Traffic index prediction method and device, electronic equipment and storage medium
Through the hierarchical time sequence causal network prediction model HTCNP combines causal inference and deep learning technology, the problem of inaccurate network traffic prediction is solved, accurate traffic indicator prediction and device expansion are achieved, user experience is improved and resources are saved.
Patent Information
- Application Number
- CN202510373930.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-07-01
AI Technical Summary
The prior art cannot accurately predict network traffic, resulting in inaccurate expansion of network equipment, affecting user experience and wasting resources.
The hierarchical time sequence causal network prediction model HTCNP is used, combined with causal inference testing and deep learning technology, and accurately predicted by obtaining historical business data sequences.
Provide more accurate and interpretable traffic indicator prediction results, realize accurate expansion of equipment, ensure user experience and save network resources.
Smart Images
Figure CN120238917A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of big data technology, and in particular, to a traffic index prediction method, apparatus, electronic device, and storage medium. Background Art
[0002] Since the development of the network, users' demand for wireless traffic has been continuously increasing. However, the cell network sometimes fails to meet these demands, resulting in a decrease in network speed, which seriously affects the user experience. To cope with the surge in network traffic, network equipment is usually expanded based on business experience to ensure that users' demands are met, reduce the occurrence of network congestion, improve the user experience, and reduce user complaints and churn caused by slow network speed.
[0003] Therefore, accurately predicting traffic usage and achieving precise expansion of equipment are not only the key to ensuring the user experience but also contribute to the conservation of network resources. Summary of the Invention
[0004] The present disclosure provides a traffic index prediction method, apparatus, electronic device, storage medium, and computer program product, aiming to solve the technical problems in the related art to at least some extent.
[0005] In a first aspect of an embodiment of the present disclosure, a traffic index prediction method is provided, including: obtaining a historical service data sequence; determining a to-be-predicted index from the historical service data sequence; inputting the historical service data sequence into a pre-trained hierarchical time-series causal network prediction model (HTCNP) to obtain a target prediction result corresponding to the to-be-predicted index output by the HTCNP.
[0006] In a second aspect of an embodiment of the present disclosure, a traffic index prediction apparatus is provided, including: a first obtaining module for obtaining a historical service data sequence; a determining module for determining a to-be-predicted index from the historical service data sequence; a second obtaining module for inputting the historical service data sequence into a pre-trained hierarchical time-series causal network prediction model (HTCNP) to obtain a target prediction result corresponding to the to-be-predicted index output by the HTCNP.
[0007] In a third aspect of an embodiment of the present disclosure, an electronic device is provided, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the traffic index prediction method.
[0008] In a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is enabled to execute the traffic index prediction method.
[0009] A computer program product according to an embodiment of the fifth aspect of the present disclosure includes a computer program, characterized in that the computer program is executed by a processor to perform a traffic metric prediction method.
[0010] The traffic metric prediction method, device, electronic device, storage medium, and computer program product proposed in this embodiment at least have the following beneficial effects: obtaining a historical service data sequence; determining a metric to be predicted from the historical service data sequence, and inputting the historical service data sequence into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain a target prediction result corresponding to the metric to be predicted output by the HTCNP. Since the HTCNP model can provide more accurate and interpretable prediction results by combining causal inference testing and deep learning techniques, it can effectively improve the traffic metric prediction effect.
[0011] Additional aspects and advantages of the present disclosure will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] The above and / or additional aspects and advantages of the present disclosure will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:
[0013] Figure 1 is a flowchart of a traffic metric prediction method according to a first embodiment of the present disclosure;
[0014] Figure 2 is a flowchart of a traffic metric prediction method according to a second embodiment of the present disclosure;
[0015] Figure 3 is a flowchart of a traffic metric prediction method according to a third embodiment of the present disclosure;
[0016] Figure 4 is a flowchart of a causal convolution proposed in an embodiment of the present disclosure;
[0017] Figure 5 is a block diagram of a traffic metric prediction device according to the present disclosure;
[0018] Figure 6 shows a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0019] Embodiments of the present disclosure will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary only for explaining the present disclosure and should not be construed as limiting the present disclosure. On the contrary, the embodiments of the present disclosure include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.
