the network function virtualization infrastructure NFVI (Network Function Virtualization Infrastructure), and NFVI provides computing, storage, network interworking and other virtual resources for virtual network elements. KPI (Key Performance Indication) data is the key performance indicator data when the network element is running, and is used to represent the running state of the network element. The target network element is a certain network element among all network elements at the VNF layer. For the determined target network element of the VNF layer, the associated network element associated with the VNF layer and the target network element and the KPI data corresponding to the NFVI layer and the target network element are determined.

Both the KPI data of the target network element and the KPI data of the associated network element include attributes such as business load, business success rate, throughput, and number of error codes, and the KPI data corresponding to the NFVI layer and the target network element include CPU utilization and I/O rate , memory usage, read and write response time, and other attributes. In the specific implementation process, several attributes can be selected as KPI data. The KPI attribute of the target network element and the KPI attribute category of the associated network element can be the same or different. The number of selected attribute categories can also be the same or different. Considering that the KPI data of a network element changes with time, the KPI data of a recent period of time is obtained through a preset time window to more accurately predict the KPI data of a target network element in a certain period of time in the future .

Step 120: Input the test data into the prediction model to obtain the prediction KPI data of the target network element within the preset time period.

Among them, the predictive model is trained according to multiple sets of training data, and each set of multiple sets of training data includes the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the NFVI The KPI data corresponding to the layer and the target network element.

It is worth noting that when training the forecasting model with multiple sets of training data, the training can be performed on the forecasting model that can predict a certain item of KPI data, or can be trained on the forecasting model that can predict all items of KPI data. When training a prediction model capable of predicting a certain item of KPI data, in order to effectively diagnose abnormalities of network elements, a prediction model needs to be trained for each item of KPI data to predict all items of KPI data.

其中，C_t表示当前时刻神经元的状态，W_i、W_c分别表示输入门的权重，b_i和b_c分别为输入层偏置向量。输出包含两个部分，一部分是全部输出，一部分是用于输入下一LSTM神经元的输出，全部输出的计算公式为：o_t＝σ(W_o[h_t-1,x_t]+b_o)，用于输入下一LSTM神经元的输出的计算公式为：h_t＝o_t*tanh(C_t)，其中，W_o为输出层权重，b_o为输出层偏置向量。在得到全部输出o_t后，对o_t去归一化即可得到输出结果。In some embodiments, the prediction model is obtained by training the LSTM neural network framework according to multiple sets of training data. The LSTM neural network is a neural network with a memory function. The output of the LSTM neuron in each hidden layer will be stored in the cache. When the LSTM neuron has data input next time, the data in the cache will be used as input. At each time point, the output of the LSTM neuron will be put into the cache, and at the next time point, the value in the cache will be overwritten. The LSTM neuron contains three gates, the forget gate, the input gate and the output gate. The forget gate determines the discarded and retained information in the information h _t-1 stored at the last moment, and its calculation formula is: f _t =σ(W _f ·[h _t-1 ,x _t ]+b _f ), where x _t represents the input information, b _f is the input layer bias vector, σ represents the sigmoid function, W _f represents the weight of the forget gate, and the output of the forget gate The result is a number between 0 and 1, where 1 means "keep the information completely" and 0 means "discard the information completely". The input gate determines the information that needs to be updated in the neuron state C _t-1 at the last moment, and its calculation formula is: it ＝σ(W _i ·[h _{t -1} ,x _t ]+ _bi ),

Among them, C _t represents the state of the neuron at the current moment, W _i and W _c represent the weight of the input gate respectively, and b _i and b _c are the bias vectors of the input layer respectively. The output consists of two parts, one part is the whole output, and the other part is the output used to input the next LSTM neuron. The calculation formula of the whole output is: o _t = σ(W _o [h _t-1 ,x _t ]+b _o ), the calculation formula for inputting the output of the next LSTM neuron is: h _t =o _t *tanh(C _t ), where W _o is the weight of the output layer, and b _o is the bias vector of the output layer. After obtaining all the output o _t , the output result can be obtained by denormalizing o _t .

Before performing the test step of step 110, construct the LSTM neural network framework, carry out normalization processing to the multiple sets of training data obtained, obtain standard training data; convert the dimension of each group of training data in the standard training data into three-dimensional training Data; according to the three-dimensional training data, the LSTM neural network framework is trained to obtain the prediction model. Normalization is to scale the test data proportionally so that it falls into a specific interval, so as to eliminate the order of magnitude difference between different types of data. In a specific implementation, the specific interval is generally [0, 1]. In a specific implementation, normalization is performed according to the following formula:

X _std =X _sca ×(X _max -X _min )+X _min

Among them, X _std is a set of standard training data, X is a set of training data, X _max and X _max are the maximum and minimum values of this set of training data, respectively.

