Method and device for predicting traffic of base station, and computer readable storage medium

Abstract

The invention provides a method and device for predicting the traffic of a base station and a computer readable storage medium. The method for predicting the traffic of the base station comprises the following steps: acquiring historical traffic data of the base station; and generating future traffic data of the base station using a predictor based on the historical traffic data, where the predictor is trained according to a first rule. The first rule comprises that N first training curves are clustered into M categories based on similarity, the N first training curves are graphical representations of traffic values of N first training data along with time change, and M is smaller than or equal to N; a representative curve of each category is screened out, and M pieces of first training data corresponding to each representative curve are used as M pieces of final training data; and training the predictor by using the M final training data.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method, device and non-transitory computer-readable storage medium for predicting the traffic of a base station. Background Art

[0002] A base station is a radio transceiver station that transmits information between a mobile communication exchange center and a mobile terminal in a certain radio coverage area. With the development of mobile communication network services, the deployment and management of mobile communication base stations have become particularly important. For deployed base stations, the data traffic and power consumption that need to be processed vary depending on the time. For example, during a period of time when the data traffic is very small, if the base station is always in a fully turned on state, it will actually cause a lot of power waste.

[0003] It is necessary to design a power saving strategy so that the base station can adapt to the changes in data traffic to flexibly control its opening or closing, so as to achieve the purpose of power saving. If a unified power saving strategy is used to manage all base stations, this is also problematic because there are a large number of base stations across the country, and the traffic patterns of base stations in different regions vary greatly. A unified power saving strategy will undoubtedly cause additional problems.

[0004] Therefore, there is an urgent need for a method that can adaptively predict the traffic of different base stations in different regions, so that the power saving methods of the base stations can be deployed and managed in advance based on the traffic prediction results of each base station, thereby saving electricity consumption of base stations across the country. Summary of the invention

[0005] This application is made in view of the above problems. In an exemplary aspect, the present disclosure provides a method for predicting the traffic of a base station, comprising: obtaining historical traffic data of the base station; and based on the historical traffic data, using a predictor to generate future traffic data of the base station, wherein the predictor is trained according to a first rule, and the first rule comprises: clustering N first training curves into M categories based on similarity, the N first training curves are graphical representations of traffic values ​​of N first training data changing over time, M is less than or equal to N; screening out representative curves of each category, and using the M first training data corresponding to each representative curve as M final training data; and using the M final training data to train the predictor.

[0006] In some embodiments, training the predictor using the M final training data includes: adding M representative curves corresponding to the M final training data to a curve library and setting a label for each representative curve; and training the predictor using the M final training data and their labels.

[0007] In some embodiments, based on the historical traffic data, using a predictor to generate future traffic data for the base station includes: determining from the curve library a curve that best matches the curve of the historical traffic data, the curve of the historical traffic data being a graphical representation of traffic values ​​in the historical traffic data changing over time; and generating future traffic data for the base station based on the historical traffic data and a label corresponding to the best matching curve.

[0008] In some embodiments, when the matching degree between the curve of the historical traffic data and the best matching curve is lower than a first threshold, the method further includes: performing scaling processing on the generated future traffic data.

[0009] In some embodiments, scaling the generated future traffic data includes: calculating the proportional value corresponding to each time point based on the curve of the historical traffic data and the best matching curve; and multiplying the generated future traffic data point by the proportional value of the corresponding time point to generate scaled future traffic data.

[0010] In some embodiments, when the degree of match between the curve of the historical traffic data and the best matching curve is lower than a second threshold, the method further includes: dividing the curve of the historical traffic data into multiple partial curves; for each of the multiple partial curves: determining from the curve library the curve that best matches the partial curve; and generating partial future traffic data of the base station based on the historical traffic data and the label corresponding to the best matching curve, and splicing multiple partial future traffic data corresponding to the multiple partial curves to generate the future traffic data of the base station.

[0011] In some embodiments, generating partial future traffic data of the base station based on the historical traffic data and the label corresponding to the best matching curve includes: generating first future traffic data of the base station based on the historical traffic data and the label corresponding to the best matching curve; and intercepting a part corresponding to the time period of the partial curve from the first future traffic data as the partial future traffic data.

[0012] In some embodiments, when the degree of match between the curve of the historical traffic data and the best matching curve is lower than a third threshold, the method further includes: identifying at least one segment in the curve of the historical traffic data, wherein the degree of match between the at least one segment and the corresponding segment of the best matching curve is lower than the fourth threshold; replacing each segment of the at least one segment with the corresponding segment of the best matching curve to generate corrected historical traffic data; and generating future traffic data of the base station based on the corrected historical traffic data and the label corresponding to the best matching curve.