[0020] Here, it should be noted that the execution subject of the traffic index prediction method in this embodiment can be a traffic index prediction device, which can be implemented in software and / or hardware, and this device can be configured in an electronic device, and the electronic device can include but is not limited to a terminal, a server, etc.
[0021] It should be noted that in the technical solution of the present disclosure, the processes of obtaining, storing, using, processing, etc. of information all comply with the relevant regulations of national laws and regulations and do not violate public order and good customs.
[0022] Figure 1 is a flowchart of the traffic index prediction method shown in the first embodiment of the present disclosure, as Figure 1 shown, the method includes:
[0023] S101: Obtain a historical service data sequence.
[0024] In the embodiments of the present disclosure, the historical service data includes at least one of the following: radio connection establishment success rate; radio resource control (RRC) connection establishment success rate; uplink physical resource block (PRB) utilization rate; downlink PRB utilization rate.
[0025] That is to say, in the embodiments of the present disclosure, multiple key performance indicators can be collected from a network system, such as historical service data such as radio connection establishment success rate, average number of radio resource control (RRC) connections, and average utilization rate of uplink physical resource blocks (PRBs). Then, subsequent traffic index prediction methods can be executed based on the historical service data sequence. For details, refer to the subsequent embodiments and will not be elaborated here.
[0026] In the embodiments of the present disclosure, the radio connection establishment success rate, average number of RRC connections, and average utilization rate of uplink PRBs can be collected, and the collected historical service data sequence can be expressed as s1, s2,..., s n , where n is the number of characteristics of the collected historical service data, and the historical service data sequence can be expressed as S = {s1, s2,..., s n}.
[0027] In the embodiments of the present disclosure, by obtaining a historical business data sequence, the operating conditions of the network can be monitored and analyzed in real time, so as to achieve accurate identification and evaluation of network traffic changes. In addition, we deeply extract network scenario features, such as region, environment, equipment vendor, and frequency band features, to accurately identify and distinguish traffic characteristics in different scenarios. The data collection process adopts a sliding window method to ensure the real-time update and processing of key data.
[0028] S102: Determine the index to be predicted from the historical business data sequence.
[0029] In the embodiments of the present disclosure, after obtaining the historical business data sequence, the index to be predicted can be determined from the historical business data sequence.
[0030] That is to say, in the embodiments of the present disclosure, an index in the historical business data sequence S is designated as the index to be predicted. Then, the index prediction result corresponding to the index to be predicted can be jointly predicted with the historical business data sequence.
[0031] S103: Input the historical business data sequence into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain the target prediction result corresponding to the index to be predicted output by the HTCNP.
[0032] Among them, the hierarchical temporal causal network prediction model (HTCNP), which is based on causal relationship analysis and deep learning, is specifically designed to capture and analyze causal relationships in time series data and handle the non-linear characteristics of data. The HTCNP model can not only identify the key factors affecting network traffic but also adjust the prediction strategy according to the dynamic changes of these factors by combining causal inference tests and deep learning techniques. Through hierarchical processing and in-depth understanding of causal relationships, this method can provide more accurate and interpretable prediction results.
[0033] In the embodiments of the present disclosure, after determining the index to be predicted from the historical business data sequence, the historical business data sequence can be input into the pre-trained HTCNP to obtain the target prediction result corresponding to the index to be predicted output by the HTCNP.
[0034] In the embodiments of the present disclosure, let T = {t1, t2,..., t7} represent the dataset for seven consecutive days, where where t represents the summary data for one day. The task of the prediction model is to use the data T to predict t8, that is, the data for the eighth day. This prediction mode can be expressed as:
[0035] t8 = f(t1, t2,..., t7)
[0036] Among them, f represents our HTCNP model. This model adopts an iterative prediction method. That is, after the data on the eighth day is collected, it will be incorporated into the prediction data set, and the data {t2, t3, …, t8} from the second day to the eighth day of the proposal is used to predict the data t9 on the ninth day. For example, we collect the data of the RRC average connection rate from the first day to the seventh day to predict the data on the eighth day. When the data on the eighth day is collected, it will be incorporated into the training data, and the data from the second day to the eighth day is used to predict the data on the ninth day. Figure 2 It represents a schematic diagram of the prediction mode of this process. And so on, m represents the prediction time (the mth day), then the prediction formula is:
[0037] t m+1 = f(t m-6 , t m-5 , …, t m ).