Considering that the LSTM neural network requires the input data shape to be a three-dimensional array, therefore, after obtaining the standard test data, it is necessary to perform data transformation on the standard test data to convert it into three-dimensional training data. For example, the standard training data includes the data number of each data and the KPI data corresponding to the data. During the conversion, time features are added to convert the standard test data into three-dimensional training data. It is worth noting that the multiple sets of training data are obtained by dividing the acquired historical KPI data according to a preset time window, which is consistent with the preset time window corresponding to the test data in step 110, and the added time features are preset Set the time corresponding to the time window.

It should be understood that the processing process of the test data is consistent with the processing process of the training data, and will not be repeated here.

In a specific embodiment, the structure of the constructed LSTM neural network is shown in Figure 2, the LSTM neural network includes an input layer, sixteen hidden layers and an output layer, wherein the sixteen hidden layers include eight LSTM layers and eight dropout layers. The input layer is used to input the KPI data of the VNF layer target network element within the preset time window, the KPI data of the associated network elements associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element. The number of neurons contained in the input layer is related to the number of attributes of the KPI data contained in each set of training data. For example, each set of training data contains j KPI attributes of the target network element, which are associated with the target network element The KPI attribute of the associated network element is k, and the KPI attribute of the NFVI layer corresponding to the target network element is l, then, the number of neurons in the input layer should be set to j+k+l. The number of neurons in the output layer is related to the time corresponding to the preset time period. If the preset time period is m minutes, the number of neurons in the corresponding output layer is m, and each neuron is used to output every minute Corresponding KPI data, if the KPI data is a certain item of KPI data, the output result of each neuron is a certain value, if the KPI data is all items of KPI data corresponding to a certain minute, then the output of each neuron The result is an array. The eight LSTM layers and the eight dropout layers are in one-to-one correspondence, and each LSTM layer is connected to a dropout layer, which is used to discard neurons with a preset probability p, thereby avoiding the occurrence of overfitting. In a specific embodiment, the first layer and the second layer LSTM layer are provided with 128 neurons, the third layer and the fourth layer LSTM layer are provided with 64 neurons, and the fifth layer and the sixth layer LSTM layer are provided with 16 neurons. neurons.

During training, the weight of the LSTM neural network framework is obtained according to multiple sets of training data, and the loss function value is calculated according to the weight; the weight is updated repeatedly according to the optimization algorithm until the loss function value is the smallest; according to the weight with the smallest loss function value, the prediction model is obtained . Wherein, the loss function may be manually set by those skilled in the art when implementing the embodiments of the present invention. In a specific implementation manner, the loss function is selected as a Mean Squared Error (MSE) function. The optimization algorithm selects the gradient descent optimization algorithm, which is used to improve the learning speed of traditional gradient descent.

After the training is completed and the prediction model is obtained, the prediction model is verified according to multiple sets of verification data; when the prediction accuracy of multiple sets of verification data is lower than the preset threshold, the prediction model is retrained. Multiple sets of verification data and multiple sets of training data come from the acquired historical KPI data. In the specific implementation process, multiple sets of training data can be divided into training data for training the prediction model and verification data for testing the model in proportion.

Step 130: Perform abnormality diagnosis on the target network element according to the error between the predicted KPI data of the target network element and the real KPI data of the target network element.

In this step, if the error between the predicted KPI data of the target network element and the real KPI data of the target network element meets the condition of an abnormal network element, it can be determined that the target network element is abnormal. In a specific implementation manner, the abnormal condition of the target network element is a preset threshold, and when the error between the predicted KPI data and the real KPI data exceeds the preset threshold, it indicates that the KPI data has signs of degradation and the target network element is abnormal.

In the embodiment of the present invention, the predicted KPI data of the target network element within the preset time period is obtained by inputting the obtained test data into the prediction model, and the target network element is calculated according to the error between the predicted KPI data and the real KPI data within the preset time period. Abnormal diagnosis is performed, so that the abnormal diagnosis of the network element can be performed before the network element fails, and the efficiency of network operation and maintenance is improved.

Fig. 3 shows a flow chart of a method for diagnosing a network element abnormality according to a second embodiment of the present invention. As shown in Fig. 3, the method includes the following steps:

Step 210: Obtain the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element within the preset time window to obtain test data.

Step 220: Input the test data into the prediction model to obtain the prediction KPI data of the target network element within the preset time period.

