Business traffic prediction method and device, electronic equipment and storage medium

Abstract

The application provides a service traffic prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: obtaining first historical service traffic data of a first time period; predicting the service traffic based on the first historical service traffic data by using a first traffic prediction model to obtain first traffic prediction data; the first traffic prediction model comprises a plurality of trend factors and weights of the plurality of trend factors; the plurality of trend factors correspond to a plurality of time sequence characteristics of the service traffic respectively, and the weights are used to indicate the correlation between the corresponding trend factors and the service traffic; predicting the service traffic based on the first historical service traffic data by using a second traffic prediction model to obtain second traffic prediction data; and fusing the first traffic prediction data and the second traffic prediction data by using a third model to obtain a traffic prediction result. The accuracy of traffic prediction is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a business traffic prediction method and device, electronic equipment and storage medium. BACKGROUND

[0002] In related technologies, most of the existing traffic prediction methods are to extract features from historical traffic and directly predict future traffic based on the extracted features. However, network operation can only passively accept and deal with problems in an after-treatment manner when VPN business traffic is overloaded. The current VPN business has traffic diversification (a seasonal change rule of VPN business traffic time series over time, which is constantly growing and diversified), VPN business traffic burstiness (traffic in the network is real-time variable, and when encountering various holidays or user peak traffic, it will show a burst trend), and VPN business traffic randomness (uncertain or unpredictable random phenomena of business traffic under certain conditions). The existing technical solutions cannot accurately predict the above traffic characteristics. SUMMARY

[0003] The present application aims to at least solve one of the technical problems in the related art to some extent.

[0004] To this end, the first purpose of the present application is to propose a business traffic prediction method to improve the accuracy of the prediction result.

[0005] The second purpose of the present application is to propose a business traffic prediction device.

[0006] The third purpose of the present application is to propose an electronic device.

[0007] The fourth purpose of the present application is to propose a computer-readable storage medium.

[0008] The fifth purpose of the present application is to propose a computer program product.

[0009] To achieve the above purpose, the first aspect of the present application proposes a business traffic prediction method, comprising:

[0010] obtaining first historical business traffic data of a first time period;

[0011] predicting business traffic based on the first historical business traffic data by a first traffic prediction model to obtain first traffic prediction data; wherein the first traffic prediction model comprises a plurality of trend factors and weights of the plurality of trend factors; the plurality of trend factors correspond to a plurality of time sequence characteristics of business traffic respectively, and the weights are used to indicate the correlation size between the corresponding trend factors and business traffic;

[0012] predicting, by a second traffic prediction model, the service traffic based on the first historical service traffic data, to obtain second traffic prediction data;

[0013] fusing, by a third model, the first traffic prediction data and the second traffic prediction data, to obtain a traffic prediction result.

[0014] In some embodiments, the predicting, by a first traffic prediction model, the service traffic based on the first historical service traffic data, to obtain first traffic prediction data comprises:

[0015] extracting time sequence features of the first historical service traffic data to obtain prediction basis data; wherein the time sequence features comprise a trend item, a periodic item, a holiday item, and a history item;

[0016] determining weight values of the plurality of trend factors based on the prediction basis data and a service traffic data source; wherein the service traffic data source comprises historical service traffic and index values corresponding to the historical service traffic, time sequence features;

[0017] fusing the plurality of trend factors and corresponding weight values to obtain first traffic prediction data.

[0018] In some embodiments, the determining weight values of the plurality of trend factors based on the prediction basis data and a service traffic data source comprises:

[0019] for a first trend factor corresponding to the trend item, the periodic item, and the holiday item, determining one item in the first trend factor as a target item, and determining trend factors other than the target item in the first trend factor as matching items;

[0020] selecting, from the service traffic data source, first target traffic data with different target items and same matching items; and calculating information entropy corresponding to the target items based on the first target traffic data;

[0021] determining weight values of the trend factors in the first trend factor based on the information entropy corresponding to each trend factor in the first trend factor.

[0022] In some embodiments, the determining weight values of the plurality of trend factors based on the prediction basis data and a service traffic data source comprises:

[0023] for a trend factor corresponding to the history item, obtaining, from the service traffic data source, second target traffic data with a similarity to the prediction basis data satisfying a preset condition;

[0024] determining the trend factor corresponding to the history item and a weight value based on the second target traffic data.

[0025] In some embodiments, the obtaining, from the service traffic data source, second target traffic data with a similarity to the prediction base data satisfying a preset condition comprises:

[0026] calculating a Euclidean distance between first data in each historical traffic data in the service traffic data source and corresponding data in the prediction base data, wherein the first data is at least one of historical service traffic, time sequence features, and index values;

[0027] obtaining, based on the Euclidean distance, second target traffic data with a similarity to the prediction base data satisfying a preset condition.

