A communication traffic data prediction method and device, electronic equipment and storage medium
By combining shared-weight neural networks and ridge regression algorithms, the problem of gradient vanishing in deep networks is solved, improving the prediction accuracy and stability of communication traffic data.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA TELECOM CORP LTD
- Filing Date
- 2024-06-19
- Publication Date
- 2026-05-19
AI Technical Summary
Existing deep networks suffer from the vanishing gradient problem in predicting communication traffic data, resulting in low prediction accuracy.
By employing a shared-weight neural network to reduce the number of neural network layers, and combining it with the ridge regression algorithm to correct the prediction results, the accuracy and stability of the prediction are improved.
By reducing error accumulation, the accuracy and speed of communication traffic data prediction are improved, and the stability of prediction is enhanced.
Smart Images
Figure CN118612092B_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of Internet technology, and in particular relates to a method, apparatus, electronic device and storage medium for predicting communication traffic data. Background Technology
[0002] With the development of information technology, communication traffic forecasting has become a crucial task for network operators and related industries. Deep learning methods have demonstrated superior performance in communication traffic forecasting, primarily because they can automatically extract features from communication traffic data and model time-series communication traffic data.
[0003] Predicting communication traffic relies on long-term statistical sequences. However, for such sequences, the prediction accuracy of deep networks needs further improvement. In deep networks, gradients propagate through multiple neural network layers, each introducing some error. These errors accumulate, eventually leading to gradient vanishing and consequently, lower accuracy in predicting communication traffic data.
[0004] Therefore, the accuracy of current predictions of communication traffic data needs to be improved. Summary of the Invention
[0005] This application provides a method, apparatus, device, and storage medium for predicting communication traffic data, which can solve the problem of the need to improve the accuracy of current predictions of communication traffic data.
[0006] In a first aspect, embodiments of this application provide a method for predicting communication traffic data, the method comprising:
[0007] Acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period;
[0008] Based on the trained shared weight neural network, the first communication traffic data is predicted to obtain the second communication traffic data for the second time period, which is after the first time period.
[0009] The second communication traffic data is corrected using the ridge regression algorithm to obtain the third communication traffic data.
[0010] Secondly, embodiments of this application provide a communication traffic data prediction device, which includes:
[0011] The acquisition module is used to acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period;
[0012] The prediction module is used to predict the first communication traffic data based on the trained shared weight neural network to obtain the second communication traffic data for the second time period, which is after the first time period.
[0013] The correction module is used to correct the second communication traffic data using the ridge regression algorithm to obtain the third communication traffic data.
[0014] Thirdly, embodiments of this application provide an electronic device, the device including: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the method as described in the first aspect or any possible implementation of the first aspect.
[0015] Fourthly, embodiments of this application provide a readable storage medium storing computer program instructions that, when executed by a processor, implement the method as described in the first aspect or any possible implementation thereof.
[0016] In this embodiment, first communication traffic data is acquired, including communication traffic time series data and communication traffic change characteristic data for a first time period. Then, based on a trained shared weight neural network, the first communication traffic data is predicted to obtain second communication traffic data for a second time period. Since the second time period is after the first time period, the trained shared weight neural network has fewer network layers and a faster fitting speed, which can reduce the accumulation of errors during prediction, improve the prediction accuracy and speed of communication traffic data, and quickly and accurately predict the second communication traffic data for future time periods. Finally, the second communication traffic data is corrected using a ridge regression algorithm to obtain third communication traffic data, which can improve the stability of communication traffic data prediction. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a method for predicting communication traffic data provided in an embodiment of this application;
[0019] Figure 2 This is a schematic diagram of a shared weight neural network provided in an embodiment of this application;
[0020] Figure 3 This is a schematic diagram of another shared weight neural network structure provided in an embodiment of this application;
[0021] Figure 4 This is a schematic diagram of the structure of a communication traffic data prediction device provided in an embodiment of this application;
[0022] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0023] The features and exemplary embodiments of various aspects of this application will now be described in detail. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only configured to explain this application and are not configured to limit this application. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples of this application.