For specific descriptions of steps 210 to 220, reference may be made to the specific descriptions of steps 110 to 120 in the first embodiment, and details are not repeated here.

Step 230: Calculate the error between each item of KPI data in the predicted KPI data of the target network element and the corresponding item of KPI data in the real KPI data of the target network element.

The embodiment of the present invention is applicable to the case where the predicted KPI data is the KPI data of all items. In this case, it is necessary to calculate the error between each item of predicted KPI data and the corresponding real KPI data, so as to further determine whether the target network element is abnormal.

Step 240: Calculate the root mean square error of the error within a preset time period.

预设时间段内每一分钟对应的KPI真实值为：

则预测值与真实值之间的误差分别为：

均方根误差用于表示预设时间段内的平均误差，均方根误差的计算公式为：

其中，

和

分别表示预测时间为t+a时刻的第i项预测KPI数据及真实KPI数据，m表示预测时间段包含的分钟数。In this step, the KPI prediction values corresponding to each time minute in the preset time period are respectively

The actual KPI value corresponding to each minute within the preset time period is:

Then the error between the predicted value and the actual value is:

The root mean square error is used to represent the average error within a preset time period, and the formula for calculating the root mean square error is:

in,

and

Represents the i-th forecasted KPI data and real KPI data at time t+a respectively, and m represents the number of minutes included in the forecasted time period.

Step 250: When the root mean square error of at least one item of KPI data exceeds its corresponding set threshold, determine that the target network element is abnormal.

In this step, each item of KPI data is preset with a corresponding set threshold value, which is used to indicate the boundary of abnormal KPI data, when the root mean square error of at least one KPI data exceeds its corresponding When setting the threshold, it means that at least one KPI data is abnormal, and the target network element is abnormal.

In the embodiment of the present invention, it is determined whether the target network element is abnormal according to whether the root mean square error of the target network element exceeds its corresponding set threshold within the preset time period, so that the abnormality of the target network element can be predicted before the target network element fails , to facilitate network operation and maintenance.

Fig. 4 shows a schematic structural diagram of an apparatus for diagnosing an abnormality of a network element according to a fourth embodiment of the present invention. As shown in FIG. 4 , the device includes: an acquisition module 410, an input module 420, and an abnormality diagnosis module 430, wherein the acquisition module 410 is used to acquire the KPI data of the VNF layer target network element within a preset time window, the VNF layer and all The KPI data of the associated network element associated with the target network element and the KPI data corresponding to the NFVI layer and the target network element are obtained to obtain test data. The input module 420 is configured to input the test data into the forecasting model to obtain the forecasted KPI data of the target network element within a preset time period, wherein the forecasting model is trained according to multiple sets of training data, and the multiple sets of training Each group of data includes KPI data of the target network element at the VNF layer, KPI data of associated network elements associated with the target network element at the VNF layer, and KPI data corresponding to the target network element at the NFVI layer. The abnormality diagnosis module 430 is configured to perform abnormality diagnosis on the target network element according to the error between the predicted KPI data of the target network element and the real KPI data of the target network element.

In an optional manner, the abnormality diagnosis module 430 is further configured to calculate the difference between each item of KPI data in the predicted KPI data of the target network element and the corresponding item of KPI data in the real KPI data of the target network element. Calculate the root mean square error of the error within the preset time period; when the root mean square error of at least one KPI data exceeds its corresponding set threshold, determine that the target network element is abnormal.

计算所述误差在所述预设时间段内的均方根误差，其中，

和

分别表示预测时间为t+a时刻的第i项预测KPI数据及真实KPI数据，m表示所述预设时间段包含的分钟数。In an optional manner, calculating the root mean square error of the error within the preset time period includes: according to the formula

calculating the root mean square error of the error within the preset time period, wherein,

and

Respectively represent the predicted KPI data and real KPI data of the i-th item whose predicted time is t+a, and m represents the number of minutes included in the preset time period.

In an optional manner, the device further includes: a construction module 450, configured to construct an LSTM neural network framework. A normalization module 460, configured to normalize the multiple sets of acquired training data to obtain standard training data. A conversion module 470, configured to convert the dimensions of each set of training data in the standard training data into three-dimensional training data. The training module 480 is configured to train the LSTM neural network framework according to the three-dimensional training data to obtain the prediction model.

In an optional manner, the construction module 450 is further used to construct an LSTM neural network framework comprising an output layer, sixteen hidden layers and an output layer, wherein the sixteen hidden layers include eight LSTM layers and Eight dropout layers.

In an optional manner, the training module 470 is further used to obtain the weight of the LSTM neural network framework according to the multiple sets of training data; calculate the loss function value according to the weight; and repeatedly update the weight according to the optimization algorithm , until the value of the loss function is minimum; according to the weight with the minimum value of the loss function, a prediction model is obtained.