[0028] In some embodiments, the obtaining, from the service traffic data source, second target traffic data with a similarity to the prediction base data satisfying a preset condition comprises:

[0029] calculating a weighted sum of the first traffic prediction data and the second traffic prediction data by the third model, wherein a weight value of the third model is obtained through model training;

[0030] taking the weighted sum as the traffic prediction result.

[0031] In some embodiments, the method further comprises:

[0032] determining whether the traffic prediction result is greater than a preset traffic threshold;

[0033] when the traffic prediction result is greater than the preset traffic threshold, performing an alarm operation.

[0034] To achieve the above object, a second aspect embodiment of the present application provides a service traffic prediction device, comprising:

[0035] a data acquisition module configured to acquire first historical service traffic data of a first time period;

[0036] a first traffic prediction module configured to predict service traffic based on the first historical service traffic data by a first traffic prediction model to obtain first traffic prediction data, wherein the first traffic prediction model comprises a plurality of trend factors and respective weights of the plurality of trend factors; the plurality of trend factors correspond to a plurality of time sequence features of service traffic respectively, and the weights are used to indicate the correlation size between the corresponding trend factors and service traffic;

[0037] a second traffic prediction module configured to predict service traffic based on the first historical service traffic data by a second traffic prediction model to obtain second traffic prediction data;

[0038] A result fusion module is configured to fuse the first traffic prediction data and the second traffic prediction data by a third model to obtain a traffic prediction result.

[0039] In some embodiments, the first traffic prediction module is specifically configured to:

[0040] extracting time sequence features of the first historical traffic data to obtain prediction basis data, wherein the time sequence features include a trend item, a periodic item, a holiday item, and a history item;

[0041] determining weight values of the plurality of trend factors based on the prediction basis data and a traffic data source, wherein the traffic data source includes historical traffic and index values corresponding to the historical traffic and time sequence features;

[0042] fusing the plurality of trend factors and the corresponding weight values to obtain the first traffic prediction data.

[0043] In some embodiments, when the first traffic prediction module determines the weight values of the plurality of trend factors based on the prediction basis data and the traffic data source, the first traffic prediction module is configured to:

[0044] for the first trend factors corresponding to the trend item, the periodic item, and the holiday item, respectively determining one of the first trend factors as a target item, and determining trend factors other than the target item in the first trend factors as matching items;

[0045] selecting first target traffic data with different target items and the same matching items from the traffic data source, and calculating information entropy corresponding to the target items based on the first target traffic data;

[0046] determining the weight values of the trend factors in the first trend factors based on the information entropy corresponding to each of the trend factors in the first trend factors.

[0047] In some embodiments, when the first traffic prediction module determines the weight values of the plurality of trend factors based on the prediction basis data and the traffic data source, the first traffic prediction module is configured to:

[0048] for the trend factor corresponding to the history item, obtaining second target traffic data with a preset similarity to the prediction basis data from the traffic data source;

[0049] determining the trend factor corresponding to the history item and the weight value based on the second target traffic data.

[0050] In some embodiments, when the first traffic prediction module obtains the second target traffic data with a preset similarity to the prediction basis data from the traffic data source, the first traffic prediction module is configured to:

[0051] calculate a Euclidean distance between first data in each historical traffic data in the traffic flow data source and corresponding data in the prediction base data, wherein the first data is at least one of historical traffic flow, time sequence feature and index value;

[0052] obtain second target traffic data with a similarity to the prediction base data satisfying a preset condition based on the Euclidean distance.

[0053] In some embodiments, when the result fusion module fuses the first traffic prediction data and the second traffic prediction data to obtain a traffic prediction result through a third model, the result fusion module is configured to:

[0054] calculate a weighted sum of the first traffic prediction data and the second traffic prediction data through the third model, wherein a weight value of the third model is obtained through model training;

[0055] use the weighted sum as the traffic prediction result.

[0056] In some embodiments, the apparatus further comprises a pre-warning module configured to:

[0057] determine whether the traffic prediction result is greater than a preset traffic threshold;

[0058] when the traffic prediction result is greater than the preset traffic threshold, perform an alarm operation.

[0059] To achieve the above purpose, a third aspect of the present application provides an electronic device, comprising a processor and a memory connected with the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method of the first aspect.

[0060] To achieve the above purpose, a fourth aspect of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the method of the first aspect.

[0061] To achieve the above purpose, a fifth aspect of the present application provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the method of the first aspect.

[0062] The service traffic prediction method, apparatus, electronic device, and storage medium provided in this application predict service traffic using a first traffic prediction model. During traffic prediction, the first traffic prediction model fits trend factors and their respective weights corresponding to multiple time-series characteristics of the service traffic. Each trend factor is assigned its own weight, and the correlation between the corresponding trend factors and the service traffic is indicated by weights used to differentiate the trend factors, thereby improving the accuracy of traffic prediction. Furthermore, the prediction results from the first traffic prediction model are fused with those from a second traffic prediction model to further improve the accuracy of the prediction results.