[0024] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.
[0025] The following is a brief introduction to the technical terms used in this application.
[0026] Traffic forecasting refers to the process of predicting and analyzing network data traffic. Traffic forecasting is a crucial task for network operators and related industries. Accurate and efficient traffic forecasting can help these industries better plan network resources and improve network performance and service quality.
[0027] Shared-weight neural networks (SNNs) are a type of network optimization that employs a strategy of optimizing long short-term memory (LSTM) neural networks. By optimizing the number of layers in the LSM network, the errors introduced by the propagation layers are reduced, thus addressing the network optimization problem. This is why they are named shared-weight neural networks.
[0028] Ridge Regression (RR) is an optimized least squares estimation method that addresses the overfitting problem in regression by sacrificing some of the unbiasedness of the least squares method, at the cost of losing some of the mined customer traffic information. This improves the model's generalization performance and stability.
[0029] The communication traffic data prediction method provided in this application embodiment can be applied to at least the following application scenarios, which will be described below.
[0030] Traffic forecasting is a crucial task for network operators and related industries. Accurate and efficient traffic forecasting can help these industries better plan network resources and improve network performance and service quality.
[0031] Machine learning methods are widely used in communication traffic forecasting. By learning from a large amount of historical communication traffic data, a communication traffic forecasting model is built, and then this model is used to predict future data.
[0032] Machine learning methods include regression analysis, decision trees, neural networks, clustering, and dimensionality reduction. By learning from a large amount of communication traffic data, a multi-layered neural network model is built, and then this model is used to predict future communication traffic data.
[0033] Commonly used deep learning algorithms include recurrent neural networks, long short-term memory neural networks, and convolutional neural networks. Deep learning methods perform exceptionally well in communication traffic prediction, primarily because they can automatically extract features from communication traffic data and model time-series communication traffic data.
[0034] Recurrent neural networks suffer from the vanishing gradient problem. Long Short-Term Memory (LSTM) neural networks have alleviated this problem to some extent, but have not ultimately solved it. For longer sequences, their prediction accuracy needs further improvement, and the prediction of communication traffic depends on long-term statistical sequences.
[0035] In deep networks, gradients need to propagate through multiple neural network layers, and each layer introduces a certain amount of error. These errors accumulate and eventually lead to gradient vanishing. Optimized long short-term memory neural networks are used to mitigate the gradient vanishing problem.
[0036] Therefore, the accuracy of current predictions of communication traffic data needs to be improved.
[0037] Figure 1 This is a flowchart of a method for predicting communication traffic data provided in an embodiment of this application.
[0038] like Figure 1 As shown, the method for predicting communication traffic data may include steps 110-130. This method is applied to a communication traffic data prediction device, as detailed below:
[0039] Step 110: Obtain first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period;
[0040] Step 120: Based on the trained shared weight neural network, predict the first communication traffic data to obtain the second communication traffic data for the second time period, which is after the first time period;
[0041] This paper studies communication traffic prediction, analyzes communication traffic prediction based on a shared-weight neural network, and makes timely improvements to the neural network to make it suitable for accurate communication traffic prediction. Experiments show that the improved neural network achieves good communication traffic prediction results, improves the speed at which the neural network fits communication traffic data to a certain extent, and further maintains the accuracy and precision of traffic prediction while maintaining training speed.
[0042] Step 130: Correct the second communication traffic data using the ridge regression algorithm to obtain the third communication traffic data.
[0043] The ridge regression algorithm is used to perform regression analysis on the predicted communication traffic for future periods based on a shared weight neural network. Ridge regression includes a regression residual process, through which the coefficients of the regression amplitude are adaptively adjusted. Principal component analysis (PCA) is performed on the factors affecting customer traffic, and the predicted communication traffic for future periods is analyzed. The analyzed predicted communication traffic is then output, thereby improving the stability of communication traffic prediction.
