Method, apparatus, device, medium, and program product for traffic prediction

Abstract

The application discloses a traffic prediction method, device, equipment, medium and program product. The method uses a preset target prediction model to predict the traffic usage of a target cell in a first preset time period, thereby obtaining a first prediction value of a target traffic index in the first preset time period, then obtains a corresponding target prediction error value, and then determines a first target prediction value of the target traffic index in the first preset time period according to the first prediction value of the target traffic index and the corresponding target prediction error value, which is equivalent to compensating for the error of the target traffic index, thereby improving the prediction accuracy of each target traffic index of the model, accurately predicting the traffic usage of the base station cell, enabling the network equipment provider to accurately master the traffic trend of the base station cell and accurately expand or shrink the capacity of the base station cell communication equipment, thereby ensuring the quality of experience of users while avoiding waste of resources.

Description

Technical Field

[0001] This application belongs to the field of communication technology, and in particular relates to a method, apparatus, device, medium and program product for traffic prediction. Background Technology

[0002] As the carrier of mobile communication and network services, base stations undertake important functions such as communication connection with user terminals, signal transmission, and signal processing. Network equipment providers can use traffic prediction models to predict the traffic usage of base station cells over a future period, thereby understanding the traffic trends of base station cells and dynamically expanding or shrinking the capacity of base station cell communication equipment, ensuring the quality of user experience while avoiding resource waste.

[0003] Existing traffic prediction models often only predict a single indicator when predicting traffic usage in base station cells, which cannot accurately reflect traffic trends. They are not very useful for expanding or shrinking the capacity of base station cell communication equipment, and the models have certain prediction errors, resulting in inaccurate predictions. Therefore, they cannot accurately expand or shrink the capacity of base station cell communication equipment based on traffic predictions and trends, and cannot avoid wasting resources while ensuring the quality of user experience.

[0004] Therefore, the problem with existing technologies is that existing traffic prediction models cannot accurately predict the traffic usage of base station cells, and cannot avoid resource waste while ensuring the quality of user experience. Summary of the Invention

[0005] This application provides a method, apparatus, device, medium, and program product for traffic prediction, which solves the problem that existing traffic prediction models cannot accurately predict the traffic usage of base station cells and cannot avoid resource waste while ensuring the quality of user experience.

[0006] In a first aspect, embodiments of this application provide a method for traffic prediction, including:

[0007] Using a pre-defined target prediction model, the traffic usage of the target cell in the first pre-defined time period is predicted to obtain the first predicted value of the target traffic index in the first pre-defined time period. There are multiple target traffic indices.

[0008] In the preset correspondence between traffic indicators and prediction error values, the target prediction error value corresponding to the target traffic indicator is obtained, where the prediction error value is the error value of the target prediction model in predicting the traffic indicator.

[0009] Based on the first predicted value and the target prediction error value, the first target predicted value of the target flow index is determined within the first preset time period.

[0010] In one possible implementation of the first aspect, the target traffic metric includes at least two of mobile network traffic values, resource utilization, and number of user connections.

[0011] In one possible implementation of the first aspect, before predicting traffic usage within a first preset time period using a preset target prediction model to obtain a first predicted value of the target traffic index within the first preset time period, the method further includes:

[0012] Obtain the actual value of the target traffic indicator within the first historical period and the actual value within the second target preset period. There are multiple first historical periods, and multiple first historical periods constitute the first target historical period.

[0013] Using a preset initial prediction model, based on the real values ​​in the first historical time period, the traffic usage of the target cell in the second preset time period is predicted to obtain the second predicted value of the target traffic index in the second preset time period. There are multiple second preset time periods, and multiple second preset time periods constitute the second target preset time period.

[0014] The second predicted values ​​of the target flow index within multiple second preset time periods are added together to obtain the third predicted value of the target flow index within the second target preset time period.

[0015] Based on the actual value and the third predicted value of the target flow index within the second target preset time period, the initial prediction model is corrected to obtain the target prediction model.

[0016] In one possible implementation of the first aspect, the true value of the target flow index in the first historical period and the true value in the second target preset period are both greater than or equal to 0 and less than or equal to 1; the method further includes:

[0017] The first target predicted value of the target flow index within the first preset time period is inversely normalized to obtain the second target predicted value of the target flow index within the first preset time period.

[0018] In one possible implementation of the first aspect, there are M first historical time periods, where M is a positive integer; before obtaining the true value of the target traffic indicator within the first historical time period and the true value within the second target preset time period, the method further includes:

[0019] Obtain N true values ​​of the target traffic metric within the second target historical period, where the second target historical period includes N first historical periods, where N is a positive integer and N is less than M;

[0020] Based on preset expansion rules, the actual values ​​of the target traffic metric are generated within M first historical time periods.

[0021] In one possible implementation of the first aspect, before obtaining N true values ​​of the target traffic indicator within a second target historical time period, the method further includes:

[0022] Obtain N raw values ​​of the target traffic metric within the second target historical time period;

[0023] Anomaly detection is performed on the original value according to the preset anomaly detection rules to obtain the target detection result, wherein the target detection result includes anomalies in the original value;

[0024] If the target detection result shows that the original value is abnormal, the abnormal original value is replaced with the normal value according to the preset interpolation rule to obtain N true values ​​of the target flow index in the second target historical period.

[0025] In one possible implementation of the first aspect, there are multiple anomaly detection rules; anomaly detection is performed on the original value according to the preset anomaly detection rules to obtain the target detection result, including:

[0026] Anomaly detection is performed on the original value according to multiple anomaly detection rules, resulting in multiple sub-detection results;

[0027] The number of target sub-detection results is compared with the preset number to obtain the target detection results, where the target sub-detection results are the sub-detection results of the original values ​​that are abnormal.

[0028] Based on the same inventive concept, in a second aspect, embodiments of this application also provide a traffic flow prediction apparatus, comprising:

[0029] The prediction module is used to predict the traffic usage of the target cell in the first preset time period using a preset target prediction model, and to obtain the first predicted value of the target traffic index in the first preset time period. There are multiple target traffic indices.

[0030] The acquisition module is used to acquire the target prediction error value corresponding to the target traffic indicator from the preset correspondence between traffic indicators and prediction error values, wherein the prediction error value is the error value of the target prediction model in predicting the traffic indicator.

[0031] The determination module is used to determine the first target predicted value of the target flow index within a first preset time period based on the first predicted value and the target prediction error value.

[0032] Based on the same inventive concept, in a third aspect, embodiments of this application also provide a traffic prediction device, the device including a processor and a memory storing computer program instructions; the processor executes the computer program instructions to implement the traffic prediction method in the first aspect, or any possible implementation of the first aspect.

[0033] Based on the same inventive concept, in a fourth aspect, embodiments of this application also provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the method of traffic prediction in the first aspect or any possible implementation of the first aspect.