[0063] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0064] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0065] Figure 1 This is a flowchart illustrating a service traffic prediction method provided in Embodiment 1 of this application.

[0066] Figure 2 An example diagram illustrating a traffic prediction method provided in an embodiment of this application;

[0067] Figure 3 This is a schematic diagram of the traffic prediction results provided in an embodiment of this application.

[0068] Figure 4 This is a flowchart illustrating a service traffic prediction method provided in Embodiment 2 of this application;

[0069] Figure 5 This is a block diagram of a traffic prediction device provided in Embodiment 3 of this application. Detailed Implementation

[0070] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0071] The following description, with reference to the accompanying drawings, describes a service traffic prediction method, apparatus, and device according to embodiments of this application.

[0072] Figure 1 This is a flowchart illustrating a service traffic prediction method provided in an embodiment of this application.

[0073] It should be noted that the execution subject of the service traffic prediction method in the embodiments of the present application is a service traffic prediction device in the embodiments of the present application, which can be configured in an electronic device to enable the electronic device to perform a service traffic prediction function.

[0074] As shown in Figure 1 The service traffic prediction method comprises the following steps:

[0075] Step 101, obtaining first historical service traffic data of a first time period.

[0076] The embodiments of the present application are based on the prediction of future time series traffic data based on historical traffic data, and therefore, it is necessary to obtain historical service traffic data in a recent time period.

[0077] Exemplarily, VPN service traffic data can be directly collected and stored, and historical service traffic data of a required time period can be obtained from the stored historical traffic data.

[0078] Step 102, predicting service traffic based on the first historical service traffic data by using a first traffic prediction model to obtain first traffic prediction data; wherein the first traffic prediction model comprises a plurality of trend factors and respective weights of the plurality of trend factors; and the plurality of trend factors correspond to a plurality of time series characteristics of service traffic respectively.

[0079] The first traffic prediction model in the embodiments of the present application is an improved Prophet model (hereinafter referred to as the Prophet model), which is a time series data prediction model based on an additive model, which can automatically detect trends and seasonality in data and combine them together to obtain a prediction value.

[0080] The basic form of the Prophet prediction model in the embodiments of the present application is as follows:

[0081] y(t) =αg(t)+βs(t) +λh(t) +θp(t)+ε(t)

[0082] Wherein, y(t) is a predicted value, g(t), s(t), h(t), p(t) and ε(t) are all trend factors, g(t) is a trend item, indicating the trend change of the time series on the non-period, s(t) is a periodic item, indicating the periodic influence in the time series, h(t) is a holiday item, indicating the influence of potential holidays with non-fixed periods on the predicted value in the time series; p(t) is a historical reference factor, which is a reference factor based on the change trend of the business traffic data with the same characteristics in the historical traffic data; ε(t) is an error item subject to normal distribution, which can also be called a noise item, indicating the random fluctuations that the model cannot predict; and α, β, λ and θ are weights of the corresponding trend factors.

[0083] The prediction process of the Prophet model is to obtain the predicted value corresponding to the time by fitting the trend factors respectively, and then to obtain the predicted value of the time series by accumulating the predicted values of different times according to the time sequence. The specific fitting method is to perform weighted summation on the trend item, the holiday item, the periodic item, the historical reference factor and the error item corresponding to the same time.

[0084] When performing weighted fitting, for different trend factors such as the trend item g(t), the periodic item s(t) and the holiday item h(t), adaptive weights are used to obtain the weight values for weighting.

[0085] As shown in FIG. 1, the first historical business traffic data is input into the Prophet model to obtain first traffic prediction data y(t). Figure 2

[0086] Step 103, predicting the business traffic based on the first historical business traffic data through a second traffic prediction model to obtain second traffic prediction data.

[0087] In the embodiments of the present application, the second traffic prediction model is an Xgboost model, which is a special gradient boosting decision tree and has the characteristics of fast and efficient prediction, and can perform classification prediction based on the decision tree.

[0088] The expression of the second traffic prediction model is as follows:

[0089]

[0090] Wherein, represents the predicted value of the i th data, K is the number of decision trees, F is a set of regression trees, x i represents the i th input data. The final prediction result y'(t) is obtained by predicting all input traffic data. That is, the prediction result y'(t) is obtained by fitting the Xgboost model.

[0091] As​Figure 2 The first historical traffic flow data is input into the Xgboost model to obtain second traffic prediction data y'(t).

[0092] In step 104, the first traffic prediction data and the second traffic prediction data are fused by a third model to obtain a traffic prediction result.

[0093] As an implementation manner, the first traffic prediction data and the second traffic prediction data are fused by a third model to obtain a traffic prediction result; including:

[0094] The weighted sum of the first traffic prediction data and the second traffic prediction data is calculated by the third model; wherein the weight value of the third model is obtained by model training; and the weighted sum is taken as the traffic prediction result.