In an optional manner, the device further includes: a verification module 490, configured to verify the prediction model according to multiple sets of verification data. The training module 470 is further used for retraining the prediction model when the accuracy rate of multiple sets of verification data is lower than a preset threshold.

In the embodiment of the present invention, the acquired test data is input into the prediction model through the input module 420 to obtain the predicted KPI data of the target network element within the preset time period, and the abnormality diagnosis of the target network element is performed through the abnormality diagnosis module 430, so that the network element can be Abnormal diagnosis is performed on network elements before a fault occurs, which improves the efficiency of network operation and maintenance.

An embodiment of the present invention provides a non-volatile computer storage medium, where at least one executable instruction is stored in the computer storage medium, and the computer executable instruction can execute the method corresponding to the network element abnormality diagnosis method in any of the above method embodiments. operate.

An embodiment of the present invention provides a computer program product, the computer program product includes a computer program stored on a computer storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer Execute the operations corresponding to the network element abnormality diagnosis method in any of the foregoing method embodiments.

Fig. 5 shows a schematic structural diagram of a network element abnormality diagnosis device according to a fifth embodiment of the present invention, and the specific embodiment of the present invention does not limit the specific implementation of the device.

As shown in FIG. 5 , the device may include: a processor (processor) 502 , a communication interface (Communications Interface) 504 , a memory (memory) 506 , and a communication bus 508 .

Wherein: the processor 502 , the communication interface 504 , and the memory 506 communicate with each other through the communication bus 508 . The communication interface 504 is configured to communicate with network elements of other devices such as clients or other servers. The processor 502 is configured to execute the program 510, and may specifically execute the above-mentioned relevant steps in the embodiment of the above-mentioned network element abnormality diagnosis method.

Specifically, the program 510 may include program codes including computer operation instructions.

The processor 502 may be a central processing unit CPU, or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention. One or more processors included in the network element abnormality diagnosis device can be the same type of processor, such as one or more CPUs; or different types of processors, such as one or more CPUs and one or more ASICs .

The memory 506 is used for storing the program 510 . The memory 506 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.

The program 510 can specifically be used to make the processor 502 perform the following operations:

Obtain the KPI data of the VNF layer target network element within the preset time window, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element, to obtain test data; inputting the test data into a predictive model to obtain predictive KPI data of the target network element within a preset period of time, wherein the predictive model is trained according to multiple sets of training data, each of the multiple sets of training data All include the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element; according to the predicted KPI of the target network element An error between the data and the real KPI data of the target network element is used to diagnose the abnormality of the target network element.

In an optional manner, the program 510 may be specifically configured to enable the processor 502 to perform the following operations: calculate the difference between each item of KPI data in the predicted KPI data of the target network element and the actual KPI data of the target network element The error between the corresponding KPI data; calculate the root mean square error of the error in the preset time period; when there is at least one KPI data whose root mean square error exceeds its corresponding set threshold, determine the The target NE is abnormal.

In an optional manner, the program 510 may be specifically configured to enable the processor 502 to perform the following operations: triggering the alarm device to issue an alarm of the abnormality of the target network element to the operation and maintenance personnel.

In an optional manner, the program 510 may be specifically configured to cause the processor 502 to perform the following operations: perform normalization processing on the test data to obtain standard test data; input the test data into the prediction model to obtain the predicted Setting the predicted KPI data of the target network element within a time period includes: inputting the standard test data into a prediction model to obtain the predicted KPI data of the target network element within a preset time period.

In an optional manner, the program 510 may be specifically configured to cause the processor 502 to perform the following operations: construct an LSTM neural network framework; train the LSTM neural network framework according to the multiple sets of training data to obtain the prediction Model.

In an optional manner, the program 510 can specifically be used to make the processor 502 perform the following operations: construct an LSTM neural network framework including an output layer, sixteen hidden layers and an output layer, wherein sixteen hidden layers The layers include eight LSTM layers and eight dropout layers.

In an optional manner, the program 510 may be specifically configured to cause the processor 502 to perform the following operations: obtain the weight of the LSTM neural network framework according to the multiple sets of training data; calculate the loss function value according to the weight; The optimization algorithm repeatedly updates the weights until the loss function value is the smallest; according to the weight with the smallest loss function value, a prediction model is obtained.

