Model training method, flow prediction method, device, equipment, medium and product

Abstract

The application relates to the computer technical field, in particular to a model training method, a traffic prediction method, a device, equipment, a medium and a product, the method comprises the following steps: training a model by using a last update result of historical traffic data to obtain a prediction result of the model; in the case that the prediction result is greater than a threshold value, determining a current update result based on the last update result and an influence factor; in the case that the prediction result obtained based on the current update result is greater than the threshold value, updating the last update result to the current update result until the prediction result is less than or equal to the threshold value and the model converges; and performing traffic prediction based on the obtained final converged model. By updating the historical traffic data by using the influence factor, the final converged model is obtained, the influence of real-time operator strategies in the network on traffic prediction is considered, the prediction result is more in line with the actual situation of the network, and the prediction accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a model training method, a traffic prediction method, an apparatus, a device, a medium, and a product. Background Technology

[0002] With the development of internet technology, network traffic prediction technology is also constantly improving and evolving. The importance of network traffic prediction lies in its ability to help network operators, enterprises, or service providers better understand and predict user behavior, thereby optimizing resource allocation and improving service quality. Therefore, the focus of network traffic prediction is accuracy.

[0003] There are many algorithms for traffic prediction, and to improve prediction accuracy, various algorithms have developed their own optimization methods. For example, data preprocessing can accelerate model training, improving model stability and generalization ability, or optimizing model parameters can accelerate convergence and enhance network generalization. However, these existing optimization methods do not consider the impact of real-time updated traffic scheduling or management strategies in the real network environment on the prediction results, which may lead to predictions that do not match the actual network situation, thus making it difficult to guarantee prediction accuracy. Summary of the Invention

[0004] In view of the above problems, this application is made to provide a model training method, a flow prediction method, an apparatus, a device, a medium, and a product.

[0005] According to one aspect of this application, a model training method for traffic prediction is provided, comprising:

[0006] The model is trained using the last updated result of historical traffic data to obtain the model's prediction result; if the last updated result is the first updated result, the last updated result is obtained based on the historical traffic data and the current influencing factors; the influencing factors are operator strategies that affect network traffic usage; the prediction result is the network traffic usage at a future set time, the future set time being later than the time when the historical traffic data was generated.

[0007] If the predicted result is greater than the threshold, the current update result of the historical traffic data is determined based on the previous update result of the historical traffic data and the influencing factor;

[0008] If the prediction result of the model obtained based on the current update result is greater than the threshold, the previous update result is updated to the current update result, and the process returns to the step of training the model using the previous update result with historical traffic data to obtain the model's prediction result, until the model's prediction result is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0009] Furthermore, the model training method for traffic prediction according to one aspect of this application also includes:

[0010] Before obtaining the model's prediction results based on the previous update results of historical traffic data, the following steps are also included:

[0011] Obtain raw historical traffic data;

[0012] The original historical traffic data is grouped according to preset data attributes to obtain multiple groups of historical traffic data. The data attributes are the geographical region to which the traffic data belongs, the length of the traffic data, or the resource type of the traffic data.

[0013] Furthermore, the model training method for traffic prediction according to one aspect of this application also includes:

[0014] Obtain raw historical traffic data, including:

[0015] Based on a preset maximum future time, the original historical traffic data for a historical time period is obtained, wherein the historical time period has the same length as the maximum future time.

[0016] According to another aspect of this application, a traffic prediction method is provided, comprising:

[0017] Obtain the traffic data to be predicted for a historical time period;

[0018] The final convergence model obtained above is used to predict the traffic data to be predicted, and the prediction result of the future set time of the traffic data demand is obtained.

[0019] Furthermore, according to another aspect of the traffic prediction method of this application, it also includes:

[0020] The predicted flow data is then predicted using the final convergence model obtained above, including:

[0021] According to preset data attributes, the traffic data to be predicted is grouped to obtain multiple data groups; and the data attributes of historical traffic data used to train the final convergence model are obtained.

[0022] Select a target data group from the plurality of data groups that has the same data attributes as the historical traffic data, and use the final convergence model to predict the traffic of the target data group.

[0023] Furthermore, according to another aspect of the traffic prediction method of this application, it also includes:

[0024] Using the final convergence model, traffic prediction is performed on the target data set, including:

[0025] Having obtained all the final converged models, obtain the last prediction result of each final converged model during the training phase, with each final converged model corresponding to a different future time setting;

[0026] Select the n target prediction results with the highest accuracy from the multiple last prediction results, and use the final convergence model corresponding to the target prediction results as the traffic prediction model;

[0027] If the future set time of the target data group's demand is consistent with the future set time corresponding to the traffic prediction model, the traffic prediction model is used to predict the traffic of the target data group.