[0095] That is, the traffic prediction is performed based on historical traffic data by the trained Prophet+Xgboost prediction model, and the change trend of time series data is fitted according to different time granularities such as week, month and year. A summable model is constructed based on the Prophet+Xgboost algorithm to perform prediction.

[0096] For example, as shown in Figure 2 The final traffic prediction result is the weighted sum of the prediction results of the Prophet model and the Xgboost model, and the formula is as follows:

[0097] Y(t) = w1y(t) + w2y'(t)

[0098] Wherein, Y(t) is the final traffic prediction result, w1 and w2 are weight parameters determined by model training, and are also the weights of the prediction results of the Prophet model and the Xgboost model. In the model training process, the values of w1 and w2 are determined according to the errors of the prediction results of the Prophet model and the Xgboost model.

[0099]

[0100] Wherein, ε1 is the average relative error of the Prophet model, and ε2 is the average relative error of the Xgboost model. The inverse error method is used to obtain the weight of the model prediction result, which can give a larger weight to the model with smaller average relative error, so that the average relative error of the combined model tends to decrease, thereby making the prediction result more accurate.

[0101] It can be understood that the final traffic prediction model is constructed in the manner of Prophet+Xgboost combination model, the prediction results of the two models are fused through the prediction manner of Prophet+Xgboost combination model, the reciprocal error method is used to give a larger weight to the model with a smaller average relative error, so that the average relative error of the combined model tends to decrease, the prediction results of the two models are complementary to each other, so that the prediction result is more accurate.

[0102] The Prophet+Xgboost combination model is actually an optimization of the Prophet model, which uses the characteristics of the Xgboost model to compensate for the shortcomings of the Prophet model. In the combined model, the Prophet model mainly considers the time correlation and has excellent adaptability to the trend change point and holiday effect of the data, especially the robustness to a large number of abnormal values, missing values and trend changes, which reduces the adverse effects of data volatility on prediction, while the Xgboost model can consider the influence of various factors on the predicted value, especially the characteristics of the data itself, so as to supplement the Prophet model.

[0103] It should be noted that after obtaining the real traffic value of the prediction time period, the original value and the predicted value of the business link traffic data are compared to post-evaluate the prediction accuracy of the Prophet+Xgboost combination model, determine the prediction error of the model, and verify the effectiveness of the model. If the prediction error is less than the preset error threshold, the model prediction accuracy meets the requirements, and the model is effective; otherwise, if the prediction error is greater than or equal to the preset error threshold, the model prediction accuracy does not meet the requirements, and the model is not effective, and the model parameters can be adjusted to retrain the model.

[0104] For example, in the scenario of predicting IP network VPN service traffic, in the training process of the Prophet+Xgboost combination model, historical two or three layer VPN service traffic data is collected to support the prediction of incoming historical data for a period of time, and a traffic baseline training model based on a certain VPN service link can be generated. Based on the traffic baseline training model based on a certain VPN service link generated in the model training stage, the traffic prediction model of each business link is constructed. Through training, after obtaining the trained traffic prediction model of the business link, a single business traffic data can be predicted, the traffic trend change is analyzed by inputting the historical business traffic data for a period of time, and the traffic data for a future period of time is predicted to support multi-granularity time series prediction, including minute, hour, day granularity prediction.

[0105] For example, Figure 3The flow prediction result is shown in the figure. The left curve represents the real data of the historical time series, and the right curve represents the predicted time series based on the historical time series.

[0106] In some embodiments, after obtaining the flow prediction result, the method further comprises: determining whether the flow prediction result is greater than a preset flow threshold; and performing an alarm operation when the flow prediction result is greater than the preset flow threshold.

[0107] That is, after obtaining the prediction result, the prediction result is compared with the normal flow threshold set for the link. If the flow threshold is exceeded, an alarm is given. This allows management and operation personnel to identify, warn and solve problems before the occurrence of traffic overload, avoids the impact of link congestion on user perception, and improves the availability of the network.

[0108] In the scenario of predicting IP network VPN service traffic, the present scheme realizes analysis of the future traffic growth trend of VPN services, supports VPN service traffic threshold judgment, allows management and operation personnel to identify, warn and solve problems before the occurrence of traffic overload, avoids the impact of link congestion on user perception, and improves the availability of the network. The technology has a wide coverage and can be applied in many scenarios. IP network VPN services are widely used, such as telecom operator value-added services, large enterprise and institution information networks, and special industry informatization construction. The technology is practical, simple and effective, convenient to use, and can be widely used in IP networks with VPN demand.

[0109] The embodiment improves the Prophet prediction model by weighting. In flow prediction, the trend factors are fitted in an adaptive weighting manner according to the correlation size between the trend factors and the service traffic, thereby improving the accuracy of flow prediction. The prediction results of the XGBoost prediction model and the Prophet prediction model are fused, further improving the accuracy of the prediction results.