[0038] In the embodiments of the present disclosure, by obtaining a historical service data sequence; determining a to-be-predicted indicator from the historical service data sequence, and inputting the historical service data sequence into a pre-trained hierarchical time series causal network prediction model HTCNP to obtain a target prediction result corresponding to the to-be-predicted indicator output by the HTCNP. Since the HTCNP model can provide more accurate and interpretable prediction results by combining causal inference testing and deep learning techniques, it can effectively improve the prediction effect of traffic indicators.
[0039] Figure 2 is a schematic flowchart of a traffic indicator prediction method according to the second embodiment of the present disclosure. As Figure 2 shown, the method includes:
[0040] S201: Obtain a historical service data sequence.
[0041] S202: Determine a to-be-predicted indicator from the historical service data sequence.
[0042] For the descriptions of S201 - S202, specific reference can be made to the above embodiments, which will not be elaborated here.
[0043] S203: Perform an ADF test on the historical service data sequence for time series stationarity to obtain a first test result.
[0044] In the embodiments of the present disclosure, in the hierarchical time series causal network prediction model (HTCNP), ensuring the stationarity of the data is crucial for improving the accuracy of model prediction. The ADF test is a statistical method for determining whether a time series has a unit root. Its core principle is to judge the stationarity of time series data by testing the existence of a unit root.
[0045] In an embodiment of the present disclosure, after determining the to-be-predicted indicator from the historical business data sequence, the time series stationarity (Augmented Dickey Fuller, ADF) detection can be performed on the historical business data sequence to obtain a first detection result.
[0046] S204: Determine the target business data sequence according to the to-be-predicted indicator, the first detection result, and the historical business data sequence.
[0047] In an embodiment of the present disclosure, after performing ADF detection on the historical business data sequence to obtain a first detection result, the target business data sequence can be determined according to the to-be-predicted indicator, the first detection result, and the historical business data sequence.
[0048] In some embodiments, it can be a joint large model to implement determining the target business data sequence according to the to-be-predicted indicator, the first detection result, and the historical business data sequence, that is, the to-be-predicted indicator, the first detection result, and the historical business data sequence can be input into the large model together to obtain the target historical business data sequence output by the large model.
[0049] In other embodiments, to determine the target business data sequence according to the to-be-predicted indicator, the first detection result, and the historical business data sequence, it can also be that when the first detection result indicates that the historical business data sequence is a time-stationary sequence, the historical business data sequence is directly used as the target business data sequence, and when the first detection result indicates that the historical business data sequence is not a time-stationary sequence, the target business data sequence is determined from the historical business data sequence by combining the to-be-predicted indicator and the first detection result.
[0050] S205: Input the target business data sequence into the pre-trained hierarchical time-series causal network prediction model HTCNP to obtain the target prediction result corresponding to the to-be-predicted indicator output by the HTCNP.
[0051] In an embodiment of the present disclosure, after determining the target business data sequence according to the to-be-predicted indicator, the first detection result, and the historical business data sequence, it can be to input the target business data sequence into the pre-trained hierarchical time-series causal network prediction model HTCNP to obtain the target prediction result corresponding to the to-be-predicted indicator output by the HTCNP.
[0052] In the embodiments of the present disclosure, by obtaining a historical business data sequence, determining a to-be-predicted indicator from the historical business data sequence, performing an ADF test for time series stationarity on the historical business data sequence to obtain a first test result, determining a target business data sequence according to the to-be-predicted indicator, the first test result, and the historical business data sequence, and inputting the target business data sequence into a pre-trained hierarchical time series causal network prediction model (HTCNP) to obtain a target prediction result corresponding to the to-be-predicted indicator, it is thus possible to efficiently convert the to-be-distributed data into target storage data that can be stored in the to-be-stored database according to the field type mapping relationship between the source data and the to-be-stored database, thereby improving the prediction efficiency of traffic indicators.