[0028] According to another aspect of this application, a flow prediction apparatus is provided, comprising:

[0029] The training module is used to train the model using the last update result of historical traffic data to obtain the model's prediction result; if the last update result is the first update result, the last update result is obtained based on the historical traffic data and the current influencing factor; the influencing factor is the operator's strategy that affects the network's traffic usage.

[0030] The data update module is used to determine the current update result of the historical traffic data based on the previous update result of the historical traffic data and the influencing factor when the prediction result is greater than the threshold.

[0031] The training update module is used to update the previous update result to the current update result when the prediction result of the model obtained based on the current update result is greater than the threshold, and return to the step of training the model using the previous update result of historical traffic data to obtain the prediction result of the model, until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0032] The acquisition module is used to acquire traffic data to be predicted for historical time periods;

[0033] The prediction module is used to predict the traffic data to be predicted using the final convergence model obtained above, and to obtain the prediction result of the future set time of the traffic data demand.

[0034] According to another aspect of this application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method of yet another aspect above.

[0035] According to another aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method of yet another aspect described above.

[0036] According to another aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the method of yet another aspect described above.

[0037] As will be described in detail below, a traffic prediction method, apparatus, device, storage medium, and program product according to embodiments of this application uses the comparison result of the prediction result with a threshold as the trigger condition for calling the influence factor, updates the historical traffic data using the influence factor, and trains the model with the updated historical traffic data to obtain a final converged model. This takes into account the impact of real-time operator policies in the network on traffic prediction. When using the final converged model for traffic prediction, the traffic prediction result can be made to better match the actual network situation, which is beneficial to improving the accuracy of traffic prediction.

[0038] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0039] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0040] Figure 1 This is a flowchart illustrating a model training method for traffic prediction according to an embodiment of this application.

[0041] Figure 2 This is a flowchart illustrating a traffic prediction method according to an embodiment of this application.

[0042] Figure 3This is a flowchart illustrating yet another traffic prediction method according to an embodiment of this application.

[0043] Figure 4 This is a diagram illustrating the module call logic of a traffic prediction method according to an embodiment of this application.

[0044] Figure 5 This is a schematic diagram of the structure of a flow prediction device according to an embodiment of this application.

[0045] Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of this application.

[0046] Figure 7 This is a schematic diagram illustrating a computer program product according to an embodiment of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application more apparent, exemplary embodiments according to this application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0048] With the development of internet technology, network traffic prediction technology is also constantly improving and evolving. The importance of network traffic prediction lies in its ability to help network operators, enterprises, or service providers better understand and predict user behavior, thereby optimizing resource allocation and improving service quality. Therefore, the focus of network traffic prediction is accuracy.

[0049] There are many algorithms for traffic prediction, and to improve prediction accuracy, various algorithms have developed their own optimization methods. For example, data preprocessing can accelerate model training, improving model stability and generalization ability, or optimizing model parameters can accelerate convergence and enhance network generalization. However, these existing optimization methods do not consider the impact of real-time updated traffic scheduling or management strategies in the real network environment on the prediction results, which may lead to predictions that do not match the actual network situation, thus making it difficult to guarantee prediction accuracy.

[0050] The above description, with reference to the accompanying drawings, illustrates a model training method, traffic prediction method, apparatus, device, medium, and product according to embodiments of this application. By using the comparison between the prediction result and a threshold as the trigger condition for calling the influencing factor, the historical traffic data is updated using the influencing factor. The updated historical traffic data is then used to train the model to obtain a final converged model. This method takes into account the impact of real-time operator policies on traffic prediction in the network. When using the final converged model for traffic prediction, the traffic prediction result can better match the actual network situation, which is beneficial to improving the accuracy of traffic prediction.

[0051] To facilitate understanding of this embodiment, a detailed description of the model training method for traffic prediction disclosed in this application embodiment is provided first. The execution subject of the model training method provided in this application embodiment is generally a computer device with certain computing capabilities. This computer device includes, for example, a terminal device, a server, or other processing devices. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, this model training method can be implemented by the processor calling computer-readable instructions stored in memory.

[0052] like Figure 1 The diagram shows a flowchart of a model training method for traffic prediction provided in an embodiment of this application. The method includes steps S101-S103:

[0053] S101: Train the model using the last update result of historical traffic data to obtain the model's prediction results.

[0054] Historical traffic data refers to traffic data collected over a historical period, such as the past 12 months, which can be obtained from base stations, servers, or cloud platforms. The model is used for traffic prediction, such as the SVR model. The prediction result represents network traffic usage at a future time, which is later than the time the historical traffic data was generated, such as one month or three months from now. The metric represented by the prediction result is the same as the threshold; for example, if the threshold represents link bandwidth utilization, then the prediction result also represents link bandwidth utilization.