[0110] The service flow prediction method of the embodiment of the present application predicts the service flow by using the first flow prediction model. In flow prediction, the first flow prediction model fits the trend factors corresponding to the multiple time sequence characteristics of the service flow and their respective weights. The multiple trend factors set their respective weights, and the corresponding trend factors are distinguished by the weights indicating the correlation size between the corresponding trend factors and the service flow, thereby improving the accuracy of flow prediction. The prediction results of the first flow prediction model and the second flow prediction model are fused, further improving the accuracy of the prediction results.

[0111] The implementation process of the first traffic prediction model based on the first historical traffic data to predict the traffic and obtain the first traffic prediction data is described in detail below. The embodiment provides another traffic prediction method, Figure 4 The flowchart of another traffic prediction method provided by the embodiment of the application is shown in Figure 4 The traffic prediction method can include the following steps:

[0112] Step 201, extracting the time sequence characteristics of the first historical traffic data to obtain prediction basis data; wherein the time sequence characteristics include a trend item, a period item, a holiday item and a history item.

[0113] It can be understood that based on the current time and the time granularity to be predicted, the historical traffic data for predicting the traffic of the future time or the future period is obtained, and the time sequence characteristics of the historical traffic data are extracted as the prediction basis data according to the time granularity to be predicted.

[0114] Step 202, determining the weight values of the plurality of trend factors based on the prediction basis data and the traffic data source; wherein the traffic data source includes the historical traffic and the index values corresponding to the historical traffic, the time sequence characteristics.

[0115] It should be noted that different ways are used to calculate the weight values of the trend factors with different characteristics, so that the calculated weight values are more consistent with the characteristics of the data itself. For example, the weight α of the trend item g(t), the weight β of the period item s(t) and the weight λ of the holiday item h(t) can be determined by calculating the information entropy, and the weight θ of the historical reference factor p(t) can be determined by the feature similarity method. The information entropy method and the feature similarity method are described in detail below.

[0116] As an implementation manner, for the first trend factors corresponding to the trend item, the period item and the holiday item, one of the first trend factors is determined as a target item, and the trend factors other than the target item in the first trend factors are determined as matching items; the first target traffic data with different target items and the same matching items is selected from the traffic data source; and the information entropy corresponding to the target item is calculated based on the first target traffic data; and the weight values of the trend factors in the first trend factors are determined based on the information entropy corresponding to each trend factor in the first trend factors.

[0117] It should be noted that the traffic data source is a traffic database, which is established based on the historical traffic data of each link, including the historical traffic, the index values of the historical traffic and the time sequence characteristics of the historical traffic, the time sequence characteristics including the holiday item, the trend item and the period item and a plurality of trend factors.

[0118] It can be understood that, for the weight α of the trend item g(t), the weight β of the periodic item s(t) and the weight λ of the holiday item h(t), the correlation between each trend factor and the business traffic is calculated based on the prediction basis data by using the entropy method, and the weight value of each trend factor is determined based on the correlation. Wherein, the greater the correlation with the business traffic, the greater the weight value of the trend factor.

[0119] In calculating the correlation between each trend factor and the business traffic: based on the prediction basis data, one of the trend factors is taken as a target item, and the other trend factors are taken as matching items, and historical traffic data with only the target item different and the other matching items same is selected from the traffic database by controlling variables; according to the historical traffic data matched from the traffic database, the information entropy of the business traffic is calculated, and the greater the information entropy value, the greater the correlation between the business traffic and the target item, and the greater the weight of the target item. Information entropy represents the order of data, and by controlling variables, the information entropy under the variable is calculated, which can reflect the correlation between the data and the variable to a certain extent. When the variable and the data have a large correlation, the probability of the data changing with the variable is large, the order of the data is low, the information entropy is large, and vice versa. In the extreme case, for example, the data and the variable have no correlation, and under the condition that other time series characteristics are the same, the data with different variable values should have basically the same trend or value, and the order of the data is high, and the information entropy is small.

[0120] In calculating the information entropy: for example, taking the periodic item as the target item and the trend item and the holiday item as the matching items, historical traffic data with the same trend item and holiday item and different periodic items is selected from the traffic database; the information entropy of the historical traffic data is calculated, and the information entropy is the information entropy corresponding to the periodic item. When the periodic item is different and the other trend factors are the same, the greater the information entropy of the business traffic, the greater the correlation between the business traffic and the periodic item.

[0121] In calculating the information entropy, the difference between the maximum value and the minimum value of the business traffic in the obtained historical traffic data is taken as the interval range of the information entropy, and adaptive feature value interval division is performed in a fixed interval size or a fixed interval number manner, the probability of each sub-interval falling into the data is calculated, and the corresponding information entropy is obtained. For example, the calculation process of the information entropy is as follows:

[0122] According to the matched historical traffic data, the feature value interval [a, b] of the periodic item is determined, the interval is divided according to the preset interval scale or the preset interval number, m feature value intervals are obtained, and then the information entropy IE corresponding to the periodic item is calculated according to the following formula:

[0123]

[0124]

[0125] wherein P i is the probability of the i-th characteristic value interval possessing data, n i is the amount of data possessed by the i-th interval, and N is the total amount of matched data.