[0034] Based on the same inventive concept, in a fifth aspect, embodiments of this application also provide a computer program product, wherein instructions in the computer program product, when executed by a processor of a device, enable the device to perform the flow prediction method of the first aspect or any possible implementation thereof.

[0035] The traffic prediction method, apparatus, device, medium, and program products of this application embodiment utilize a preset target prediction model to predict the traffic usage of a target cell within a first preset time period, thereby obtaining a first predicted value of a target traffic indicator within the first preset time period. There are multiple target traffic indicators. Then, in a preset correspondence between traffic indicators and prediction error values, the target prediction error value corresponding to each target traffic indicator can be obtained. The prediction error value is the error value of the target prediction model in predicting the traffic indicator. Next, based on the first predicted value of the target traffic indicator and the corresponding target prediction error value, a first target predicted value of the target traffic indicator within the first preset time period can be determined, which is equivalent to compensating for the error of the target traffic indicator. This improves the model's prediction accuracy for each target traffic indicator, enabling accurate prediction of the traffic usage of base station cells. This allows network equipment providers to accurately grasp the traffic trends of base station cells and accurately scale up or down the communication equipment of base station cells, thereby ensuring user experience quality while avoiding resource waste. Attached Figure Description

[0036] 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.

[0037] Figure 1 This is a schematic flowchart of a traffic prediction method provided in an embodiment of this application;

[0038] Figure 2 This is another flowchart illustrating the traffic prediction method provided in the embodiments of this application;

[0039] Figure 3 This is a flowchart illustrating an expansion of the sample dataset in the traffic prediction method provided in this application embodiment;

[0040] Figure 4This is a flowchart illustrating the preprocessing of a sample dataset in the traffic prediction method provided in this application embodiment;

[0041] Figure 5 This is a schematic diagram of the training process of the target prediction model in the traffic prediction method provided in the embodiments of this application;

[0042] Figure 6 This is a flowchart illustrating the method for detecting outliers in the traffic prediction method provided in this application embodiment;

[0043] Figure 7 This is a schematic diagram of a flow prediction device provided in an embodiment of this application;

[0044] Figure 8 This is a schematic diagram of a traffic prediction device provided in an embodiment of this application. Detailed Implementation

[0045] The features and exemplary embodiments of various aspects of this application will be described in detail below. 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 intended to explain this application and not to limit it. 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.

[0046] 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 said element.

[0047] Various modifications and variations can be made to this application without departing from its spirit or scope, which will be apparent to those skilled in the art. Therefore, this application is intended to cover modifications and variations falling within the scope of the corresponding claims (the claimed technical solutions) and their equivalents. It should be noted that the implementation methods provided in the embodiments of this application can be combined with each other without contradiction.

[0048] Before describing the technical solutions provided in the embodiments of this application, in order to facilitate understanding of the embodiments of this application, this application first specifically explains the problems existing in the related technologies:

[0049] As described in the background section, base stations, as carriers of mobile communication and network services, undertake important functions such as communication connection with user terminals, signal transmission, and signal processing. Network equipment providers can use traffic prediction models to predict the traffic usage of base station cells over a future period, thereby understanding the traffic trends of base station cells and dynamically expanding or shrinking the capacity of base station cell communication equipment, ensuring user experience quality while avoiding resource waste.

[0050] Existing traffic prediction models often only predict a single indicator when forecasting traffic usage in base station cells. They fail to filter out multiple features closely related to traffic, cannot reflect the essential changing patterns of network traffic, and cannot accurately reflect traffic trends. Therefore, they offer limited reference value for scaling up or down base station cell communication equipment. Furthermore, these models have a certain degree of prediction error, resulting in inaccurate predictions. Consequently, they cannot accurately scale up or down base station cell communication equipment based on traffic predictions and trends, failing to ensure user experience quality while avoiding resource waste. Therefore, existing traffic prediction models cannot accurately predict base station cell traffic usage and cannot guarantee user experience quality while avoiding resource waste.

[0051] Based on this, embodiments of this application provide a method, apparatus, device, medium, and program product for traffic prediction, which can solve the problem that existing traffic prediction models cannot accurately predict the traffic usage of base station cells and cannot avoid resource waste while ensuring the quality of user experience.

[0052] The method for traffic prediction provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0053] Figure 1 This is a schematic flowchart of a traffic prediction method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method may include steps S110 to S130.

[0054] First, let's introduce step S110.

[0055] S110, using a preset target prediction model, predicts the traffic usage of the target cell in the first preset time period, and obtains the first predicted value of the target traffic index in the first preset time period, wherein there are multiple target traffic indices.

[0056] The target prediction model refers to a pre-trained prediction model that can be used to predict the traffic usage of a target cell within a first preset time period.

[0057] The target community refers to the residential community covered by the base station.

[0058] The first preset time period refers to a period of time in the future, such as one week, two weeks, or one month. The duration of the first preset time period can be set according to the user's needs.

[0059] The first predicted value refers to the predicted value of the target traffic indicator within a first preset time period. If there are multiple target traffic indicators, each target traffic indicator corresponds to a first predicted value. For example, if the target traffic indicator is mobile network traffic, the first predicted value of mobile network traffic within the first preset time period is 10G. Another example is resource utilization, where the first predicted value of resource utilization within the first preset time period is 70%.

[0060] Specifically, step S110 can be understood as using a pre-trained target prediction model to predict the traffic usage of the target cell covered by the target base station in the first preset time period, thereby obtaining the traffic usage of the target cell in the first preset time period, which is also the first predicted value of each target traffic index in the first preset time period.

[0061] In some embodiments, the target traffic metric includes at least two of mobile network traffic values, resource utilization, and number of user connections.

[0062] Among them, the mobile network traffic value refers to the mobile network traffic value of the target cell.

[0063] Resource utilization rate refers to the percentage of resources actually used by users relative to the total available resource flow.

[0064] User connection count refers to the number of users connected to the base station in the target cell.

[0065] Specifically, the target traffic indicators may include at least two of the following: mobile network traffic value, resource utilization rate, and number of user connections. Furthermore, the inventors have discovered that resource utilization rate and number of user connections are characteristics closely related to traffic, reflecting the essential changing patterns of network traffic and effectively indicating traffic trends. These indicators are highly valuable for expanding or reducing the capacity of base station cell communication equipment. Therefore, when predicting traffic usage in a target cell during a first preset time period, multiple indicators, such as mobile network traffic value, resource utilization rate, and number of user connections, can be predicted. Based on these multiple indicators, the traffic trend of the target cell can be better understood, enabling accurate expansion or reduction of base station cell communication equipment.

[0066] Next, we will introduce step S120.

[0067] S120, in the preset correspondence between flow indicators and prediction error values, obtain the target prediction error value corresponding to the target flow indicator, wherein the prediction error value is the error value of the target prediction model in predicting the flow indicator.