[0044] In this embodiment, first communication traffic data is acquired, including communication traffic time series data and communication traffic change characteristic data for a first time period. Then, based on a trained shared weight neural network, the first communication traffic data is predicted to obtain second communication traffic data for a second time period. Since the second time period is after the first time period, the trained shared weight neural network has fewer network layers and a faster fitting speed, which can reduce the accumulation of errors during prediction, improve the prediction accuracy and speed of communication traffic data, and quickly and accurately predict the second communication traffic data for future time periods. Finally, the second communication traffic data is corrected using a ridge regression algorithm to obtain third communication traffic data, which can improve the stability of communication traffic data prediction.
[0045] The following describes the contents of steps 110-130 respectively:
[0046] Step 110 is involved.
[0047] Acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period.
[0048] Acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period.
[0049] First, the original communication traffic time series data for the first time period is obtained. The original communication traffic time series data is then cleaned by supplementing missing values, removing outliers, and deleting duplicate values.
[0050] Then, the original communication traffic time series data is arranged into uniform time intervals. The original communication traffic time series data records are recorded in monthly units and arranged in chronological order to facilitate subsequent data processing.
[0051] The original communication traffic time series data records are standardized. After standardization, the value of each element in the original communication traffic time series data records is mapped to the [0,1] interval according to the normalization method.
[0052] Therefore, after normalization, the original communication traffic time series data is processed to a range suitable for the input of the shared weight neural network, thus obtaining the first communication traffic data.
[0053] In one possible embodiment, step 110 includes:
[0054] Obtain the time series data of communication traffic for the first time period;
[0055] Extract the communication traffic change feature data from the communication traffic time series data;
[0056] The first communication traffic data is obtained by combining the communication traffic time series data and the communication traffic change characteristic data.
[0057] Communication traffic change characteristics data, such as communication traffic change characteristics data associated with seasons, years, and months.
[0058] The communication traffic change feature data is extracted from the communication traffic time series data, and the extracted communication traffic change feature data is combined with the communication traffic time series data as input data for predicting communication traffic in future periods based on a shared weight neural network.
[0059] In one possible embodiment, the step of extracting the communication traffic change feature data from the communication traffic time series data mentioned above may specifically include the following steps:
[0060] Determine candidate change features associated with the time series data of the communication traffic;
[0061] Based on the communication traffic time series data, determine the combined solutions of different combinations of candidate change features;
[0062] Based on each of the combined solutions, the target change feature is determined from the candidate change features;
[0063] Extract the change feature data corresponding to the target change feature from the communication traffic time series data.
[0064] Characteristics of target changes, such as time factors like season, year, and month.
[0065] During the feature extraction process, the current communication traffic time series data is input, and the candidate features of the communication traffic time series data are listed. The data is set to two states: 0 or 1. 0 represents discarding the current feature, and 1 represents selecting the current feature. All solutions of feature combinations are listed, that is, the combined solutions of different combinations of candidate changing features are determined.
[0066] Solving the optimal feature combination problem is equivalent to a 0-1 programming problem. A set of target change features is iteratively solved using a genetic algorithm, i.e., a set of target change features is obtained by screening. The change feature data corresponding to the target change features is extracted from the communication traffic time series data. The change feature data corresponding to the target change features is combined with the communication traffic time series data as input data for predicting communication traffic in future periods based on a shared weight neural network.
[0067] Step 120 is involved.
[0068] Based on the trained shared weight neural network, the first communication traffic data is predicted to obtain the second communication traffic data for the second time period, which is after the first time period.
[0069] In neural networks, weight sharing is an optimization technique commonly used to reduce the number of model parameters and improve the model's generalization ability. It is widely used in convolutional neural networks (CNNs), especially when processing images and other similarly structured data.
[0070] The concept of weight sharing involves using the same parameter weight values in different parts of a network. This means using the same weights at multiple locations or layers within a neural network, rather than learning different weights independently for each location or layer. This method of sharing weights can effectively capture local features of the input data, thereby reducing the number of parameters that need to be learned.