[0126] According to the above manner, the information entropy of other trend factors is calculated respectively, and the weight of each trend factor is obtained based on the proportional relationship between the information entropy of each trend factor.

[0127] Therefore, when predicting the traffic, the information entropy is used to determine the correlation between different trend factors and the business traffic, so as to adaptively assign different weights to the trend factors, and improve the accuracy of the prediction result.

[0128] For example, the trend factor based on adaptive weight is used to realize the IP network VPN business traffic prediction, analyze the future traffic growth trend of VPN business, and support the threshold value judgment of VPN business traffic.

[0129] As an implementation manner, for the trend factor corresponding to the historical item, second target traffic data with a similarity to the prediction basis data satisfying a preset condition is obtained from a business traffic data source; and the trend factor corresponding to the historical item and the weight value are determined based on the second target traffic data.

[0130] As an implementation manner, the implementation manner of obtaining second target traffic data with a similarity to the prediction basis data satisfying a preset condition from a business traffic data source; includes:

[0131] The Euclidean distance between the first data in each historical traffic data in the business traffic data source and the corresponding data in the prediction basis data is calculated; wherein the first data is at least one of historical business traffic, time sequence characteristics and index value; and the second target traffic data with a similarity to the prediction basis data satisfying a preset condition is obtained based on the Euclidean distance.

[0132] For example, for the weight θ of the historical reference factor p(t), the historical reference data with the highest similarity to the prediction basis data is selected from the business traffic database, and the trend factor corresponding to the historical reference data is taken as the historical reference factor p(t).

[0133] In the process of obtaining the historical reference data with the highest similarity to the prediction basis data, the Euclidean distance of the prediction basis data and at least one of the service traffic, the trend factor and the index value of the historical traffic data in the traffic database is calculated, the historical reference data with the minimum Euclidean distance is selected as the reference basis, and the similarity of the prediction basis data to the reference basis is taken as the weight θ value of the historical reference factor p(t).

[0134] For example, the calculation formula of the Euclidean distance d of the trend factor is as follows:

[0135]

[0136] wherein x j is the value of the jth time sequence feature in the prediction basis data, y j is the value of the jth time sequence feature in the historical traffic data, and M is the total number of time sequence features in the trend factor.

[0137] Step 203: fusing the multiple trend factors and the corresponding weight values to obtain the first traffic prediction data.

[0138] After obtaining the multiple trend factors each corresponding to a weight value, the weight value is fused with the corresponding trend factor to obtain the first traffic prediction data.

[0139] The service traffic prediction method of the embodiment of the application calculates the weight values in different ways for different characteristic trend factors in the first traffic prediction model, so that the calculated weight values are more in line with the characteristics of the data itself, thereby further improving the prediction accuracy.

[0140] To implement the above-mentioned embodiment, the application further provides a service traffic prediction device. Figure 5 A structural schematic diagram of a service traffic prediction device provided by the embodiment of the application is shown in FIG. 1. Figure 5 As shown in the figure, the service traffic prediction device can include a data acquisition module 301, a first traffic prediction module 302, a second traffic prediction module 303 and a result fusion module 304.

[0141] The data acquisition module 301 is configured to acquire first historical service traffic data of a first time period.

[0142] The first traffic prediction module 302 is configured to predict the service traffic based on the first historical service traffic data through a first traffic prediction model to obtain first traffic prediction data; wherein the first traffic prediction model includes multiple trend factors and weights of the multiple trend factors respectively; the multiple trend factors correspond to multiple time sequence features of the service traffic respectively, and the weight is used to indicate the correlation size between the corresponding trend factor and the service traffic.

[0143] The second traffic prediction module 303 is configured to predict the service traffic based on the first historical service traffic data by using a second traffic prediction model, and obtain second traffic prediction data.

[0144] The result fusion module 304 is configured to fuse the first traffic prediction data and the second traffic prediction data by using a third model, and obtain a traffic prediction result.

[0145] Further, in a possible implementation manner of the embodiment of the present application, the first traffic prediction module 302 is specifically configured to:

[0146] extract time sequence features of the first historical service traffic data to obtain prediction basis data; wherein the time sequence features include a trend item, a periodic item, a holiday item and a history item;

[0147] determine weight values of the plurality of trend factors respectively based on the prediction basis data and a service traffic data source; wherein the service traffic data source includes historical service traffic and index values corresponding to the historical service traffic, time sequence features;

[0148] fuse the plurality of trend factors and the corresponding weight values to obtain the first traffic prediction data.