[0053] Figure 3 is a schematic flowchart of a traffic indicator prediction method shown in the third embodiment of the present disclosure. As Figure 3 shown, the method includes:
[0054] S301: Obtain a historical business data sequence.
[0055] S302: Determine a to-be-predicted indicator from the historical business data sequence.
[0056] S303: Perform an ADF test for time series stationarity on the historical business data sequence to obtain a first test result.
[0057] For the descriptions of S301 - S303, specific reference may be made to the above embodiments and will not be elaborated here.
[0058] S304: Determine a first business data sequence according to the first test result and the historical business data sequence.
[0059] In the embodiments of the present disclosure, after performing an ADF test for time series stationarity on the historical business data sequence to obtain a first test result, a first business data sequence may be determined according to the first test result and the historical business data sequence.
[0060] Optionally, in some embodiments, after determining the first business data sequence according to the first test result and the historical business data sequence, when the first test result indicates that the historical business data sequence is not a stationary time series, the historical business data sequence may be differenced to obtain the first business data sequence, or when the first test result indicates that the historical business data sequence is a stationary time series, the historical business data sequence may be used as the first business data sequence.
[0061] That is to say, in the embodiments of the present disclosure, when the first detection result indicates that the historical service data sequence is not a stationary time series, the historical service data sequence may be differenced to convert it into a first service data sequence with time stationarity. When the first detection result indicates that the historical service data sequence is a stationary time series, the historical service data sequence is used as the first service data sequence, so as to ensure the stationarity of the service data sequence and improve the prediction effect of traffic indicators.
[0062] S305: Determine a second service data sequence having a causal relationship with the to-be-predicted indicator from the first service data sequence.
[0063] In the embodiments of the present disclosure, after determining the first service data sequence according to the first detection result and the historical service data sequence, a second service data sequence having a causal relationship with the to-be-predicted indicator may be determined from the first service data sequence, so as to help the HTCNP model understand the potential causal relationship between sequences.
[0064] Optionally, in some embodiments, to determine a second service data sequence having a causal relationship with the to-be-predicted indicator from the first service data sequence, a Granger causality test is performed on the first service data sequence according to the to-be-predicted indicator to obtain a second test result. When the second test result indicates that there is a causal relationship between the first service data sequence and the to-be-predicted indicator, the first service data sequence is used as the second service data sequence.
[0065] Among them, the Granger causality test can help the model understand the causal relationship between sequences, thereby improving the accuracy and interpretability of the model.
[0066] Among them, the Granger causality test is used to detect whether a time series can predict the future value of another time series. The basic idea of this test is that if the past values of sequence s1 can significantly improve the prediction accuracy of the future values of sequence s2, then s1 is called the Granger cause of s2. For example, when we predict the average RPC connection rate and the E-RAB establishment success rate is one of the features, it can be shown that the E-RAB establishment success rate is the Granger cause of the average RPC connection rate, and there is no limitation to this.
[0067] In the embodiments of the present disclosure, after the Granger causality test, the first service data sequence is divided into a second service data sequence s having a causal relationship with the to-be-predicted indicator k and a sequence s having no causal relationship v , where the second service data sequence will be used in the subsequent traffic indicator prediction process.
[0068] S306: Determine the similarity between each two data in the second service data sequence respectively.
[0069] In the embodiments of the present disclosure, after determining the second business data sequence having a causal relationship with the to-be-predicted indicator from the first business data sequence, the similarity between each two data in the second business data sequence can be determined respectively.
[0070] For example, in the embodiments of the present disclosure, the second business data sequence can be processed by Dynamic Time Warping (DTW) respectively to determine the similarity between each two data in the second business data sequence.
[0071] Among them, DTW is an algorithm used to measure the similarity between two time series. DTW finds the best match between two time series by non-linearly aligning the sequences to minimize the difference between them. DTW is used to process those sequences that have been confirmed to have a causal relationship through the Granger causality test. Through DTW, we can quantify the similarity between different time series and effectively align the sequences, so as to mine the sequence correlation between the prediction sequence and the feature sequence, thereby providing a basis for subsequent clustering of the sequences.