[0068] The prediction error value refers to the prediction error value of the trained target prediction model when predicting each target flow indicator. The prediction error value of the target prediction model is different when predicting different target flow indicators, so a correspondence between flow indicators and prediction error values ​​will be generated.

[0069] Specifically, step S120 can be understood as obtaining the target prediction error value corresponding to the target flow index from the preset correspondence between the flow index and the prediction error value of the target prediction model. Each target flow index corresponds to a target prediction error value.

[0070] Next, we will introduce step S130.

[0071] S130, based on the first predicted value and the target prediction error value, determine the first target predicted value of the target flow index within the first preset time period.

[0072] In this context, the first predicted value and the target prediction error value are both the predicted value and the prediction error value corresponding to the same target flow index. In other words, target flow index_1 corresponds to (first predicted value_1, target prediction error_1), target flow index_2 corresponds to (first predicted value_2, target prediction error_2), target flow index_3 corresponds to (first predicted value_3, target prediction error_3), and so on.

[0073] The inventors discovered that because the target prediction model has different prediction error values ​​when predicting different target flow indicators, that is, the prediction accuracy of different flow indicators is different, the predicted value and error value corresponding to each target flow indicator can be combined separately to make up for the error of each target flow indicator and improve the prediction accuracy of each target flow indicator.

[0074] The first target predicted value refers to the final predicted value after combining the predicted value corresponding to the target flow indicator with the error value. Each target flow indicator corresponds to a final first target predicted value.

[0075] Specifically, step S130 can be understood as determining the first target predicted value of the target traffic indicator within a first preset time period based on the first predicted value and the target prediction error value corresponding to the target traffic indicator, i.e., the predicted value and prediction error value corresponding to the same target traffic indicator. For example, if the target traffic indicator is mobile network traffic value, and the first predicted value of the mobile network traffic value within the first preset time period is 200G, and the prediction error value of the target prediction model when predicting the mobile network traffic value is -2% (i.e., the first predicted value is too low), then the final predicted value of the mobile network traffic value can be determined to be 200 × (1 + 2%) = 204G, that is, the first target predicted value of the mobile network traffic value within the first preset time period is 204G. As another example, if the target traffic indicator is resource utilization rate, and the first predicted value of the resource utilization rate within the first preset time period is 65%, and the prediction error value of the target prediction model when predicting the resource utilization rate is +3% (i.e., the first predicted value is too high), then the final predicted value of the resource utilization rate can be determined to be 65% × (1 - 3%) = 63%, that is, the first target predicted value of the resource utilization rate within the first preset time period is 63%.

[0076] According to the traffic prediction method provided in this application embodiment, by utilizing a preset target prediction model, the traffic usage of a target cell within a first preset time period can be predicted, thereby obtaining a first predicted value of the target traffic index within the first preset time period. Then, in the preset correspondence between traffic index and prediction error value, the target prediction error value corresponding to the target traffic index can be obtained, wherein the prediction error value is the error value of the target prediction model in predicting the traffic index. Next, based on the first predicted value of the target traffic index and the corresponding target prediction error value, the first target predicted value of the target traffic index within the first preset time period can be determined, which is equivalent to compensating for and correcting the error of the target traffic index, thereby improving the prediction accuracy of each target traffic index of the model. It can accurately predict the traffic usage of the base station cell, enabling network equipment providers to accurately grasp the traffic trend of the base station cell and accurately scale up and down the communication equipment of the base station cell, thereby ensuring the quality of user experience while avoiding resource waste.

[0077] The above describes the specific implementation method of using a pre-trained target prediction model in steps S110 to S130 to determine the first target prediction value of the target flow index within the first preset time period. Next, we will introduce the specific training process of the target prediction model before using the target prediction model in step S110.

[0078] In some embodiments, such as Figure 2 As shown, before step S110, which uses a preset target prediction model to predict traffic usage within a first preset time period and obtains the first predicted value of the target traffic index within the first preset time period, the method further includes:

[0079] S141, obtain the true value of the target traffic indicator in the first historical period and the true value in the second target preset period, wherein there are multiple first historical periods, and multiple first historical periods constitute the first target historical period;

[0080] S142, using a preset initial prediction model, based on the real value in the first historical time period, predict the traffic usage of the target cell in the second preset time period, and obtain the second predicted value of the target traffic index in the second preset time period. There are multiple second preset time periods, and multiple second preset time periods constitute the second target preset time period.

[0081] S143, add the second predicted values ​​of the target flow index in multiple second preset time periods to obtain the third predicted value of the target flow index in the second target preset time period;

[0082] S144. Based on the actual value and the third predicted value of the target flow index within the second target preset time period, the initial prediction model is corrected to obtain the target prediction model.

[0083] The first target historical period can include multiple consecutive first historical periods, and the duration of each first historical period can be set according to requirements. For example, if the first target historical period consists of 10 first historical periods, then "First Target Historical Period = First Historical Period_1 + First Historical Period_2 + First Historical Period_3 + ... + First Historical Period_10". For example, if the first target historical period is from January to November 2033, and each first historical period is one month long, then the first target historical period = January + February + ... + November.

[0084] The true value of the target traffic indicator in the first historical period means that the target traffic indicator has a corresponding true value in each first historical period. It is a known value that reflects the actual historical usage of the target traffic indicator.

[0085] For example, the first target historical period is from January to November 2033, with each first historical period lasting one month. The target traffic indicator is the traffic value of the mobile network. The actual values ​​of the target traffic indicator of the target cell in multiple first historical periods are as follows: 300G actual usage in January 2033, 203G actual usage in February 2033, 120G actual usage in March 2033, ..., 55G actual usage in November 2033.

[0086] The second target preset period can include multiple consecutive second preset periods, and the duration of each second preset period can be set according to needs. For example, if the second target preset period consists of 10 second preset periods, then "Second target preset period = Second preset period_1 + Second preset period_2 + Second preset period_3 + ... + Second preset period_10". For example, if the second target preset period is from February to December 2033, and each second preset period is one month long, then the second target preset period = February + March + ... + December.

[0087] The true value within the second target preset time period refers to the historical actual value of the target traffic index within the second target preset time period. It is a known total value used as the result label for supervised training of the initial prediction model, and is used to correct the parameters of the initial prediction model to obtain more accurate prediction results.