[0055] If the previous update result is the first update result, the previous update result is obtained based on historical traffic data and current influencing factors. The influencing factors are operator policies that affect network traffic usage, such as traffic scheduling policies and data management policies with real-time update characteristics.

[0056] In this preferred embodiment, before S101, the following steps are also included:

[0057] (1) Obtain the original historical traffic data.

[0058] Specifically, based on a preset maximum future time, raw historical traffic data for a historical period is obtained. The historical time period has the same length as the maximum future time. For example, if the preset future times are 1 month, 2 months, ... 12 months, then the maximum future time is 12 months, and raw historical traffic data for the past 12 months is obtained. Preferably, in this embodiment, the raw historical traffic data is daily granular data, meaning it is raw historical traffic data for the past 12 months collected in "days".

[0059] (2) According to the preset data attributes, the original historical traffic data is grouped to obtain multiple sets of historical traffic data.

[0060] In this preferred embodiment, the data attributes are as follows:

[0061] 1) Geographical region to which the traffic data belongs: such as various provinces and cities, or various districts within a city.

[0062] 2) Resource types of traffic data: such as traffic data of device ports, traffic data of links, etc.

[0063] 3) Length of traffic data: Traffic data for some network resource indicators may not be monitored continuously, or due to other operator policies, traffic data for certain resource indicators may only be monitored periodically. Therefore, the monitoring duration of traffic data (i.e., the length of traffic data) may vary. For example, traffic data for port A may have been monitored for 12 months in the past 12 months, while traffic data for port B may have been monitored for 6 months in the past 12 months.

[0064] This application embodiment sets data attributes based on the characteristics of the network itself, classifying the original historical traffic data to provide more accurate sample data for model training, which is beneficial to improving the prediction accuracy. Specifically, according to the above three data attributes, the original historical traffic data is divided into three groups, resulting in three groups of historical traffic data. In step S101, each group of historical traffic data is used sequentially to train the model.

[0065] S102: If the prediction result is greater than the threshold, determine the current update result of the historical traffic data based on the previous update result of the historical traffic data, the influencing factor, and the threshold.

[0066] Among them, the threshold represents an indicator related to the influencing factor. Assuming that the threshold represents the link bandwidth utilization rate, the algorithm of the influencing factor can combine the historical traffic data related to the link bandwidth of the entire network to calculate and obtain a batch of new traffic data (i.e., the current update result). Under this batch of new traffic data, the link bandwidth utilization rate of the entire network can reach a balanced state and all of them do not exceed the threshold.

[0067] S103: If the prediction result of the model obtained based on the current update result is greater than the threshold, update the previous update result to the current update result, and return to execute S101 until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0068] If the model's prediction result is less than or equal to the threshold, but the model has not converged, then in subsequent training phases, the obtained prediction result will no longer be compared with the threshold until the model converges and the final converged model is obtained. Alternatively, if the model has converged, but the model's prediction result is still greater than the threshold, then the model will continue to be trained until the prediction result is less than or equal to the threshold and the final converged model is obtained.

[0069] like Figure 2 The diagram shown is a flowchart of a traffic prediction method provided in an embodiment of this application. The method includes steps S201-S202:

[0070] S201: Obtain the traffic data to be predicted for a historical time period.

[0071] The traffic data to be predicted is the sample historical traffic data used when applying the final convergence model for traffic prediction, such as historical traffic data of the past 12 months. In this embodiment, the traffic data to be predicted is preferably daily granular data, that is, historical traffic data of the past 12 months collected with "day" as the time unit.

[0072] S202: The final convergence model obtained in the above embodiments is used to predict the traffic data to be predicted, and the prediction result of the future set time of the traffic data demand to be predicted is obtained.

[0073] Specifically, the traffic data to be predicted is grouped according to preset data attributes to obtain multiple data groups; and the data attributes of historical traffic data used to train the final converged model are obtained.

[0074] Select a target data group from multiple data groups that has the same data attributes as the historical traffic data, and use the final convergence model to predict the traffic of the target data group.

[0075] Combining the three data attributes described in the above embodiments, the traffic data to be predicted is grouped into three data groups. Assuming that the data attribute of the historical traffic data used to train the traffic prediction model is resource type, the target data group with the resource type attribute is selected from the three data groups. By selecting a final convergence model with the same data attribute for each data group, the accuracy of traffic prediction can be improved.

[0076] Specifically, S202 includes steps 11-13:

[0077] Step 11: Given all the final converged models, obtain the last prediction result of each final converged model during the training phase.