[0072] S307: Determine the target business data sequence according to the first business data sequence, the similarity, and the second business data sequence.
[0073] In the embodiments of the present disclosure, after determining the similarity between each two data in the second business data sequence respectively, the target business data sequence can be determined according to the first business data sequence, the similarity, and the second business data sequence.
[0074] In some embodiments, the joint large model can be used to determine the target business data sequence according to the first business data sequence, the similarity, and the second business data sequence, that is, the first business data sequence, the similarity, and the second business data sequence can be input into the large model together to obtain the target business data sequence output by the large model.
[0075] Optionally, in some embodiments, to determine the target business data sequence according to the first business data sequence, the similarity, and the second business data sequence, the data in the second business data sequence can be clustered according to the similarity to obtain the clustering result, the data in the first business data sequence other than the second business data sequence can be convolved to obtain the convolution result, and the clustering result and the convolution result are jointly used as the target business data sequence.
[0076] That is to say, in the embodiments of the present disclosure, for the sequence s without causal relationship vwill be subjected to a convolution operation, which includes a convolutional layer (with a convolutional kernel size of 3x3 and a stride of 1) and a pooling layer (max pooling), that is, the sequence s v After convolution with the convolutional kernel, max pooling is performed to ensure that the sequence after the convolution operation is consistent with the original sequence, and finally the output is flattened into a one-dimensional vector. Such a design can extract basic features from the data, prevent the loss of information in the sequence s without causal relationship v and solve the problem of manually designing feature extractors in neural networks, and can improve the training and inference speed of the model while ensuring accuracy, ensuring the comprehensiveness of information and the efficiency of the model. This can optimize the training and inference speed of the model without sacrificing complexity, ensuring the comprehensiveness of information and the efficiency of the model.
[0077] Among them, the clustering process can be, for example, a hierarchical clustering process. Hierarchical clustering is an analysis method used to group data points into a tree-like clustering structure. The hierarchical clustering method is used to further classify the time series with similarity after DTW processing. The input sequence will be clustered into two clusters according to the sequence correlation distance matrix calculated by DTW, which can improve the model accuracy while ensuring the calculation efficiency (classifying into too many clusters will lead to an increase in the number of sub-models to be trained and a decrease in efficiency).
[0078] In the embodiment of the present disclosure, during the process of performing hierarchical clustering on the data in the second service data sequence, it can be to first calculate the distance between each pair of sequences, and then merge the closest sequence pairs into a new cluster, repeating this process until all sequences are merged into a single cluster or reach a preset number of clusters.
[0079] S308: Input the target service data sequence into the pre-trained hierarchical time-series causal network prediction model HTCNP to obtain the target prediction result corresponding to the index to be predicted output by HTCNP.
[0080] In the embodiment of the present disclosure, after determining the target service data sequence according to the first service data sequence, similarity, and the second service data sequence, it can be to input the target service data sequence into the pre-trained hierarchical time-series causal network prediction model HTCNP to obtain the target prediction result corresponding to the index to be predicted output by HTCNP.
[0081] Optionally, in some embodiments, inputting the target business data sequence into a pre-trained hierarchical temporal causal network prediction model (HTCNP) to obtain a target prediction result corresponding to the metric to be predicted may be to input the clustering result into the first temporal convolutional network (TCN) in the HTCNP to obtain a first prediction result output by the first TCN, input the convolution result into the second temporal convolutional network (TCN) in the HTCNP to obtain a second prediction result output by the second TCN, and perform weighted summation on the first prediction result and the second prediction result to obtain the target prediction result.
[0082] In the embodiments of the present disclosure, in the HTCNP model, we use TCN to process the two clusters of sequences obtained after DTW and hierarchical clustering. Each cluster of sequences is separately trained on a TCN model with different hyperparameter settings based on its characteristics. This allows each model to better adapt to the specific features and structures of its input data, thereby improving the accuracy of prediction. The key steps of TCN include causal convolution, dilated convolution, and residual connection.