[0088] For example, the second target preset period is from February to December 2033, with each second preset period lasting one month. Using the initial prediction model based on the actual values ​​of the first historical period, the target flow indicators for each second preset period are predicted, resulting in multiple predicted values ​​(i.e., multiple second predicted values): 150G is expected to be used in February 2033, 100G in March 2033, ..., 50G in December 2033. Adding these multiple second predicted values, the total predicted value for the second target preset period from February to December 2033 is 150 + 100 + ... + 50 = 900G. This 900G is the total predicted value for February to December 2033 obtained by the initial prediction model based on the actual values ​​of multiple first historical periods from January to November 2033. At this point, the "actual value within the second target preset period" is obtained as 820G. Therefore, the error between 900G and 820G can be calculated, and the parameters of the initial prediction model can be corrected. The revised initial prediction model is based on the "true value within the first historical period" for prediction (predicting multiple short periods (i.e., the second preset period) to obtain multiple sub-predicted values ​​(i.e., the second predicted value)). The resulting predicted value (here, the predicted value is the sum of multiple sub-predicted values, i.e., the third predicted value) is compared with the "true value within the second target preset period." The resulting error can be used to revise the initial model. After revision, multiple rounds of "prediction and revision" can be performed using the above method until the prediction error value of each target flow indicator is less than a preset threshold. At this point, the prediction model is considered to meet the requirements, and the target prediction model is obtained and can be put into use.

[0089] Specifically, steps S141 to S144 can be understood as follows: obtaining multiple real values ​​of the target traffic indicator within multiple first historical time periods to predict the second predicted value of the target traffic indicator within multiple second preset time periods; simultaneously obtaining the total real value within the second target preset time period to compare with the total predicted value (i.e., the third predicted value) within the second target preset time period to calculate the error; then, using a preset initial prediction model, based on the real values ​​within the first historical time periods, predicting the traffic usage of the target cell within the second preset time period to obtain multiple second predicted values ​​of the target traffic indicator within multiple second preset time periods; and then adding the multiple second predicted values ​​to obtain the third predicted value of the target traffic indicator within the second target preset time period. Thus, the real value and the third predicted value of the target traffic indicator within the second target preset time period can be compared to calculate the prediction error of the initial prediction model, and the initial prediction model can be corrected until the prediction error is less than a preset threshold to obtain the target prediction model. In this embodiment of the application, by using an initial model to predict the index values ​​of multiple consecutive "short time periods" and then adding them together, a total index value for a "long time period" can be predicted. The predicted total index value is then compared with the actual value within the "long time period" to obtain the error and correct the original model. This process is repeated until the prediction error is less than a preset threshold, thereby training a target prediction model with higher prediction accuracy.

[0090] The above describes the specific implementation method of using a pre-trained target prediction model in steps S110 to S130 to determine the first target prediction value of the target flow index within the first preset time period. It also describes the specific training process of training the initial prediction model to obtain the target prediction model in steps S141 to S144 before using the target prediction model in step S110. Next, the preprocessing process of the sample data used for training will be introduced before training the initial prediction model in step S141.

[0091] The preprocessing of sample data can include anomaly detection and replacement with normal values, data augmentation, and data normalization. The following section describes the data normalization process, which can assist model training and accelerate its speed and accuracy.

[0092] In some embodiments, the true value of the target traffic indicator in the first historical period and the true value in the second target preset period are both greater than or equal to 0 and less than or equal to 1; the method further includes:

[0093] The first target predicted value of the target flow index within the first preset time period is inversely normalized to obtain the second target predicted value of the target flow index within the first preset time period.

[0094] The fact that the true value of the target flow index in the first historical period and the true value in the second target preset period are both greater than or equal to 0 and less than or equal to 1 means that after normalizing the true value of the target flow index in the first historical period and the true value in the second target preset period, the true value of the target flow index in the first historical period and the true value in the second target preset period are both in the range of [0, 1].

[0095] Specifically, by performing inverse normalization on the first target predicted value of the target traffic indicator within the first preset time period, a second target predicted value of the target traffic indicator within the first preset time period can be obtained. For example, when predicting the traffic indicator based on the trained target prediction model, the first target predicted value is also in the range of [0, 1]. In this case, inverse normalization can be performed. For example, if the target traffic indicator is the number of user connections, the first target predicted value corresponding to the number of user connections is 0.3. After inverse normalization, it becomes 300 people. The normalized value is used for model training, which results in faster training speed and better training effect. Performing inverse normalization after obtaining the predicted value makes it more intuitive.

[0096] The above introduced the model-assisted training methods of normalization and denormalization. Next, we will introduce the process of expanding the sample data before normalization.

[0097] In some embodiments, such as Figure 3 As shown, there are M first historical time periods, where M is a positive integer; before obtaining the true value of the target flow index in the first historical time period and the true value in the second target preset time period in step S141, the method further includes:

[0098] S151, obtain N real values ​​of the target traffic indicator within the second target historical period, where the second target historical period includes N first historical periods, N is a positive integer, and N is less than M;

[0099] S152, based on preset expansion rules, generates the actual values ​​of the target traffic metric within M first historical time periods.

[0100] The second target historical period is the historical period before expansion. The N real values ​​in the second target historical period are the actual usage values ​​of users in the target cell. Directly using them as sample data for model training is insufficient to support the model training process. Therefore, it is necessary to expand them according to the preset expansion rules.

[0101] Specifically, steps S151 and S152 can be understood as follows: First, obtain the N true values ​​of the target traffic indicator corresponding to the N first historical periods within the second target historical period. Then, based on the preset expansion rules, generate the true values ​​of the target traffic indicator within M first historical periods, where M > N. Before expansion, there are N true values: the second target historical period = N first historical periods; after expansion, there are M true values: the first target historical period = M first historical periods, where M > N. That is, expanding the N first historical periods to obtain M first historical periods, expanding from N true values ​​to M true values, thereby generating the true values ​​of the target traffic indicator within M first historical periods. This increases the amount of sample data, which can support multiple training processes of the model and is beneficial to improving the accuracy of model training.

[0102] The previous section introduced the process of expanding the sample data in steps S151 and S152, and the normalization process after expansion. Next, we will introduce how to perform anomaly detection on the sample data and replace it with normal values ​​before expanding the sample data in step S151.

[0103] In some embodiments, such as Figure 4 As shown, before obtaining N real values ​​of the target traffic index within the second target historical time period in step S151, the method further includes:

[0104] S161, Obtain N raw values ​​of the target flow index within the second target historical time period;

[0105] S162, perform anomaly detection on the original value according to the preset anomaly detection rules to obtain the target detection result, wherein the target detection result includes anomalies in the original value;

[0106] S163, if the target detection result is abnormal, the abnormal original value is replaced with a normal value according to the preset interpolation rule to obtain N true values ​​of the target flow index in the second target historical period.

[0107] The raw value refers to the original data corresponding to the target traffic indicators in the second target historical period of the target cell, which is directly collected from the operator's server. The raw data often has problems such as missing data points, abnormal data points, and non-standard data formats.