[0078] Each final convergent model corresponds to a different future timeframe. Based on the preset number of future timeframes, historical traffic data is used to train the models, resulting in the same number of final convergent models. For example, assuming the preset future timeframes are 1 month, 2 months, 3 months…12 months, and the number is 12, then 12 final convergent models will be obtained. For instance, the first final convergent model is used to predict traffic data for the next month, and the third final convergent model is used to predict traffic data for the next 3 months.

[0079] Step 12: Select the n target prediction results with the highest accuracy from the multiple last prediction results, and use the final converged model corresponding to the target prediction results as the traffic prediction model.

[0080] Each prediction result output by the model includes an accuracy rate, indicating the accuracy of the prediction. The n models with the highest accuracy are selected from the 12 ultimately converged models as the traffic prediction models. n is a positive integer, and its specific value can be chosen according to actual needs. In this embodiment, n=2. The two ultimately converged models with the highest accuracy predict the next 7 months and 12 months respectively. Therefore, there are two traffic prediction models: the first predicts traffic data for the next 7 months, and the second predicts traffic data for the next 12 months.

[0081] Step 13: If the future set time of the target data group's demand is consistent with the future set time corresponding to the traffic prediction model, use the traffic prediction model to predict the traffic for the target data group.

[0082] Specifically, assuming the target data group's future target timeframe is 7 months, the first traffic prediction model can be used. If the target data group's future target timeframe is 12 months, the second traffic prediction model can be used. Preferably, in practical applications, if the target data group's future target timeframe differs from the future target timeframe of the traffic prediction model, the corresponding final convergence model can be selected for traffic prediction based on the target data group's future target timeframe. For example, if the target data group's future target timeframe is 6 months, the sixth final convergence model (used to predict traffic data for the next 6 months) can be selected.

[0083] like Figure 3 The diagram shown is another flowchart of the traffic prediction method provided in this application embodiment, the method including S301-S309:

[0084] S301: Obtain raw historical traffic data.

[0085] Specifically, based on the preset maximum future setting time, the raw historical traffic data for the historical time period is obtained. The historical time period is the same as the maximum future setting time. For example, if the preset future setting time is 1 month, 2 months...12 months, then the maximum future setting time is 12 months. In this case, the raw historical traffic data for the past 12 months is obtained. The daily granularity means that the raw historical traffic data is collected in "days".

[0086] S302: Data storage.

[0087] The collected raw historical traffic data at the 12-month daily granularity is stored in the storage unit as sample data used in the model training phase.

[0088] S303: Data processing.

[0089] The original historical traffic data is cleaned and preprocessed to remove discontinuous, missing, or outlier values, and to fill in minor flaws in the data, or to change the time node and move the later data forward as a whole, so as to preserve the original form of the data.

[0090] S304: Group the original historical traffic data according to the preset data attributes to obtain multiple groups of historical traffic data.

[0091] In this preferred embodiment, the data attributes are as follows:

[0092] 1) Geographical region to which the traffic data belongs: such as various provinces and cities, or various districts within a city.

[0093] 2) Resource type: such as traffic data of device ports, traffic data of links, etc.

[0094] 3) Length of traffic data: Traffic data for some network resource indicators may not be monitored continuously, or due to other operator policies, traffic data for certain resource indicators may only be monitored periodically. Therefore, the monitoring duration of traffic data (i.e., the length of traffic data) may vary. For example, traffic data for port A may have been monitored for 12 months in the past 12 months, while traffic data for port B may have been monitored for 6 months in the past 12 months.

[0095] Data attributes were set based on the characteristics of the network itself, and the original historical traffic data was divided into three categories. Each category of historical traffic data was trained and predicted independently.

[0096] S305: Use historical traffic data to train the model.

[0097] Specifically, S305 includes steps 31-33:

[0098] Step 31: Train the model using the last update result of the historical traffic data to obtain the model's prediction results.

[0099] Historical traffic data refers to traffic data collected over the past 12 months. Data attributes include resource type, which can be obtained from base stations, servers, or cloud platforms. The model can specifically be an SVR model. The metric represented by the prediction result is the same as the threshold. For example, if the threshold represents link bandwidth utilization, then the prediction result will also represent link bandwidth utilization.

[0100] If the previous update result is the initial update result, it was obtained based on historical traffic data and current influencing factors. These influencing factors are operator policies that affect network traffic usage, such as traffic scheduling policies and data management policies with real-time update capabilities. The threshold indicator is related to the influencing factors. Assuming the threshold indicator represents link bandwidth utilization, the algorithm for calculating the influencing factors can combine historical traffic data related to link bandwidth across the entire network to obtain a new set of traffic data. Under this new set of traffic data, the link bandwidth utilization across the entire network can reach a balanced state, and none of the data should exceed the threshold.