[0149] Further, in a possible implementation manner of the embodiment of the present application, when the first traffic prediction module 302 determines the weight values of the plurality of trend factors respectively based on the prediction basis data and the service traffic data source, the first traffic prediction module 302 is configured to:

[0150] for the first trend factor corresponding to the trend item, the periodic item and the holiday item, one of the first trend factor is determined as a target item, and the trend factors other than the target item in the first trend factor are determined as matching items;

[0151] select first target traffic data with different target items and the same matching items from the service traffic data source; and calculate information entropy corresponding to the target items based on the first target traffic data;

[0152] determine the weight values of the trend factors in the first trend factor based on the information entropy corresponding to the trend factors in the first trend factor.

[0153] Further, in a possible implementation manner of the embodiment of the present application, when the first traffic prediction module 302 determines the weight values of the plurality of trend factors respectively based on the prediction basis data and the service traffic data source, the first traffic prediction module 302 is configured to:

[0154] for the trend factor corresponding to the history item, obtain second target traffic data with a similarity to the prediction basis data satisfying a preset condition from the service traffic data source;

[0155] determine the trend factor corresponding to the history item and the weight value based on the second target traffic data.

[0156] Further, in a possible implementation of the embodiment of the application, when the first traffic prediction module 302 acquires the second target traffic data with a similarity to the prediction base data satisfying a preset condition from the service traffic data source, it is used for:

[0157] calculating the Euclidean distance between the first data in each historical traffic data in the service traffic data source and the corresponding data in the prediction base data; wherein the first data is at least one of historical service traffic, time sequence characteristics and index values;

[0158] based on the Euclidean distance, acquiring the second target traffic data with a similarity to the prediction base data satisfying a preset condition.

[0159] Further, in a possible implementation of the embodiment of the application, when the result fusion module 304 fuses the first traffic prediction data and the second traffic prediction data through a third model to obtain a traffic prediction result, it is used for:

[0160] calculating the weighted sum of the first traffic prediction data and the second traffic prediction data through the third model; wherein the weight value of the third model is obtained through model training;

[0161] taking the weighted sum as the traffic prediction result.

[0162] Further, in a possible implementation of the embodiment of the application, the device further comprises a warning module 305, which is used for:

[0163] judging whether the traffic prediction result is greater than a preset traffic threshold;

[0164] when the traffic prediction result is greater than the preset traffic threshold, performing a warning operation.

[0165] It should be noted that the foregoing explanation and description of the service traffic prediction method embodiment also applies to the service traffic prediction device of this embodiment, which will not be described here.

[0166] The service traffic prediction device of the embodiment of the application predicts service traffic through a first traffic prediction model. The first traffic prediction model fits the trend factors corresponding to the multiple time sequence characteristics of the service traffic and their respective weights when predicting traffic, and the multiple trend factors set their respective weights. The weights distinguish the trend factors by indicating the correlation between the corresponding trend factors and the service traffic, and improve the accuracy of traffic prediction. Meanwhile, the prediction result of the first traffic prediction model is fused with the prediction result of a second traffic prediction model, further improving the accuracy of the prediction result.

[0167] To achieve the above-mentioned embodiments, the present application further provides an electronic device, comprising: a processor, and a memory connected with the processor in communication; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to realize the method provided by the foregoing embodiments.

[0168] To achieve the above-mentioned embodiments, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to realize the method provided by the foregoing embodiments.

[0169] To achieve the above-mentioned embodiments, the present application further provides a computer program product, comprising a computer program, wherein the computer program is executed by a processor to realize the method provided by the foregoing embodiments.

[0170] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in the present application comply with relevant laws and regulations and do not violate public order and good customs.

[0171] It should be noted that the personal information from the user should be collected for legal and reasonable purposes, and should not be shared or sold outside these legal uses. In addition, such collection / sharing should be carried out after the user's informed consent is received, including but not limited to informing the user to read the user agreement / user notice before the user uses the function, and signing the agreement / authorization including authorization of relevant user information. In addition, any necessary steps should be taken to protect and ensure access to such personal information data, and to ensure that other people with access to personal information data comply with their privacy policy and processes.

[0172] The present application is expected to provide embodiments in which the user can selectively prevent the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or prevent access to such personal information data. Once the personal information data is no longer needed, the risk is minimized by limiting data collection and deleting data. In addition, such personal information is de-identified, if applicable, to protect the privacy of the user.

[0173] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0174] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0175] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0176] The logic and / or steps represented in flow diagrams or otherwise described herein, for example, can be considered as a sequence of executable instructions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a product of the manufacturing and / or processing. The computer-readable medium can include, but is not limited to, the following: an electronic connection (an electronic device having one or more wires), a portable computer diskette (a magnetic device), a RAM (random access memory), a ROM (read-only memory), an EPROM (erasable programmable ROM) or Flash memory, an optical fiber device, and a portable CD ROM. Additionally, the computer-readable medium can be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via, for example, an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in order to be executed.