[0108] Specifically, steps S161 to S163 can be understood as follows: before expanding the sample data, firstly, N original values ​​of the target traffic index within the second target historical period are obtained. Then, anomaly detection is performed on the original values ​​according to the preset anomaly detection rules to obtain the target detection results. The target detection results include original values ​​that are abnormal and original values ​​that are not abnormal. Next, if the target detection result is that the original value is abnormal, the abnormal original value can be replaced with a normal value according to the preset interpolation rules, thereby obtaining N real values ​​of the target traffic index within the second target historical period. This can solve the problem of abnormal data points in the original data directly collected from the operator's server.

[0109] In some embodiments, there are multiple anomaly detection rules; step S162 performs anomaly detection on the original value according to the preset anomaly detection rules to obtain the target detection result, including:

[0110] Anomaly detection is performed on the original value according to multiple anomaly detection rules, resulting in multiple sub-detection results;

[0111] The number of target sub-detection results is compared with the preset number to obtain the target detection results, where the target sub-detection results are the sub-detection results of the original values ​​that are abnormal.

[0112] Among them, the sub-detection result refers to the detection result that can be obtained by performing anomaly detection on the original value according to multiple anomaly detection rules.

[0113] The target sub-detection result refers to the sub-detection result where the original value is an outlier.

[0114] The preset quantity refers to the number of target sub-detection results corresponding to the original value. If this preset quantity is exceeded, the original value is determined to be an outlier. For example, if there are 3 outlier detection rules, the preset quantity is 2, and the sub-detection results obtained for the original value _1 are: original value outlier, original value not outlier, and original value outlier, then there are 2 results for "original value outlier" that are less than or equal to the preset quantity. Therefore, the original value _1 can be determined to be an outlier, which is equivalent to using a voting method to determine the outlier of the original value.

[0115] Specifically, step S162, which involves performing anomaly detection on the original value according to preset anomaly detection rules to obtain the target detection result, may include: performing anomaly detection on the original value according to multiple anomaly detection rules, resulting in multiple sub-detection results, the number of which is the same as the number of anomaly detection rules; then counting the number of target sub-detection results indicating that the original value is an anomaly, and comparing this number with a preset number; when the number of target sub-detection results is greater than or equal to the preset number, the target detection result is "original value is anomaly"; when the number of target sub-detection results is less than the preset number, the target detection result is "original value is not anomaly". For example, if there are 3 anomaly detection methods, there will be 3 corresponding results. If 2 or more detection results are anomaly, it indicates that the original value is an anomaly. Using a voting method to determine the final target detection result can more accurately determine whether the original value is anomaly, ensuring the authenticity and accuracy of the sample data.

[0116] In one example, the preprocessing of sample data can include anomaly detection and replacement with normal values, sample data augmentation, and sample data normalization. After obtaining the original values ​​of the sample data, it is determined whether they are blank or outliers, for example, by using a voting algorithm to detect outliers. If the original values ​​are outliers or blank, they are replaced with normal values, for example, by using an interpolation algorithm. Next, the sample data is augmented by expanding the time period; for example, if there are N time periods before augmentation and M time periods after augmentation, M > N. Then, the sample data is normalized to ensure that the sample data ∈ [0, 1]. This facilitates subsequent training of the original prediction model based on the obtained sample data, resulting in the target prediction model. The higher the quantity and accuracy of the sample data, the higher the prediction accuracy of the trained target prediction model. Therefore, the preprocessing steps of anomaly detection and replacement with normal values, sample data augmentation, and sample data normalization can significantly improve the speed and accuracy of model training.

[0117] In one example, such as Figure 5 As shown, the training and use process of the target prediction model may specifically include steps S10 to S80.

[0118] S10, Obtain the raw sample data.

[0119] Obtain N raw values ​​of the target traffic indicator within the second target historical period. The second target historical period consists of N first historical periods, and each first historical period corresponds to one raw value of the target traffic indicator.

[0120] For example, the statistics of cell traffic data and related information of a certain base station cell collected from the operator are shown in Table 1.

[0121] Table 1

[0122]

[0123] In Table 1, the explicit feature "Mobile Network Traffic Value" and the implicit features "PRB Resource Utilization Rate" and "User Connections RRC" are closely related to traffic trends, and therefore can be used as target traffic indicators. Thus, we can obtain N original values ​​of "Mobile Network Traffic Value, PRB Resource Utilization Rate, and User Connections RRC" in the second target historical period through Table 1.

[0124] S20, Outlier Identification.

[0125] If the original value is missing (i.e., the field has no data), a preset interpolation algorithm can be used to fill it with a normal value. If the original value is not missing, it can be further checked for outliers. This involves performing anomaly detection on the original value according to preset anomaly detection rules. If the target detection result indicates that the original value is an outlier, the outlier original value is replaced with a normal value according to the preset interpolation rules. This allows us to obtain N true values ​​of the target traffic indicator within the second target historical time period.

[0126] In one example, a voting method is used to identify outliers, and cubic spline interpolation is used to replace blank and outlier points with normal values. Outlier detection, such as... Figure 6 As shown, year-on-year detection, short-term month-on-month detection, and density-based detection can be used to detect excessively abnormal data points. Since each of these three methods has its advantages and disadvantages in anomaly detection, a voting rule can be adopted for detection. The voting rule can balance the processing results of the three anomaly detection methods mentioned above; that is, a point is determined to be an anomaly only if two or more anomaly detection methods determine that the point is abnormal. Data interpolation: To achieve predictive effects, the continuity of data needs to be maintained. Therefore, after detecting existing blank value points and anomalies, data interpolation can be performed on these points. For example, the cubic spline interpolation method can be used to obtain the values ​​of fill points that are infinitely close to the surrounding existing data points and maintain the smoothness of the line segments through piecewise polynomials. This solves the problems of missing data points and abnormal data points in community mobile data traffic information directly collected from the operator's server, i.e., Table 1.

[0127] In another example, after addressing issues such as missing and outlier data points in Table 1, data standardization can be performed. This transforms the dataset into a standard data format suitable for model training, mapping the data to the [0,1] interval. For example, the min-max method can be used for normalization, with the following formula:

[0128] (1)

[0129] in, These are the values ​​before normalization. It is the largest value in the data. It is the smallest value in the data. These are normalized values. Data normalization can resolve issues such as missing or abnormal data points in community mobile data traffic information directly collected from operator servers (i.e., Table 1), as well as the problem of non-standard data formats.

[0130] S30 augments the sample dataset.

[0131] By expanding N first historical time periods to obtain M first historical time periods, and expanding from N true values ​​to M true values, the true values ​​of the target traffic indicators within the M first historical time periods can be generated. This increases the amount of sample data, which can support multiple training processes of the model and is conducive to improving the accuracy of model training.

[0132] For example, generative adversarial networks (GANs) can be used to augment sample datasets. During training, the data is reconstructed into an H×W image as input. The specific steps are as follows:

[0133] 1) Initialize the parameters of the generator and discriminator networks;

[0134] 2) The generator uses normally randomly distributed noise vectors to generate base station traffic data. The generator is fixed, and the discriminator is trained to distinguish between real and fake data as much as possible.