[0071] Among them, such as Figure 2 As shown, X1, X2, and X3 share the following... Figure 2 The weight values of the shared weight neural network are shown.
[0072] In one possible embodiment, the following steps may be included before step 120:
[0073] While maintaining the neural network unit structure of the initial neural network, the forget gate and the output gate in the initial neural network are merged to obtain the merged gate;
[0074] The merging gate and the input gate are coupled to obtain a shared weight neural network, wherein the number of network layers in the shared weight neural network is less than the number of network layers in the initial neural network.
[0075] Train the shared weight neural network to obtain the trained shared weight neural network.
[0076] Long Short-Term Memory (LSTM) networks introduce memory cells, or simply cells. To control these memory cells, several gates are needed. One gate is used to output entries from the cell; this is called the output gate. Another gate determines when data is read into the cell; this is called the input gate. A mechanism is also needed to reset the cell's contents, managed by the forget gate. This design is motivated by the same principle as gated recurrent units, which use a dedicated mechanism to determine when to remember or ignore input from the hidden state.
[0077] like Figure 2 As shown, a shared weight neural network includes an input layer, a hidden layer, and an output layer.
[0078] x t s t a t This is the input to the input layer. C t C represents the current cell state. t-1 Indicates the state of the previous unit. h t h t-1 `in` represents the current and previous hidden layer states, respectively. tAs intermediate variables, tanh(x) and σ(x) are activation functions. This is the symbol for matrix element multiplication.
[0079] At time t, calculate the shared gate. The calculation formula is as follows:
[0080]
[0081] Next, update cell state C t The calculation formula is as follows:
[0082] C t =s t *C t-1 +(1-s t )*a t-1
[0083] Then, calculate the hidden layer output h. t The calculation formula is as follows:
[0084] h t =s t *tanh(C t )
[0085] Finally, calculate the output value yt of the output layer. The calculation formula is as follows:
[0086]
[0087] In the embodiments of this application, while maintaining the neural network unit structure of the initial neural network, the forget gate and the output gate in the initial neural network are merged to obtain a merged gate, which can reduce the number of neural network layers of the long short-term memory neural network. By coupling the merged gate and the input gate, a shared weight neural network is obtained, which can further reduce the number of neural network layers of the long short-term memory neural network.
[0088] The optimized Long Short-Term Memory (LSTM) neural network aims to reduce the propagation error between layers by reducing the number of neural network layers.
[0089] The neural network with reduced layers is named the shared-weight neural network. In the shared-weight neural network, while maintaining the structure of the neural network units, the forget gate and the output gate in the neural network are merged, and the merged gate is coupled with the input gate to reduce the number of layers in the deep network to be optimized, reduce the propagation error between layers, and improve prediction accuracy.
[0090] In one possible embodiment, the following steps may be included before step 120:
[0091] Acquire first historical communication traffic data and second historical communication traffic data, wherein the time period of the second historical communication traffic data is after the time period of the first historical communication traffic data;
[0092] Based on the first historical communication traffic data and the second historical communication traffic data, a shared weight neural network is trained until the shared weight neural network meets the training stopping condition, thus obtaining a trained shared weight neural network.
[0093] The time period of the second historical communication traffic data is after the time period of the first historical communication traffic data. For example, the second historical communication traffic data is the communication traffic data in December 2023, and the first historical communication traffic data is the communication traffic data in November 2023.
[0094] The first historical communication traffic data is input into the shared weight neural network, and the predicted second historical communication traffic data is output. The shared weight neural network is trained based on the second historical communication traffic data and the predicted second historical communication traffic data until the shared weight neural network meets the training stopping condition, and the trained shared weight neural network is obtained.
[0095] Step 130 is involved.
[0096] The second communication traffic data is corrected using the ridge regression algorithm to obtain the third communication traffic data.