[0083] Among them, causal convolution can ensure that the model can only use current and previous information when making predictions, avoiding information leakage. See Figure 4 , Figure 4 is a schematic flowchart of causal convolution proposed in an embodiment of the present disclosure, and this causal convolution can be expressed as:
[0084]
[0085] Among them, h t is the output at time t, f i is the weight of the convolution kernel. x t-i is the value of the sequence at time t - i, and m is the size of the convolution kernel.
[0086] Among them, dilated convolution allows the network to capture longer-range data dependencies without significantly increasing the number of parameters. Dilated convolution can process the input data with a larger stride, and the process of dilated convolution can be expressed as:
[0087]
[0088] Among them, d is the dilation factor, representing the spacing between each element in the convolution kernel.
[0089] Among them, residual connection usually adds a residual connection in the TCN to help information propagate in deeper network layers, thereby avoiding the problem of vanishing gradients. The residual module can be expressed as:
[0090] H t = Activation(h t + x t )
[0091] Among them, H t is the output after passing through the activation function and the residual connection, and x t is the part directly passed from the previous layer or the input.
[0092] In the embodiments of the present disclosure, in the HTCNP model, we use TCNs with different hyperparameter configurations (such as different numbers of layers, different sizes of convolutional kernels, different dilation factors) to train sequences of different clusters. The model f of each cluster k is optimized and configured according to its sequence characteristics. The final predicted output is the weighted sum of the first prediction result output by the first TCN and the second prediction result output by the second TCN. The weighted calculation process is as follows:
[0093]
[0094] Among them, w k is the weight adjusted according to the model performance, which reflects the importance of different models in the overall prediction task.
[0095] In the embodiments of the present disclosure, by obtaining the historical business data sequence, determining the index to be predicted from the historical business data sequence, performing the ADF test for time series stationarity on the historical business data sequence to obtain the first test result, determining the first business data sequence according to the first test result and the historical business data sequence, determining the second business data sequence having a causal relationship with the index to be predicted from the first business data sequence, respectively determining the similarity between every two data in the second business data sequence, determining the target business data sequence according to the first business data sequence, the similarity, and the second business data sequence. Thus, the hierarchical clustering method is used to further classify the time series with similarity after DTW processing. The input sequence will be clustered into two clusters according to the sequence correlation distance matrix calculated by DTW, which can improve the model accuracy while ensuring the calculation efficiency (classifying into too many clusters will lead to an increase in the number of sub-models to be trained and a decrease in efficiency). The target business data sequence is input into the pre-trained hierarchical time series causal network prediction model HTCNP to obtain the target prediction result corresponding to the index to be predicted output by the HTCNP, so as to accurately predict the traffic usage, realize the precise expansion of the device, ensure the user experience, and save network resources.
[0096] Figure 5 is a block diagram of a traffic index prediction device shown according to the present disclosure. As Figure 5 shown, the traffic index prediction device 50 includes:
[0097] The first acquisition module 501 is configured to acquire the historical business data sequence;
[0098] A determination module 502, configured to determine an index to be predicted from a historical service data sequence;
[0099] A second acquisition module 503, configured to input the historical service data sequence into a pre-trained hierarchical time series causal network prediction model HTCNP, so as to obtain a target prediction result corresponding to the index to be predicted output by the HTCNP.
[0100] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0101] Perform an ADF test on the time series stationarity of the historical service data sequence to obtain a first test result;
[0102] Determine a target service data sequence according to the index to be predicted, the first test result, and the historical service data sequence;
[0103] Input the target service data sequence into a pre-trained hierarchical time series causal network prediction model HTCNP, so as to obtain a target prediction result corresponding to the index to be predicted output by the HTCNP.
[0104] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0105] Determine a first service data sequence according to the first test result and the historical service data sequence;
[0106] Determine a second service data sequence having a causal relationship with the index to be predicted from the first service data sequence;
[0107] Determine the similarity between every two data in the second service data sequence respectively;
[0108] Determine a target service data sequence according to the first service data sequence, the similarity, and the second service data sequence.