The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other device. Various generic systems can also be used with the teachings based on this. The structure required to construct such a system is apparent from the above description. Furthermore, embodiments of the present invention are not directed to any particular programming language. It should be understood that various programming languages can be used to implement the content of the present invention described herein, and the above description of specific languages is for disclosing the best mode of the present invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure the understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, in order to streamline the present disclosure and to facilitate an understanding of one or more of the various inventive aspects, various features of the embodiments of the invention are sometimes grouped together into a single implementation examples, figures, or descriptions thereof. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be adaptively changed and arranged in one or more devices different from the embodiment. Modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method or method so disclosed may be used in any combination, except that at least some of such features and/or processes or units are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will understand that although some embodiments herein include some features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention. And form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the execution order.

Claims (9)

1. A network element abnormality diagnosis method, is characterized in that, described method comprises: Obtain the KPI data of the VNF layer target network element within the preset time window, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element, to obtain test data; Inputting the test data into a predictive model to obtain predictive KPI data of the target network element within a preset time period, wherein the predictive model is obtained by training the LSTM neural network framework according to multiple sets of training data, and the multiple sets of training Each group in the data includes the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element; wherein, the The LSTM neural network framework includes an input layer, sixteen hidden layers and an output layer LSTM neural network framework, wherein the sixteen hidden layers include eight LSTM layers and eight dropout layers, eight of the LSTM layers and The eight dropout layers are in one-to-one correspondence, and each of the LSTM layers is connected to one of the dropout layers; An abnormality diagnosis is performed on the target network element according to an error between the predicted KPI data of the target network element and the real KPI data of the target network element. 2. The method according to claim 1, wherein the abnormal diagnosis is performed on the target network element according to the error between the predicted KPI data of the target network element and the real KPI data of the target network element, comprising: calculating an error between each item of KPI data in the predicted KPI data of the target network element and the corresponding KPI data in the real KPI data of the target network element; calculating the root mean square error of the error within the preset time period; When the root mean square error of at least one item of KPI data exceeds its corresponding set threshold, it is determined that the target network element is abnormal. 3. The method according to claim 2, wherein calculating the root mean square error of the error within the preset time period comprises: According to the formula

calculating the root mean square error of the error within the preset time period, wherein,

and

Respectively represent the predicted KPI data and real KPI data of the i-th item whose predicted time is t+a, and m represents the number of minutes included in the preset time period. 4. method according to claim 1, is characterized in that, before obtaining test data, described method also comprises: Construct LSTM neural network framework; Perform normalization processing on multiple sets of acquired training data to obtain standard training data; converting the dimensions of each set of training data in the standard training data into three-dimensional training data; The LSTM neural network framework is trained according to the three-dimensional training data to obtain the prediction model. 5. method according to claim 4, is characterized in that, according to described multiple groups of training data, described LSTM neural network frame is trained, obtains described predictive model, comprising: Obtain the weight of the LSTM neural network framework according to the multiple sets of training data; calculating a loss function value based on the weights; Repeatedly updating the weights according to an optimization algorithm until the loss function value is minimum; A prediction model is obtained according to the weight with the smallest loss function value. 6. method according to claim 5, is characterized in that, after obtaining prediction model, described method also comprises: Verifying the predictive model according to multiple sets of verification data; When the prediction accuracy rate of the plurality of sets of verification data is lower than a preset threshold, the prediction model is retrained. 7. A network element abnormality diagnosis device, characterized in that the device comprises: An acquisition module, configured to acquire the KPI data of the VNF layer target network element within the preset time window, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI data corresponding to the NFVI layer and the target network element , get test data; The input module is used to input the test data into the prediction model to obtain the prediction KPI data of the target network element within the preset time period, wherein the prediction model is obtained by training the LSTM neural network framework according to multiple sets of training data, Each of the multiple sets of training data includes the KPI data of the VNF layer target network element, the KPI data of the associated network element associated with the VNF layer and the target network element, and the KPI corresponding to the NFVI layer and the target network element Data; Wherein, described LSTM neural network framework comprises the LSTM neural network framework of an input layer, sixteen hidden layers and an output layer, wherein, sixteen hidden layers comprise eight LSTM layers and eight dropout layers, eight The LSTM layer is in one-to-one correspondence with the eight dropout layers, and each of the LSTM layers is connected to one of the dropout layers; An abnormality diagnosis module, configured to perform abnormality diagnosis on the target network element according to the error between the predicted KPI data of the target network element and the real KPI data of the target network element. 8. A network element abnormality diagnosis device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface complete mutual communication through the communication bus; The memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform an operation corresponding to the network element abnormality diagnosis method according to any one of claims 1-6. 9. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction causes a processor to execute a network element abnormality diagnosis method corresponding to any one of claims 1-6. operation.

Images