[0097] The ridge regression algorithm is used to perform regression analysis on the predicted values of communication traffic for future periods based on a shared weight neural network. Ridge regression includes a regression residual process, which adaptively adjusts the coefficients of the regression amplitude and outputs third-party communication traffic data, thereby improving the stability of communication traffic data prediction.
[0098] like Figure 2 As shown, the input data is: Customer monthly traffic time series training dataset X. i ;
[0099] Output data: Monthly traffic forecast sequence Y i ;
[0100] l represents the length of the input data;
[0101] R i Predict the residual sequence.
[0102] Exemplary, the algorithm steps of an embodiment of this application include:
[0103] 1. Normalize the first communication traffic data using standard normalization.
[0104] 2: Define the `series_to_supervised` function to convert the time series into a supervised learning problem;
[0105] 3: Set the number of steps;
[0106] 4: Divide the dataset into training and testing sets;
[0107] 5: Construct the training dataset for the input of the shared-weight neural network model;
[0108] 6: Initialize the shared-weight neural network model;
[0109] The number of neurons is set to 16, the number of iterations is 100, and the Adam iterator is used;
[0110] 7: for i = 1 to;
[0111] 8: LSTM modeling;
[0112] 9: i++;
[0113] 10: end for;
[0114] 11: Output the predicted flow sequence values and predicted residual sequence from the shared weight neural network;
[0115] 12: Reconstruct the inputs of the RR model;
[0116] The inputs in this case are the predicted sequence values and the predicted residual sequence values of the shared weight neural network.
[0117] 13: for j = 1 to 1;
[0118] 14: RR modeling;
[0119] 15: Ridge regression is used to analyze and predict residual sequence values;
[0120] 16: j++;
[0121] 17: end for;
[0122] 18: Introduce adjustable adaptive parameters;
[0123] 19: Introduce PCA principal component analysis;
[0124] 18: Output the RR communication traffic prediction sequence;
[0125] 19: Anti-normalization;
[0126] 20: Output the final prediction result.
[0127] Through the embodiments of this application, the customer's monthly communication traffic data is used as an experimental dataset to predict the customer's communication traffic for the following month.
[0128] The experimental dataset is shown in Table 1:
[0129] Table 1
[0130]
[0131] Based on the principle of control in the experiment, the communication traffic prediction method based on shared weight neural network is compared with the communication traffic prediction method based on long short-term memory neural network (LSTM) and the communication traffic prediction method based on ridge regression (RR).
[0132] The prediction results for Client A are shown in Table 2, and the prediction results for Client B are shown in Table 3.
[0133] Table 2
[0134]
[0135]
[0136] Table 3
[0137]
[0138] Evaluation is conducted using evaluation indicators, including RMSE (Root Mean Square Error) and R. 2 (Determination coefficient), the comparison results of various flow forecasting methods are as follows.
[0139] The root mean square error (RMSE) measures the deviation between the predicted value and the true value. The smaller the RMSE, the smaller the prediction deviation of the algorithm. The coefficient of determination is a statistic that measures the goodness of fit. The larger the coefficient, the better the goodness of fit between the predicted data and the actual data, indicating a better prediction effect.
[0140] The evaluation indicators in Table 2 are shown below:
[0141]
[0142]
[0143] The evaluation index table uses the root mean square error (RMSE) and coefficient of determination between the predicted and actual values as the evaluation criteria for the algorithm's performance. The smaller the RMSE, the smaller the error between the algorithm's predicted and actual values, indicating higher accuracy. The communication traffic prediction method based on shared weight neural networks is compared with the long short-term memory neural network (LSTM) communication traffic prediction method and the ridge regression (RR) communication traffic prediction method.
[0144] The communication traffic prediction method based on shared weight neural networks has a smaller root mean square error, indicating that the method has a smaller deviation in predicting communication traffic. At the same time, the communication traffic prediction method based on shared weight neural networks has a higher coefficient of determination, indicating that the predicted traffic value fits the actual value better and the accuracy of communication traffic prediction is higher.