[0109] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0110] In the case where the first test result indicates that the historical service data sequence is not a stationary time series, perform differencing processing on the historical service data sequence to obtain a first service data sequence; or
[0111] In the case where the first test result indicates that the historical service data sequence is a stationary time series, use the historical service data sequence as the first service data sequence.
[0112] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0113] Cluster the data in the second service data sequence according to similarity to obtain the clustering result;
[0114] Perform convolution processing on the other data in the first service data sequence except the second service data sequence to obtain the convolution result;
[0115] Use the clustering result and the convolution result together as the target service data sequence.
[0116] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0117] Perform a Granger causality test on the first service data sequence according to the index to be predicted to obtain a second test result.
[0118] In the case where the second test result indicates that there is a causal relationship between the first service data sequence and the index to be predicted, use the first service data sequence as the second service data sequence.
[0119] In some embodiments of the present disclosure, the second acquisition module 503 is further configured to:
[0120] Input the clustering result into the first temporal convolutional network TCN in the HTCNP to obtain a first prediction result output by the first TCN;
[0121] Input the convolution result into the second temporal convolutional network TCN in the HTCNP to obtain a second prediction result output by the second TCN
[0122] Perform weighted summation on the first prediction result and the second prediction result to obtain the target prediction result.
[0123] In some embodiments of the present disclosure, the historical service data includes at least one of the following:
[0124] Radio connection rate;
[0125] Radio Resource Control (RRC) connection rate;
[0126] Uplink Physical Resource Block (PRB) utilization rate;
[0127] Downlink PRB utilization rate.
[0128] In the embodiments of the present disclosure, by obtaining a historical service data sequence; determining an index to be predicted from the historical service data sequence, and inputting the historical service data sequence into a pre-trained hierarchical temporal causal network prediction model (HTCNP) to obtain a target prediction result corresponding to the index to be predicted output by the HTCNP. Since the HTCNP model combines causal inference testing and deep learning techniques, it can provide more accurate and interpretable prediction results, thereby effectively improving the prediction effect of traffic metrics.
[0129] To implement the above embodiments, the present application also provides an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the traffic metric prediction method provided in the foregoing embodiments.
[0130] To implement the above embodiments, the present application also provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the traffic metric prediction method provided in the foregoing embodiments when executed by a processor.
[0131] Figure 6 A block diagram of an exemplary electronic device suitable for implementing the embodiments of the present disclosure is shown.
[0132] Figure 6 The illustrated electronic device 6 is merely an example and should not impose any limitation on the functions and scope of use of the embodiments of the present disclosure.
[0133] As Figure 6 shown, the electronic device 6 is presented in the form of a general-purpose computing device. The components of the electronic device 6 may include, but are not limited to: one or more processors or processing units 16, a memory 28, and a bus 18 connecting different system components (including the memory 28 and the processing unit 16).
[0134] The bus 18 represents one or more of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the multiple bus architectures. By way of example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnection (PCI) bus.
[0135] The electronic device 6 typically includes a variety of computer system-readable media. These media can be any available media accessible by the electronic device 6, including volatile and non-volatile media, removable and non-removable media.
[0136] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 6 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used for reading and writing on a non-removable, non-volatile magnetic medium ( Figure 6 not shown, commonly referred to as a "hard disk drive").
[0137] Although Figure 6 not shown in, a disk drive for reading and writing on a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing on a removable non-volatile optical disk (such as a compact disc read only memory (CD-ROM), a digital video disc read only memory (DVD-ROM) or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 through one or more data medium interfaces. The memory 28 may include at least one program product having a set (such as at least one) of program modules that are configured to perform the functions of the embodiments of the present disclosure.
[0138] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. An implementation of a network environment may be included in each or some combination of these examples. The program modules 42 generally perform the functions and / or methods in the embodiments described in the present disclosure.
[0139] The electronic device 6 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a human body to interact with the electronic device 6, and / or communicate with any device that enables the electronic device 6 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Moreover, the electronic device 6 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the electronic device 6 through the bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 6, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.
[0140] The processing unit 16 executes various functional applications and parameter information determination by running the programs stored in the memory 28, such as implementing the service traffic index prediction method mentioned in the foregoing embodiments, or implementing the service data acquisition method mentioned in the foregoing embodiments.