[0101] Step 32: If the prediction result is greater than the threshold, determine the current update result of the historical traffic data based on the previous update result, the influencing factor, and the threshold.

[0102] Among them, the threshold represents an indicator related to the influencing factor. Assuming that the threshold represents the link bandwidth utilization rate, the algorithm of the influencing factor can combine the historical traffic data related to the link bandwidth of the entire network to calculate and obtain a batch of new traffic data (i.e., the current update result). Under this batch of new traffic data, the link bandwidth utilization rate of the entire network can reach a balanced state and all of them do not exceed the threshold.

[0103] Step 33: If the prediction result of the model obtained based on the current update result is greater than the threshold, update the previous update result to the current update result, and return to execute step 31 until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0104] S306: Obtain traffic data to be predicted for a historical time period.

[0105] The traffic data to be predicted is the sample historical traffic data used when applying the final convergence model for traffic prediction, such as historical traffic data of the past 12 months. In this embodiment, the traffic data to be predicted is preferably daily granular data, that is, historical traffic data of the past 12 months collected with "day" as the time unit.

[0106] S307: Using a final convergence model, predict the traffic data to be predicted and obtain the prediction results of the future set time of the traffic data demand.

[0107] Specifically, this includes steps 34-38:

[0108] Step 34: Group the traffic data to be predicted according to the preset data attributes to obtain multiple data groups; and obtain the data attributes of the historical traffic data used to train the final converged model.

[0109] Step 35: Select the target data group from multiple data groups that has the same data attributes as the historical traffic data.

[0110] Combining the three data attributes of the above embodiments, the traffic data to be predicted is grouped into three data groups. Assuming that the data attribute of the historical traffic data used to train the traffic prediction model is resource type, the target data group with the attribute of resource type is selected from the three data groups.

[0111] Step 36: Given all the final converged models, obtain the last prediction result of each final converged model during the training phase.

[0112] The final convergence models correspond to different future timeframes. Based on the preset number of future timeframes, historical traffic data is used to train the models, resulting in the same number of final convergence models. For example, assuming the preset future timeframes are 1 month, 2 months, 3 months…12 months, and the number is 12, then this set of historical traffic data with the resource type attribute will yield 12 final convergence models. For instance, the first final convergence model is used to predict traffic data for the next month, and the second final convergence model is used to predict traffic data for the next 2 months.

[0113] Step 37: Select the n target prediction results with the highest accuracy from multiple last prediction results, and use the final converged model corresponding to the target prediction results as the traffic prediction model.

[0114] Each prediction output from the model includes an accuracy rate, indicating the accuracy of the prediction. The n models with the highest accuracy from the 12 ultimately converged models are selected as the traffic prediction models. n is a positive integer, and its specific value can be chosen according to actual needs. In this embodiment, n=2. The two ultimately converged models with the highest accuracy predicting the next 8 months and 10 months respectively are selected. Therefore, for this set of historical traffic data with the resource type attribute, there are two traffic prediction models: the first predicts traffic data for the next 8 months, and the second predicts traffic data for the next 10 months.

[0115] Step 38: If the future set time of the target data group's demand is consistent with the future set time corresponding to the traffic prediction model, use the traffic prediction model to predict the traffic for the target data group.

[0116] Specifically, assuming the target data group's future target timeframe is 7 months, the first traffic prediction model can be used. If the target data group's future target timeframe is 12 months, the second traffic prediction model can be used. Preferably, in practical applications, if the target data group's future target timeframe differs from the future target timeframe of the traffic prediction model, the corresponding final convergence model can be selected for traffic prediction based on the target data group's future target timeframe. For example, if the target data group's future target timeframe is 8 months, the eighth final convergence model (used to predict traffic data for the next 8 months) can be selected.

[0117] S308: Results Display.

[0118] The traffic prediction results obtained from each data group are presented to users in a visual way. Various data indicators and analysis results of the traffic prediction results can be displayed in the form of charts and other methods. Users can query, filter and sort the predicted traffic data through an interactive interface.

[0119] S309: Optimize traffic strategy.

[0120] Based on the traffic prediction results, network traffic optimization strategies are generated, such as bandwidth adjustment and resource allocation. In this preferred embodiment, network performance can also be monitored and alerted to respond promptly to changes in network traffic.

[0121] like Figure 4 The diagram shown illustrates the module call logic of the traffic prediction method provided in this application embodiment. Based on the digital twin background, this application embodiment deploys various unit modules related to the traffic prediction method in the digital twin layer and application interaction layer of the digital twin, specifically including:

[0122] Data acquisition unit: Acquires raw historical traffic data at the daily granularity for the past 12 months.