[0145] In one possible embodiment, step 130 may specifically include the following steps:
[0146] The residual sequence is determined based on the second communication traffic data and the actual communication traffic data for the second time period;
[0147] The residual sequence is regressed using the ridge regression algorithm to obtain the regression amplitude coefficients;
[0148] The second communication traffic data is corrected using the regression amplitude coefficient to obtain the third communication traffic data.
[0149] The determination of the residual sequence based on the second communication traffic data and the actual communication traffic data of the second time period refers to the result obtained by comparing the predicted value of the communication traffic in the future time period based on the shared weight neural network with the actual value.
[0150] The residual sequence is regressed using the ridge regression algorithm to obtain the regression amplitude coefficient. Here, the gap between the learned second communication traffic data and the actual second communication traffic data is determined. An adjustable adaptive parameter λ is introduced, and the coefficient of the regression amplitude is adaptively adjusted using the adaptive parameter λ. The adjustable adaptive parameter aims to minimize the above gap, and the current adaptive parameter λ is adaptively adjusted according to the regression function.
[0151] The coefficients of the regression amplitude are adaptively adjusted using an adaptive parameter λ. If the difference between the currently learned second communication traffic data and the actual second communication traffic data is extremely small, then λ is considered to be the optimal parameter at this time. The regression amplitude coefficients learned at this time are then used to correct the communication traffic in future periods.
[0152] Therefore, by correcting the second communication traffic data using the regression amplitude coefficient, the third communication traffic data can be obtained, which can further improve the stability of streaming communication traffic prediction.
[0153] In one possible embodiment, after step 130, the following steps may also be included:
[0154] The spatiotemporal factors in the communication traffic time series data are identified as the influencing factors in the principal component analysis of the communication traffic data. The spatiotemporal factors include: seasonal factors, monthly factors, location factors, and regional factors.
[0155] By analyzing the influencing factors through principal component analysis, principal component analysis is performed on the third communication traffic data to obtain the analyzed third communication traffic data.
[0156] In the time series data of customer communication traffic, time factors, such as seasonal and monthly factors, and spatial factors, such as location and region factors, are considered as influencing factors in the PCA principal component analysis of customer traffic.
[0157] Principal component analysis (PCA) is performed on the factors influencing customer traffic, and the third-party communication traffic data for future periods is analyzed. The analyzed third-party communication traffic data is then output, thereby improving the stability of communication traffic prediction.
[0158] In this embodiment, first communication traffic data is acquired, including communication traffic time series data and communication traffic change characteristic data for a first time period. Then, based on a trained shared weight neural network, the first communication traffic data is predicted to obtain second communication traffic data for a second time period. Since the second time period is after the first time period, the trained shared weight neural network has fewer network layers and a faster fitting speed, which can reduce the accumulation of errors during prediction, improve the prediction accuracy and speed of communication traffic data, and quickly and accurately predict the second communication traffic data for future time periods. Finally, the second communication traffic data is corrected using a ridge regression algorithm to obtain third communication traffic data, which can improve the stability of communication traffic data prediction.
[0159] Based on the above Figure 1 The method for predicting communication traffic data shown in this application also provides a device for predicting communication traffic data, such as... Figure 4 As shown, the communication traffic data prediction device 400 may include:
[0160] The acquisition module 410 is used to acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period.
[0161] Prediction module 420 is used to predict the first communication traffic data based on a trained shared weight neural network to obtain the second communication traffic data for a second time period, wherein the second time period is after the first time period.
[0162] The correction module 430 is used to correct the second communication traffic data using the ridge regression algorithm to obtain the third communication traffic data.
[0163] In one possible embodiment, the acquisition module 410 is specifically used for:
[0164] Obtain the time series data of communication traffic for the first time period;
[0165] Extract the communication traffic change feature data from the communication traffic time series data;
[0166] The first communication traffic data is obtained by combining the communication traffic time series data and the communication traffic change characteristic data.