[0177] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. As such, in some embodiments, specifically configured hardware can be used to implement at least some of the functionality described herein. For example, if implemented in hardware, the hardware can include any or a combination of the following: a discrete logic circuit having logic gates for implementing logic functions upon data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0178] Those of skill in the art would understand that information and signals can be represented using any of a variety of technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that can be referenced throughout the above description can be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0179] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing module, or each of the units can be physically present separately, or two or more units can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0180] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A business traffic prediction method, characterized in that, Includes the following steps: Obtain the first historical business traffic data for the first time period; The first traffic prediction model is used to predict the traffic based on the first historical traffic data to obtain the first traffic prediction data. The first traffic prediction model includes multiple trend factors and their respective weights. The multiple trend factors correspond to multiple time-series characteristics of the traffic, and the weights are used to indicate the correlation between the corresponding trend factors and the traffic. The second traffic prediction model is used to predict the traffic based on the first historical traffic data to obtain the second traffic prediction data. The traffic prediction results are obtained by fusing the first traffic prediction data and the second traffic prediction data using the third model. The step of predicting service traffic based on the first historical service traffic data using a first traffic prediction model to obtain first traffic prediction data includes: Extract the temporal features of the first historical business traffic data to obtain the prediction base data; wherein, the temporal features include trend items, periodic items, holiday items, and historical items; Based on the aforementioned prediction base data and business traffic data source, the weight values ​​of each of the multiple trend factors are determined; wherein, the business traffic data source includes historical business traffic and the corresponding indicator values ​​and time-series characteristics of the historical business traffic; By integrating the multiple trend factors and their corresponding weight values, the first traffic prediction data is obtained.

2. The method according to claim 1, characterized in that, The step of determining the weight values ​​of the multiple trend factors based on the prediction base data and business traffic data source includes: For the first trend factors corresponding to the trend item, cycle item and holiday item, one of the first trend factors is determined as the target item, and all trend factors other than the target item are determined as matching items. Select first target traffic data with different target items but the same matching items from the business traffic data source; and calculate the information entropy corresponding to the target item based on the first target traffic data; Based on the information entropy corresponding to each trend factor in the first trend factor, the weight value of each trend factor in the first trend factor is determined.

3. The method according to claim 1, characterized in that, The step of determining the weight values ​​of the multiple trend factors based on the prediction base data and business traffic data source includes: For the trend factor corresponding to the historical item, obtain the second target traffic data from the business traffic data source that meets the preset condition of similarity with the prediction base data; Based on the second target traffic data, determine the trend factor and weight value corresponding to the historical item.

4. The method according to claim 3, characterized in that, The step of obtaining second target traffic data from the business traffic data source that meets a preset condition for similarity with the prediction base data includes: Calculate the Euclidean distance between the first data in each historical traffic data in the business traffic data source and the corresponding data in the prediction base data; wherein, the first data is at least one of historical business traffic, time series features, and indicator values; Based on the Euclidean distance, second target traffic data that meets the preset conditions for similarity with the predicted basic data is obtained.

5. The method according to claim 3, characterized in that, The step of fusing the first traffic prediction data and the second traffic prediction data through a third model to obtain the traffic prediction result includes: The weighted sum of the first traffic prediction data and the second traffic prediction data is calculated using the third model; wherein the weight values ​​of the third model are obtained through model training. The weighted sum is used as the traffic prediction result.

6. The method according to claim 1, characterized in that, The method further includes: Determine whether the traffic prediction result is greater than a preset traffic threshold; An alarm is triggered when the predicted traffic volume exceeds a preset traffic volume threshold.

7. A business traffic prediction device, characterized in that, include: The data acquisition module is used to acquire the first historical business traffic data for the first time period; A first traffic prediction module is used to predict business traffic based on the first historical business traffic data using a first traffic prediction model to obtain first traffic prediction data. The first traffic prediction model includes multiple trend factors and their respective weights. Each trend factor corresponds to multiple time-series features of the business traffic, and the weights indicate the correlation between the corresponding trend factor and the business traffic. The process of predicting business traffic based on the first historical business traffic data using the first traffic prediction model to obtain the first traffic prediction data includes: extracting time-series features from the first historical business traffic data to obtain prediction base data; wherein the time-series features include trend items, periodic items, holiday items, and historical items; determining the weight values ​​of each of the multiple trend factors based on the prediction base data and the business traffic data source; wherein the business traffic data source includes historical business traffic and corresponding indicator values ​​and time-series features; and fusing the multiple trend factors and their corresponding weight values ​​to obtain the first traffic prediction data. The second traffic prediction module is used to predict business traffic based on the first historical business traffic data using the second traffic prediction model, and obtain the second traffic prediction data. The result fusion module is used to fuse the first traffic prediction data and the second traffic prediction data through a third model to obtain the traffic prediction result.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-6.

Images