[0135] 3) After training the discriminator in a loop, train the generator so that the discriminator cannot distinguish between true and false data, that is, when the discrimination probability is 0.5, stop training. The augmented base station traffic data can be obtained, that is, the M true values ​​corresponding to the M first historical time periods can be obtained.

[0136] S40, M first historical periods.

[0137] After augmenting the sample dataset, we can obtain M true values ​​corresponding to the M first historical time periods.

[0138] S50, using the initial prediction model, predict the target flow indicators within the second preset time period to obtain the second predicted value. For example, predict the target flow indicators within the second preset time period _1 to obtain the second predicted value _1, predict the target flow indicators within the second preset time period _2 to obtain the second predicted value _2, and so on, until the second preset time period _M to obtain the second predicted value _M.

[0139] In one example, base station traffic data that has undergone data preprocessing and data augmentation (M real values ​​corresponding to M first historical time periods) are selected for empirical mode decomposition, which decomposes several base station traffic data sequences into IMF components Ci(t)={Ci(t1),Ci(t2)…Ci(tm)}, where Ci(tm) represents the component value at time n on the i-th IMF component sequence Ci(t), thus obtaining the M real values ​​{Ci(t1),Ci(t2)…Ci(tm)} corresponding to the M first historical time periods of the explicit feature "mobile network traffic value C".

[0140] Then, the user connection count (RRC) and PRB resource utilization rate, which are closely related to traffic at the corresponding time, are selected as input features. It is assumed that the M true value sequences of the RRC connection count in the M first historical time periods are Ri(t)={Ri(t1),Ri(t2)…Ri(tm)}, and the M true value sequences of the PRB resource utilization rate in the M first historical time periods are Pi(t)={Pi(t1),Pi(t2)…Pi(tm)}.

[0141] Feature selection is performed using the Pearson coefficient, and the specific formula is as follows:

[0142] (2)

[0143] in, and This represents the average of the sample.

[0144] Next, the input features are combined with the traffic decomposition sequence, and the three together form the input sequence of LSTM: LSTMi(t)={(Ci(t1),Ri(t1),Pi(t1)), (Ci(t2),Ri(t2),Pi(t2)), …, (Ci(tm),Ri(tm),Pi(tm))}. That is, the "mobile network traffic value, user connection count RRC, PRB resource utilization" corresponding to the first first historical period is (Ci(t1),Ri(t1),Pi(t1)), the "mobile network traffic value, user connection count RRC, PRB resource utilization" corresponding to the first first historical period is (Ci(t2),Ri(t2),Pi(t2)), ..., and the "mobile network traffic value, user connection count RRC, PRB resource utilization" corresponding to the Mth first historical period is (Ci(tm),Ri(tm),Pi(tm)).

[0145] Then, based on LSTMi(t)={(Ci(t1),Ri(t1),Pi(t1)), (Ci(t2),Ri(t2),Pi(t2)), …, (Ci(tm),Ri(tm),Pi(tm))}, multiple second predicted values ​​within the second preset time period are obtained. According to the characteristics of the LSTMi(t) input sequence, the network structure includes an input layer, an LSTM layer, a fully connected layer, and a regression layer in sequence. Based on this network structure, the Adam algorithm is used to iteratively learn the input sequence. The input sequence is input into each LSTM network for training. A Drop layer is added in the middle to prevent overfitting. The ReLU function is used as the activation function to prevent gradient vanishing. Thus, the predicted values ​​of each LSTM network are obtained, that is, the second predicted values ​​within the second preset time period are obtained. There are M preset time periods and M historical time periods. In terms of time sequence, the end time of each preset time period is ∆t longer than the end time of the corresponding historical time period. The smaller ∆t is, the more accurate the prediction is. The larger ∆t is, the less accurate the prediction is. ∆t can be set according to needs. For example, ∆t can be 1 hour, 6 hours, 12 hours, 1 day, etc.

[0146] S60, determine the third predicted value of the target flow rate indicator.

[0147] The second predicted values ​​of the target flow index within multiple second preset time periods are added together to obtain the third predicted value of the target flow index within the second target preset time period.

[0148] For example, the EMD-PPCS algorithm is used to decompose and reconstruct the augmented base station traffic dataset, and co-variables closely related to the traffic sequence are selected, such as multiple target traffic indicators like mobile network traffic values, resource utilization, and the number of user connections. The second predicted value, formed by the decomposition, reconstruction, and selection using the EMD-PPCS algorithm, is then used to construct the LSTM network, and the results are accumulated to form the third predicted value for the LSTM network.

[0149] S70, calculate the error value based on the true value.

[0150] By comparing the actual value and the third predicted value of the target flow index within the second target preset time period, the prediction error of the initial prediction model is calculated, so as to correct the initial prediction model.

[0151] In one example, SVR is used to perform regression prediction on the prediction error value of the LSTM network, and the parameters of SVR are optimized by the PSO algorithm to obtain the corrected prediction error value et < preset threshold.

[0152] Specifically, when using SVR to perform regression prediction on the prediction error value, the steps of simultaneously optimizing the parameters of SVR using the PSO algorithm, i.e., optimizing the parameters of the prediction model when the prediction error value is large, are as follows:

[0153] Initialize the particle swarm Determine its initial value and range of variation, determine the population size, and the maximum number of generations. The individual extreme value at the current moment is denoted as... The global extremum is denoted as ;

[0154] The fitness value of the training samples is calculated using the mean squared error. The particle with the best fitness value is selected as the initial global extremum. The specific formula is as follows:

[0155] (3)

[0156] in, This is an estimate of the new sample;

[0157] Evolutionary calculations are performed on the particle's velocity to update the particle's velocity and position;

[0158] Continue training SVR, calculate the fitness value of the samples, and update the particles based on the fitness value. and If a particle's current fitness is better than ,but Replaced by the current position;

[0159] The updated and Compare and update if superior. Otherwise, keep it;

[0160] Determine if the algorithm termination condition is met. If the maximum number of iterations is reached or the solution no longer changes, stop the iteration and output the optimal solution, i.e., the corrected prediction error value et < the preset threshold. The algorithm ends. Otherwise, continue to repeat the training and execute steps S50 to S80 in a loop.

[0161] S80, target prediction model.

[0162] Repeat steps S50 to S80 until the prediction error is less than the preset threshold to obtain the target prediction model.