[0123] Data storage unit: The raw historical traffic data collected at the 12-month daily granularity is stored in the storage unit as sample data used in the model training phase.

[0124] Data processing unit: Cleans and preprocesses the raw historical traffic data to remove discontinuous, missing or outlier values, fills in minor flaws in the data, or changes the time node to move the subsequent data forward as a whole, preserving the original form of the data.

[0125] Data classification unit: According to preset data attributes, the original historical traffic data is grouped to obtain x groups of historical traffic data.

[0126] In this preferred embodiment, the data attributes are as follows:

[0127] 1) Geographical region to which the traffic data belongs: such as various provinces and cities, or various districts within a city.

[0128] 2) Resource type: such as traffic data of device ports, traffic data of links, etc.

[0129] 3) Length of traffic data: Traffic data for some network resource indicators may not be monitored continuously, or due to other operator policies, traffic data for certain resource indicators may only be monitored periodically. Therefore, the monitoring duration of traffic data (i.e., the length of traffic data) may vary. For example, traffic data for port A may have been monitored for 12 months in the past 12 months, while traffic data for port B may have been monitored for 6 months in the past 12 months.

[0130] Model training unit: Based on historical traffic data for each group, a separate model is trained for each group of historical traffic data, ultimately deriving a traffic prediction model for each group of historical traffic data. Different traffic prediction models only predict data groups with the same data attributes. The training phase of the SVR model specifically includes steps 41-45:

[0131] Step 41: Train the model using the last update result of the historical traffic data to obtain the model's prediction results.

[0132] In this preferred embodiment, the threshold represents the link bandwidth utilization rate, and the prediction result also represents the link bandwidth utilization rate.

[0133] If the previous update result is the initial update result, it was obtained based on historical traffic data and current influencing factors. These influencing factors are operator policies that affect network traffic usage, such as traffic scheduling policies and data management policies with real-time update capabilities. The threshold indicator is related to the influencing factors. Assuming the threshold indicator represents link bandwidth utilization, the algorithm for calculating the influencing factors can combine historical traffic data related to link bandwidth across the entire network to obtain a new set of traffic data. Under this new set of traffic data, the link bandwidth utilization across the entire network can reach a balanced state, and none of the data should exceed the threshold.

[0134] Step 42: If the prediction result is greater than the threshold, calculate the last update result of the historical traffic data based on the impact factor to obtain the updated traffic data;

[0135] The threshold-based indicator calculates the indicator value using updated traffic data.

[0136] If the indicator value is less than or equal to the threshold, the updated traffic data will be used as the current update result of the historical traffic data.

[0137] Step 43: If the prediction result of the model obtained based on the current update result is greater than the threshold, update the previous update result to the current update result, and return to execute step 21 until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0138] Step 44: Given all the final converged models, obtain the last prediction result of each final converged model during the training phase.

[0139] The final convergence models correspond to different future timeframes. Based on the preset number of future timeframes, historical traffic data is used to train the models, resulting in the same number of final convergence models. For example, assuming the preset future timeframes are 1 month, 2 months, 3 months…12 months, and the number is 12, then this set of historical traffic data with the resource type attribute will yield 12 final convergence models. For instance, the first final convergence model is used to predict traffic data for the next month, and the second final convergence model is used to predict traffic data for the next 2 months.

[0140] Step 45: Select the n target prediction results with the highest accuracy from the multiple last prediction results, and use the final converged model corresponding to the target prediction results as the traffic prediction model.

[0141] Each prediction output from the model includes an accuracy rate, indicating the accuracy of the prediction. The n models with the highest accuracy from the 12 ultimately converged models are selected as the traffic prediction models. n is a positive integer, and its specific value can be chosen according to actual needs. In this embodiment, n=2. The two ultimately converged models with the highest accuracy predicting the next 8 months and 10 months respectively are selected. Therefore, for this set of historical traffic data with the resource type attribute, there are two traffic prediction models: the first predicts traffic data for the next 8 months, and the second predicts traffic data for the next 10 months.

[0142] Data Prediction Unit: Using the obtained traffic prediction model, this unit predicts the traffic data to be predicted, obtaining prediction results for the future demand of the traffic data at a predetermined time. The traffic data to be predicted consists of daily-level traffic data from the past 12 months up to the present. Based on the three data attributes mentioned above, the traffic data is grouped into three data groups. Based on the number of traffic prediction models, each data group will obtain the same number of traffic prediction results. Combining with step 38, taking the data group with the attribute of resource type as an example, two sets of traffic prediction results will be predicted: traffic prediction results for the next 8 months and the next 10 months, respectively.