[0167] In one possible embodiment, the acquisition module 410 is specifically used for:
[0168] Determine candidate change features associated with the time series data of the communication traffic;
[0169] Based on the communication traffic time series data, determine the combined solutions of different combinations of candidate change features;
[0170] Based on each of the combined solutions, the target change feature is determined from the candidate change features;
[0171] Extract the change feature data corresponding to the target change feature from the communication traffic time series data.
[0172] In one possible embodiment, the communication traffic data prediction device 400 may further include:
[0173] The merging module is used to merge the forget gate and the output gate in the initial neural network while maintaining the neural network unit structure of the initial neural network, to obtain the merged gate;
[0174] A coupling module is used to couple the merging gate and the input gate to obtain a shared weight neural network, wherein the number of network layers in the shared weight neural network is less than the number of network layers in the initial neural network.
[0175] The training module is used to train the shared weight neural network and obtain the trained shared weight neural network.
[0176] In one possible embodiment, the training module is further configured to:
[0177] Acquire first historical communication traffic data and second historical communication traffic data, wherein the time period of the second historical communication traffic data is after the time period of the first historical communication traffic data;
[0178] Based on the first historical communication traffic data and the second historical communication traffic data, a shared weight neural network is trained until the shared weight neural network meets the training stopping condition, thus obtaining a trained shared weight neural network.
[0179] In one possible embodiment, the correction module 430 is specifically used for:
[0180] The residual sequence is determined based on the second communication traffic data and the actual communication traffic data for the second time period;
[0181] The residual sequence is regressed using the ridge regression algorithm to obtain the regression amplitude coefficients;
[0182] The second communication traffic data is corrected using the regression amplitude coefficient to obtain the third communication traffic data.
[0183] In one possible embodiment, the communication traffic data prediction device 400 may further include:
[0184] The determination module is used to identify the spatiotemporal factors in the communication traffic time series data as the principal component analysis influencing factors of the communication traffic data. The spatiotemporal factors include: seasonal factors, monthly factors, location factors, and regional factors.
[0185] The analysis module is used to perform principal component analysis on the third communication traffic data by using the influencing factors of the principal component analysis, and to obtain the analyzed third communication traffic data.
[0186] In this embodiment, first communication traffic data is acquired, including communication traffic time series data and communication traffic change characteristic data for a first time period. Then, based on a trained shared weight neural network, the first communication traffic data is predicted to obtain second communication traffic data for a second time period. Since the second time period is after the first time period, the trained shared weight neural network has fewer network layers and a faster fitting speed, which can reduce the accumulation of errors during prediction, improve the prediction accuracy and speed of communication traffic data, and quickly and accurately predict the second communication traffic data for future time periods. Finally, the second communication traffic data is corrected using a ridge regression algorithm to obtain third communication traffic data, which can improve the stability of communication traffic data prediction.
[0187] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application is shown.
[0188] An electronic device may include a processor 501 and a memory 502 storing computer program instructions.
[0189] Specifically, the processor 501 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.
[0190] Memory 502 may include a large-capacity memory for data or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 502 may include removable or non-removable (or fixed) media. Where appropriate, memory 502 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 502 is a non-volatile solid-state memory. In a particular embodiment, memory 502 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.
[0191] The processor 501 reads and executes computer program instructions stored in the memory 502 to implement any of the communication traffic data prediction methods in the embodiment shown in the figure.
[0192] In one example, the electronic device may also include a communication interface 505 and a bus 510. Wherein, as... Figure 5 As shown, the processor 501, memory 502, and communication interface 505 are connected through bus 510 and complete communication with each other.
[0193] The communication interface 505 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.
[0194] Bus 510 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 510 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.
[0195] The electronic device can execute the communication traffic data prediction method in the embodiments of this application, thereby achieving a combination of Figures 1-3 The method described is for predicting communication traffic data.
[0196] Furthermore, in conjunction with the communication traffic data prediction method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement... Figures 1-3 A method for predicting communication traffic data.