[0163] In one example, when using the target prediction model to predict the traffic usage of the target cell in the first preset time period, the first predicted value of the target traffic index in the first preset time period by the LSTM network, i.e. the target prediction model, is added to the prediction error value et after PSO optimization and SVR correction, which is the prediction result of the final combined model, i.e. the first target prediction value yr=yp+et. Finally, the prediction result yr can be denormalized to obtain the final prediction value, i.e. the second target prediction value. The target prediction model can predict the traffic usage of a target cell within a first preset time period, thus obtaining the first predicted value of the target traffic indicator within the first preset time period. Then, in the preset correspondence between traffic indicators and prediction error values, the target prediction error value corresponding to the target traffic indicator can be obtained. Next, based on the first predicted value of the target traffic indicator and the corresponding target prediction error value, the first target predicted value of the target traffic indicator within the first preset time period can be determined, which is equivalent to compensating for the error of the target traffic indicator, thereby improving the prediction accuracy of each target traffic indicator in the model. It can accurately predict the traffic usage of base station cells, enabling network equipment providers to accurately grasp the traffic trends of base station cells and accurately scale up or down the communication equipment of base station cells, thereby ensuring the quality of user experience while avoiding resource waste.

[0164] In this embodiment, a base station traffic prediction method based on EMD-PCCS and LSTM-SVR models is used. This method collects traffic data (mobile network traffic values) and corresponding base station data information (number of user connections, resource usage, etc.), and preprocesses the data (detecting blank points and outliers and filling in normal values) to form a base station traffic dataset. A generative adversarial network is used to augment the base station traffic dataset. The EMD-PCCS algorithm is then used to decompose and reconstruct the base station traffic dataset, and features closely related to traffic (target traffic indicators: mobile network traffic values, number of user connections, resource usage, etc.) are selected as input sequences for the LSTM network. The method then targets the EMD-PCCS-... The PCCS algorithm decomposes, reconstructs, and filters the input sequence to construct an LSTM network, obtaining the LSTM network's predicted values ​​(the second predicted value of each target traffic indicator within multiple second preset time periods). The predicted values ​​are then subtracted from the original traffic data to obtain the prediction error term. This error term is input into the SVR model for regression prediction, while the PSO algorithm is used to optimize the SVR model's parameters, resulting in a new set of corrected error terms. This process continues until the error term is less than a preset threshold, yielding the trained target prediction model, which is then put into use. The predicted value from the target prediction model is added to the corrected error term to obtain the final prediction result, i.e., the first target predicted value. By employing multiple data preprocessing methods, iteratively training the prediction model with multiple models, and repeatedly using the initial prediction model during training to predict the values ​​of multiple target traffic indicators within multiple second preset time periods, this approach addresses the problem that traditional linear time series methods struggle to effectively capture the complex nonlinear factors in actual base station traffic sequences, and that using only a single model yields insufficient prediction results for base station traffic.

[0165] Based on the same inventive concept, embodiments of this application also provide a device for traffic flow prediction, such as... Figure 7 As shown, the device 700 may include a prediction module 710, an acquisition module 720, and a determination module 730:

[0166] The prediction module 710 is used to predict the traffic usage of the target cell in the first preset time period using a preset target prediction model, and to obtain the first predicted value of the target traffic index in the first preset time period. There are multiple target traffic indices.

[0167] The acquisition module 720 is used to acquire the target prediction error value corresponding to the target flow index from the preset correspondence between flow index and prediction error value, wherein the prediction error value is the error value of the target prediction model in predicting the flow index.

[0168] The determination module 730 is used to determine the first target predicted value of the target flow index within a first preset time period based on the first predicted value and the target prediction error value.

[0169] According to the traffic prediction apparatus provided in this application embodiment, the apparatus utilizes a preset target prediction model to predict the traffic usage of a target cell within a first preset time period, thereby obtaining a first predicted value of the target traffic index within the first preset time period. Then, in a preset correspondence between traffic indexes and prediction error values, the target prediction error value corresponding to the target traffic index can be obtained, where the prediction error value is the error value of the target prediction model in predicting the traffic index. Next, based on the first predicted value of the target traffic index and the corresponding target prediction error value, the first target predicted value of the target traffic index within the first preset time period can be determined, which is equivalent to compensating for the error of the target traffic index, thereby improving the prediction accuracy of each target traffic index of the model. This enables accurate prediction of the traffic usage of base station cells, allowing network equipment providers to accurately grasp the traffic trends of base station cells and accurately scale up or down the communication equipment of base station cells, thereby ensuring the quality of user experience while avoiding resource waste.

[0170] In some embodiments, the target traffic metric includes at least two of mobile network traffic values, resource utilization, and number of user connections.

[0171] In some embodiments, before the prediction module uses a preset target prediction model to predict traffic usage within a first preset time period and obtains a first predicted value of the target traffic index within the first preset time period, the device further includes a calculation module and a correction module:

[0172] The acquisition module is also used to acquire the real value of the target traffic indicator in the first historical period and the real value in the second target preset period. There are multiple first historical periods, and multiple first historical periods constitute the first target historical period.

[0173] The prediction module is also used to predict the traffic usage of the target cell in the second preset time period based on the real value in the first historical time period using a preset initial prediction model, and to obtain the second predicted value of the target traffic index in the second preset time period. There are multiple second preset time periods, and multiple second preset time periods constitute the second target preset time period.

[0174] The calculation module is used to add the second predicted values ​​of the target flow index in multiple second preset time periods to obtain the third predicted value of the target flow index in the second target preset time period.

[0175] The correction module is used to correct the initial prediction model based on the actual value and the third predicted value of the target flow index within the second target preset time period, so as to obtain the target prediction model.

[0176] In some embodiments, the true value of the target flow index in the first historical period and the true value in the second target preset period are both greater than or equal to 0 and less than or equal to 1; the device further includes a normalization module:

[0177] The normalization module is used to perform inverse normalization on the first target predicted value of the target flow index within the first preset time period to obtain the second target predicted value of the target flow index within the first preset time period.

[0178] In some embodiments, there are M first historical time periods, where M is a positive integer; before the acquisition module acquires the true value of the target traffic indicator within the first historical time period and the true value within the second target preset time period, the device further includes a generation module:

[0179] The acquisition module is also used to acquire N real values ​​of the target traffic indicator within the second target historical period, wherein the second target historical period includes N first historical periods, N is a positive integer, and N is less than M;

[0180] The generation module is used to generate the actual values ​​of the target traffic metric within M first historical time periods based on preset expansion rules.

[0181] In some embodiments, before the acquisition module acquires N real values ​​of the target traffic indicator within a second target historical time period, the device further includes a detection module and a replacement module:

[0182] The acquisition module is also used to acquire N raw values ​​of the target traffic metric within the second target historical time period;

[0183] The detection module is used to perform anomaly detection on the original value according to the preset anomaly detection rules and obtain the target detection result, wherein the target detection result includes anomalies in the original value;

[0184] The replacement module is used to replace the abnormal original value with a normal value according to a preset interpolation rule when the target detection result is abnormal, so as to obtain N true values ​​of the target flow index in the second target historical period.