[0143] Data Display Unit: Presents the prediction results of each data group to users in a visual manner. Various data indicators and analysis results of the prediction results can be displayed in the form of charts and other methods. Users can query, filter and sort the predicted traffic data through an interactive interface.

[0144] Decision-making and response unit: Based on the prediction results, it generates network traffic optimization strategies, such as bandwidth adjustment and resource allocation. Preferably, in this embodiment, it can also monitor and issue alarms for network performance to respond promptly to changes in network traffic.

[0145] According to another aspect of the embodiments of this application, a traffic flow prediction device is provided, such as... Figure 5 As shown, the device includes:

[0146] Training module 101 is used to train the model using the last update result of historical traffic data to obtain the prediction result of the model; when the last update result is the first update result, the last update result is obtained based on the historical traffic data and the current influencing factor; the influencing factor is the operator's strategy that affects the network's traffic usage, and the prediction result is the network's traffic usage at a future set time, the future set time being later than the time when the historical traffic data was generated.

[0147] The data update module 102 is used to determine the current update result of the historical traffic data based on the previous update result of the historical traffic data and the influencing factor when the prediction result is greater than the threshold.

[0148] The training update module 103 is used to update the previous update result to the current update result when the prediction result of the model obtained based on the current update result is greater than the threshold, and return to the step of training the model using the previous update result of historical traffic data to obtain the prediction result of the model, until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

[0149] Module 104 is used to acquire traffic data to be predicted for historical time periods.

[0150] The prediction module 105 is used to predict the traffic data to be predicted using the final convergence model obtained above, and to obtain the prediction result of the future set time of the traffic data demand to be predicted.

[0151] The traffic prediction device is also used to: acquire the original historical traffic data before obtaining the model's prediction result based on the previous update result of the historical traffic data;

[0152] The original historical traffic data is grouped according to preset data attributes to obtain multiple groups of historical traffic data. The data attributes are the geographical region to which the traffic data belongs, the length of the traffic data, or the resource type of the traffic data.

[0153] In one or more embodiments, obtaining raw historical traffic data includes:

[0154] Based on a preset maximum future time, the original historical traffic data for a historical time period is obtained, wherein the historical time period has the same length as the maximum future time.

[0155] In one or more embodiments, the prediction module 105 is used to:

[0156] According to preset data attributes, the traffic data to be predicted is grouped to obtain multiple data groups; and the data attributes of historical traffic data used to train the final convergence model are obtained.

[0157] Select a target data group from the plurality of data groups that has the same data attributes as the historical traffic data, and use the final convergence model to predict the traffic of the target data group.

[0158] In one or more embodiments, using the final convergence model to perform traffic prediction on the target data set includes:

[0159] Having obtained all the final converged models, obtain the last prediction result of each final converged model during the training phase, with each final converged model corresponding to a different future time setting;

[0160] Select the n target prediction results with the highest accuracy from the multiple last prediction results, and use the final convergence model corresponding to the target prediction results as the traffic prediction model;

[0161] If the future set time of the target data group's demand is consistent with the future set time corresponding to the traffic prediction model, the traffic prediction model is used to predict the traffic of the target data group.

[0162] The flow prediction device and the flow prediction method provided in this application are based on the same inventive concept and have the same beneficial effects as the methods they employ, operate, or implement.

[0163] This application also provides a computer device for performing the above-described traffic prediction method. Please refer to... Figure 6 It illustrates a schematic diagram of a computer device provided by some embodiments of this application. For example... Figure 6 As shown, the computer device 8 includes: a processor 800, a memory 801, a bus 802, and a communication interface 803. The processor 800, the communication interface 803, and the memory 801 are connected via the bus 802. The memory 801 stores a computer program that can run on the processor 800. When the processor 800 runs the computer program, it executes the traffic prediction method provided in any of the foregoing embodiments of this application.

[0164] The memory 801 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 803 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0165] Bus 802 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 801 is used to store programs. After receiving an execution instruction, the processor 800 executes the program. The traffic prediction method disclosed in any of the foregoing embodiments of this application can be applied to the processor 800, or implemented by the processor 800.

[0166] The processor 800 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 800 or by instructions in software form. The processor 800 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPTA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 801. Processor 800 reads the information in memory 801 and, in conjunction with its hardware, completes the steps of the above method.

[0167] The computer device provided in this application embodiment and the traffic prediction method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0168] This application also provides a computer-readable storage medium corresponding to the traffic prediction method provided in the foregoing embodiments. The computer-readable storage medium is an optical disc, on which a computer program (i.e., a computer program product) is stored. When the computer program is run by a processor, it executes the traffic prediction method provided in any of the foregoing embodiments.