[0197] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.
[0198] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.
[0199] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.
[0200] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.
Claims
1. A method for predicting communication traffic data, characterized in that, The method includes: Acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period; Based on the trained shared weight neural network, the first communication traffic data is predicted to obtain the second communication traffic data for the second time period, which is after the first time period. The second communication traffic data is corrected using the ridge regression algorithm to obtain the third communication traffic data; Before predicting the first communication traffic data based on the trained shared weight neural network, the method further includes: While maintaining the neural network unit structure of the initial neural network, the forget gate and the output gate in the initial neural network are merged to obtain the merged gate; The merging gate and the input gate are coupled to obtain a shared weight neural network, wherein the number of network layers in the shared weight neural network is less than the number of network layers in the initial neural network. Train the shared weight neural network to obtain the trained shared weight neural network.
2. The method according to claim 1, characterized in that, The acquisition of the first communication traffic data includes: Obtain the time series data of communication traffic for the first time period; Extract the communication traffic change feature data from the communication traffic time series data; The first communication traffic data is obtained by combining the communication traffic time series data and the communication traffic change characteristic data.
3. The method according to claim 2, characterized in that, The step of extracting the communication traffic change feature data from the communication traffic time series data includes: Determine candidate change features associated with the time series data of the communication traffic; Based on the communication traffic time series data, determine the combined solutions of different combinations of candidate change features; Based on each of the combined solutions, the target change feature is determined from the candidate change features; Extract the change feature data corresponding to the target change feature from the communication traffic time series data.
4. The method according to claim 1, characterized in that, Before predicting the first communication traffic data based on the trained shared weight neural network to obtain the second communication traffic data for the second time period, the method further includes: Acquire first historical communication traffic data and second historical communication traffic data, wherein the time period of the second historical communication traffic data is after the time period of the first historical communication traffic data; Based on the first historical communication traffic data and the second historical communication traffic data, a shared weight neural network is trained until the shared weight neural network meets the training stopping condition, thus obtaining a trained shared weight neural network.
5. The method according to claim 1, characterized in that, The third communication traffic data is obtained by correcting the second communication traffic data using the ridge regression algorithm, including: The residual sequence is determined based on the second communication traffic data and the actual communication traffic data for the second time period; The residual sequence is regressed using the ridge regression algorithm to obtain the regression amplitude coefficients; The second communication traffic data is corrected using the regression amplitude coefficient to obtain the third communication traffic data.
6. The method according to claim 1, characterized in that, After correcting the second communication traffic data using the ridge regression algorithm to obtain the third communication traffic data, the method further includes: The spatiotemporal factors in the communication traffic time series data are identified as the influencing factors in the principal component analysis of the communication traffic data. The spatiotemporal factors include: seasonal factors, monthly factors, location factors, and regional factors. By analyzing the influencing factors through principal component analysis, principal component analysis is performed on the third communication traffic data to obtain the analyzed third communication traffic data.
7. A device for predicting communication traffic data, characterized in that, The communication traffic data prediction device includes: The acquisition module is used to acquire first communication traffic data, which includes: communication traffic time series data and communication traffic change characteristic data for a first time period; The prediction module is used to predict the first communication traffic data based on the trained shared weight neural network to obtain the second communication traffic data for the second time period, which is after the first time period. The correction module is used to correct the second communication traffic data using the ridge regression algorithm to obtain the third communication traffic data. The communication traffic data prediction device further includes: The merging module is used to merge the forget gate and the output gate in the initial neural network while maintaining the neural network unit structure of the initial neural network, to obtain the merged gate; A coupling module is used to couple the merging gate and the input gate to obtain a shared weight neural network, wherein the number of network layers in the shared weight neural network is less than the number of network layers in the initial neural network. The training module is used to train the shared weight neural network and obtain the trained shared weight neural network.
8. An electronic device, characterized in that, The electronic device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the method for predicting communication traffic data as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the method for predicting communication traffic data as described in any one of claims 1-6.