[0185] In some embodiments, there are multiple anomaly detection rules; the detection module is used to perform anomaly detection on the original value according to the preset anomaly detection rules to obtain the target detection result, including:

[0186] The detection unit is used to perform anomaly detection on the original value according to multiple anomaly detection rules, and obtain multiple sub-detection results;

[0187] The comparison unit is used to compare the number of target sub-detection results with a preset number to obtain the target detection result, wherein the target sub-detection result is the sub-detection result with an abnormal original value.

[0188] The various modules in the traffic prediction device provided in this application embodiment can achieve... Figures 1 to 6 The functions of each step in the provided traffic prediction method, and the corresponding technical effects they achieve, will not be elaborated here for the sake of brevity.

[0189] Figure 8 A schematic diagram of the hardware structure of the traffic prediction device provided in an embodiment of this application is shown.

[0190] The device for flow prediction may include a processor 801 and a memory 802 storing computer program instructions.

[0191] Specifically, the processor 801 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.

[0192] Memory 802 may include mass storage for data or instructions. For example, and not limitingly, memory 802 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 802 may include removable or non-removable (or fixed) media. Where appropriate, memory 802 may be internal or external to a flow prediction device. In a particular embodiment, memory 802 is a non-volatile solid-state memory.

[0193] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the methods according to one aspect of this disclosure.

[0194] The processor 801 implements any of the traffic prediction methods in the above embodiments by reading and executing computer program instructions stored in the memory 802.

[0195] In one example, the traffic prediction device may also include a communication interface 803 and a bus 804. For example, Figure 8 As shown, the processor 801, memory 802, and communication interface 803 are connected through bus 804 and complete communication with each other.

[0196] The communication interface 803 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0197] Bus 804 includes hardware, software, or both, that couples components of a flow-predicting device together. For example, and not limited to, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Linear Predictive Coding (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (Peripheral Component Interconnect-X, PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VESA Local Bus, VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 804 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 interconnection.

[0198] This device can execute the flow prediction method in the embodiments of this application based on the various units / components in the flow prediction apparatus, thereby achieving a combination Figures 1 to 6 The method described is for traffic prediction.

[0199] Furthermore, in conjunction with the traffic prediction methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the traffic prediction methods in the above embodiments.

[0200] This application also provides a computer program product in which the instructions, when executed by a processor of an electronic device, cause the electronic device to perform various processes that implement any of the above-described methods for traffic prediction.

[0201] 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.

[0202] 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, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), 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.

[0203] 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.

[0204] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0205] The above are merely specific embodiments 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 traffic flow prediction, characterized in that, include: Using a preset target prediction model, the traffic usage of the target cell in the first preset time period is predicted to obtain the first predicted value of the target traffic index in the first preset time period, wherein there are multiple target traffic indices. In the preset correspondence between traffic indicators and prediction error values, the target prediction error value corresponding to the target traffic indicator is obtained, wherein the prediction error value is the error value of the target prediction model in predicting the traffic indicator, and the prediction error value is less than a preset threshold. Based on the first predicted value and the target prediction error value, determine the first target predicted value of the target flow index within the first preset time period; The target traffic metrics include at least two of the following: mobile network traffic value, resource utilization rate, and number of user connections.

2. The method according to claim 1, characterized in that, Before using a preset target prediction model to predict traffic usage within a first preset time period and obtain a first predicted value of the target traffic index within the first preset time period, the method further includes: Obtain the true value of the target traffic indicator within a first historical period and the true value within a second target preset period, wherein there are multiple first historical periods, and multiple first historical periods constitute a first target historical period; Using a preset initial prediction model, based on the real values ​​within the first historical time period, the traffic usage of the target cell within a second preset time period is predicted to obtain a second predicted value of the target traffic index within the second preset time period. There are multiple second preset time periods, and the multiple second preset time periods constitute the second target preset time period. The second predicted values ​​of the target traffic indicator in multiple second preset time periods are added together to obtain the third predicted value of the target traffic indicator in the second target preset time period; Based on the actual value of the target traffic indicator within the second target preset time period and the third predicted value, the initial prediction model is corrected to obtain the target prediction model.

3. The method according to claim 2, characterized in that, The actual value of the target flow index in the first historical period and the actual value in the second target preset period are both greater than or equal to 0 and less than or equal to 1; the method further includes: The first target predicted value of the target traffic indicator within the first preset time period is subjected to inverse normalization to obtain the second target predicted value of the target traffic indicator within the first preset time period.

4. The method according to claim 2, characterized in that, The first historical time period has M periods, where M is a positive integer; Before obtaining the actual value of the target traffic indicator within the first historical time period and the actual value within the second target preset time period, the method further includes: Obtain N true values ​​of the target traffic indicator within a second target historical period, wherein the second target historical period includes N first historical periods, where N is a positive integer and N is less than M; Based on preset expansion rules, the true values ​​of the target traffic index are generated in M ​​first historical time periods.

5. The method according to claim 4, characterized in that, Before obtaining N real values ​​of the target traffic indicator within the second target historical time period, the method further includes: Obtain N raw values ​​of the target traffic indicator within the second target historical time period; The original value is subjected to anomaly detection according to the preset anomaly detection rules to obtain the target detection result, wherein the target detection result includes anomalies in the original value; If the target detection result is that the original value is abnormal, the abnormal original value is replaced with a normal value according to a preset interpolation rule to obtain N true values ​​of the target flow index in the second target historical period.

6. The method according to claim 5, characterized in that, There are multiple anomaly detection rules; the step of performing anomaly detection on the original value according to the preset anomaly detection rules to obtain the target detection result includes: Anomaly detection is performed on the original value according to multiple anomaly detection rules to obtain multiple sub-detection results; The number of target sub-detection results is compared with a preset number to obtain the target detection result, wherein the target sub-detection result is the sub-detection result of the original value being abnormal.

7. A flow prediction device, characterized in that, include: The prediction module is used to predict the traffic usage of the target cell in a first preset time period using a preset target prediction model, and to obtain the first predicted value of the target traffic index in the first preset time period, wherein there are multiple traffic indices. The acquisition module is used to acquire the target prediction error value corresponding to the target traffic indicator from the preset correspondence between traffic indicators and prediction error values, wherein the prediction error value is the error value of the target prediction model in predicting the traffic indicator, and the prediction error value is less than a preset threshold. The determining module is used to determine the first target predicted value of the target traffic indicator within the first preset time period based on the first predicted value and the target prediction error value. The target traffic metrics include at least two of the following: mobile network traffic value, resource utilization rate, and number of user connections.

8. A flow prediction device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the traffic prediction method as described in any one of claims 1 to 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 traffic prediction method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the device, the device is able to perform the traffic prediction method as described in any one of claims 1-6.

Images