[0169] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0170] The computer-readable storage medium provided in the above embodiments of this application and the traffic prediction method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0171] This application also provides a computer program product; please refer to [reference needed]. Figure 7 The computer program product 600 carries program code, namely computer program 601. The instructions included in the computer program 601 can be used to execute the steps of the traffic prediction method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0172] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0173] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0174] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0175] Additionally, as used herein, the “or” used in a list of items beginning with “at least one” indicates a separate list, such that a list of, for example, “at least one of A, B, or C” means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word “exemplary” does not imply that the described example is preferred or better than other examples.

[0176] It should also be noted that in the system and method of this application, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions of this application.

[0177] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0178] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0179] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A model training method for traffic flow prediction, characterized in that, include: The model is trained using the last updated result of historical traffic data to obtain the model's prediction results; If the previous update result is the first update result, the previous update result is obtained based on the historical traffic data and the current influencing factors; the influencing factors are operator strategies that affect network traffic usage, and the prediction result is the network traffic usage at a future set time, which is later than the time when the historical traffic data was generated. If the predicted result is greater than the threshold, the current update result of the historical traffic data is determined based on the previous update result of the historical traffic data, the influencing factor, and the threshold. If the prediction result of the model obtained based on the current update result is greater than the threshold, the previous update result is updated to the current update result, and the process returns to the step of training the model using the previous update result with historical traffic data to obtain the model's prediction result, until the model's prediction result is less than or equal to the threshold and the model converges, thus obtaining the final converged model.

2. The model training method for traffic prediction as described in claim 1, characterized in that, Before obtaining the model's prediction results based on the previous update results of historical traffic data, the following steps are also included: Obtain raw historical traffic data; The original historical traffic data is grouped according to preset data attributes to obtain multiple groups of historical traffic data. The data attributes are the geographical region to which the traffic data belongs, the length of the traffic data, or the resource type of the traffic data.

3. The model training method for traffic prediction as described in claim 2, characterized in that, Obtain raw historical traffic data, including: Based on a preset maximum future time, the original historical traffic data for a historical time period is obtained, wherein the historical time period has the same length as the maximum future time.

4. A flow prediction method, characterized in that, include: Obtain the traffic data to be predicted for a historical time period; The final convergence model described in any one of claims 1-3 is used to predict the traffic data to be predicted, thereby obtaining the prediction result of the future set time of the traffic data demand.

5. The flow prediction method as described in claim 4, characterized in that, Predicting the traffic data to be predicted using the final convergence model obtained according to claims 1-3 includes: According to preset data attributes, the traffic data to be predicted is grouped to obtain multiple data groups; and the data attributes of historical traffic data used to train the final convergence model are obtained. Select a target data group from the plurality of data groups that has the same data attributes as the historical traffic data, and use the final convergence model to predict the traffic of the target data group.

6. The flow prediction method as described in claim 5, characterized in that, Using the final convergence model, traffic prediction is performed on the target data set, including: Having obtained all the final converged models, obtain the last prediction result of each final converged model during the training phase, with each final converged model corresponding to a different future time setting; Select the n target prediction results with the highest accuracy from the multiple last prediction results, and use the final convergence model corresponding to the target prediction results as the traffic prediction model; If the future set time of the target data group's demand is consistent with the future set time corresponding to the traffic prediction model, the traffic prediction model is used to predict the traffic of the target data group.

7. A flow prediction device, characterized in that, include: The training module is used to train the model using the last update results of historical traffic data to obtain the model's prediction results; If the previous update result is the first update result, the previous update result is obtained based on the historical traffic data and the current influencing factors; the influencing factors are operator strategies that affect network traffic usage; the prediction result is the network traffic usage at a future set time, and the future set time is later than the time when the historical traffic data was generated. The data update module is used to determine the current update result of the historical traffic data based on the previous update result of the historical traffic data and the influencing factor when the prediction result is greater than the threshold. The training update module is used to update the previous update result to the current update result when the prediction result of the model obtained based on the current update result is greater than the threshold, and return to the step of training the model using the previous update result of historical traffic data to obtain the prediction result of the model, until the prediction result of the model is less than or equal to the threshold and the model converges, thus obtaining the final converged model. The acquisition module is used to acquire traffic data to be predicted for historical time periods; The prediction module is used to predict the traffic data to be predicted using the final convergence model described in any one of claims 1-3, and to obtain the prediction result of the future set time of the traffic data demand to be predicted.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method according to any one of claims 4 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 4 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 4 to 6.

Images