[0141] It should be noted that in the description of the present disclosure, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present disclosure, unless otherwise specified, the meaning of "a plurality" is two or more.
[0142] Any process or method description in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in the reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the technical field to which the embodiments of the present disclosure belong.
[0143] It should be understood that various parts of the present disclosure can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.
[0144] Those of ordinary skill in the art can understand that all or part of the steps carried by the method of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.
[0145] In addition, in each embodiment of the present disclosure, each functional unit can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.
[0146] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disk, etc.
[0147] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.
[0148] Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.
Claims
1. A flow index prediction method, characterized in that: The method comprises: Obtain historical business data sequence; Determining the indicator to be predicted from the historical business data sequence; The historical business data sequence is input into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain a target prediction result output by the HTCNP corresponding to the indicator to be predicted.
2. The method according to claim 1, characterized in that The step of inputting the historical business data sequence into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain a target prediction result output by the HTCNP corresponding to the indicator to be predicted includes: Performing an ADF test on the time series stationarity of the historical business data sequence to obtain a first test result; Determine a target business data sequence according to the indicator to be predicted, the first detection result and the historical business data sequence; The target business data sequence is input into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain a target prediction result output by the HTCNP corresponding to the indicator to be predicted.
3. The method according to claim 2, characterized in that The determining of the target business data sequence according to the indicator to be predicted, the first detection result and the historical business data sequence includes: Determine a first service data sequence according to the first detection result and the historical service data sequence; Determining, from the first business data sequence, a second business data sequence having a causal relationship with the indicator to be predicted; respectively determining the similarity between every two data in the second business data sequence; The target service data sequence is determined according to the first service data sequence, the similarity and the second service data sequence.
4. The method according to claim 3, characterized in that The determining, according to the first detection result and the historical service data sequence, a first service data sequence includes: When the first detection result indicates that the historical service data sequence is not a stationary time sequence, performing differential processing on the historical service data sequence to obtain the first service data sequence; or, When the first detection result indicates that the historical business data sequence is the stationary time series, the historical business data sequence is used as the first business data sequence.
5. The method according to claim 3, characterized in that The determining the target service data sequence according to the first service data sequence, the similarity and the second service data sequence includes: Performing clustering processing on the data in the second service data sequence according to the similarity to obtain a clustering processing result; Performing convolution processing on other data in the first service data sequence except the second service data sequence to obtain a convolution processing result; The clustering processing result and the convolution processing are used together as the target business data sequence.
6. The method according to claim 3, characterized in that The determining, from the first business data sequence, a second business data sequence having a causal relationship with the indicator to be predicted comprises: Performing a Granger causality test on the first business data sequence according to the indicator to be predicted to obtain a second test result; When the second test result indicates that there is a causal relationship between the first business data sequence and the indicator to be predicted, the first business data sequence is used as the second business data sequence.
7. The method according to claim 2, characterized in that The step of inputting the target business data sequence into a pre-trained hierarchical temporal causal network prediction model HTCNP to obtain a target prediction result output by the HTCNP corresponding to the indicator to be predicted includes: Inputting the clustering processing result into the first temporal convolutional network TCN in the HTCNP to obtain a first prediction result output by the first TCN; Input the convolution processing result into the second temporal convolution network TCN in the HTCNP to obtain a second prediction result output by the second TCN; A weighted sum is performed on the first prediction result and the second prediction result to obtain the target prediction result.
8. The method according to claims 1 to 7, characterized in that The historical business data includes at least one of the following: Wireless connection rate; Radio Resource Control RRC connection rate; Uplink physical resource block PRB utilization; Downlink PRB utilization.
9. A flow index prediction device, characterized in that: The device comprises: A first acquisition module is used to acquire a historical business data sequence; A determination module, used to determine the indicator to be predicted from the historical business data sequence; The second acquisition module is used to input the historical business data sequence into the pre-trained hierarchical temporal causal network prediction model HTCNP to obtain the target prediction result corresponding to the indicator to be predicted output by the HTCNP.
10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method according to any one of claims 1 to 8.
11. A computer-readable storage medium, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 8.
12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.