[0013] In another exemplary aspect, the present disclosure provides an apparatus for predicting the traffic of a base station, comprising: a historical traffic data acquisition unit, for acquiring the historical traffic data of the base station; and a prediction unit, for generating future traffic data of the base station based on the historical traffic data, wherein the prediction unit is trained according to a first rule, the first rule comprising: clustering N first training curves into M categories based on similarity, the N first training curves being a graphical representation of traffic values ​​of N first training data changing over time, M being less than or equal to N; screening out representative curves of each category, and using the M first training data corresponding to each representative curve as M final training data; and training the predictor using the M final training data.

[0014] In yet another exemplary aspect, the present disclosure provides a non-transitory computer-readable storage medium having computer instructions stored thereon, wherein when the computer instructions are executed by a processor, the method for predicting the traffic of a base station as described above is performed. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram showing an example method for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0016] Figure 2 A schematic diagram of a first training method for a predictor for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0017] Figure 3 A schematic diagram of a second training method for a predictor for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0018] Figure 4 A schematic diagram showing an example method for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0019] Figure 5 A schematic diagram showing an example method for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0020] Figure 6 A schematic diagram of an example method for adjusting a prediction result of base station traffic according to an embodiment of the present disclosure is shown.

[0021] Figure 7 A schematic diagram of an example method for performing segmented prediction on base station traffic according to an embodiment of the present disclosure is shown.

[0022] Figure 8 A schematic diagram of an example method for preprocessing historical traffic data of a base station according to an embodiment of the present disclosure is shown.

[0023] Fig. 9 A schematic diagram is shown of performing power saving management on a base station according to predicted future traffic data of the base station according to an embodiment of the present disclosure.

[0024] Fig.10 A schematic diagram of an apparatus for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0025] Fig.11 A schematic diagram of a non-transitory computer-readable storage medium for predicting traffic of a base station according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0026] Preferred embodiments of the present disclosure will be described in more detail below.

[0027] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include other steps and / or omit certain steps.

[0028] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0029] It should be understood that the concepts of "first", "second", etc. mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0030] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0031] In order to flexibly control the turning on or off of a base station in accordance with changes in data traffic, thereby achieving the purpose of power saving, the present disclosure provides a method for predicting the traffic of a base station.

[0032] Predicting the traffic of base stations requires a predictor based on machine learning. In the traditional training process for the predictor, the training data is selected randomly. However, this method of randomly selecting training data often has the problems of low diversity and poor representativeness. For example, there may be a large number of repetitions in the randomly selected training data, which means that even after a long period of training with a large amount of training data, the resulting predictor still cannot learn enough comprehensive knowledge, resulting in a significant reduction in the prediction accuracy of the predictor. The usual way to solve this problem is to further expand the training data set, thereby increasing the amount of training and extending the training time.

[0033] In order to enable the training data set to include more diverse training data and at the same time reduce the amount of training data, thereby achieving more efficient training and obtaining a better predictor, the present disclosure first proposes the concept of training the predictor using screened representative training data.

[0034] Figure 1 A schematic diagram of a method for predicting future traffic data of a base station according to an embodiment of the present disclosure is shown. Figure 2 A schematic diagram of a first training method for a predictor for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0035] like Figure 1 As shown, the method for predicting the traffic of a base station proposed in the present disclosure includes: acquiring historical traffic data of the base station; and generating future traffic data of the base station using a predictor based on the historical traffic data.

[0036] Figure 1 The historical traffic data of the base station shown may be, for example, the historical traffic data of the base station in the past period of time, for example, the historical traffic data of the base station in the past week, the past month, or the past three months. In addition, the historical traffic data of the base station may be represented in the form of a sequence consisting of various time points and their corresponding traffic values, or in the form of a curve. Figure 1 2 shows the historical traffic data of the example in the form of a curve.

[0037] In addition, Figure 1 In the example, the future traffic data of the base station predicted by the predictor is represented by a dotted line following the curve of the historical traffic data. In addition, the predicted future traffic data of the base station can also be represented in the form of a sequence consisting of each time point and its corresponding traffic value.

[0038] Figure 1 The predictor shown is trained according to the first rule. Figure 2 A schematic diagram showing a method for training a predictor according to a first rule.

[0039] like Figure 2 As shown, the method for training a predictor according to the first rule can generally include three steps: clustering (step 1), screening representatives (step 2), and training (step 3).

[0040] In step 1, N first training curves are clustered into M categories based on similarity. The N first training curves are graphical representations of the flow values ​​of N first training data changing over time, and M is less than or equal to N. For example, the N first training data may correspond to flow data obtained from N base stations within a certain geographical range. Figure 2 For example, M=4, that is, N first training curves are clustered into 4 categories, each training curve in each category is similar, and different categories usually have different curve appearances.

[0041] For example, the Kmeans clustering method can be used to cluster the N first training data. Taking the number of classes as 4 as an example, first, in the first step, 4 first training curves can be randomly selected from the N first training curves as the initial center points of the four categories. Then, in the second step, the distances from all the first training curves to the respective center points are calculated, and each first training curve is classified into the category with the closest center point. Then, in the third step, for the four divided categories, the center point of each category is recalculated (for example, the mean curve of the first training curves in the same category can be used as the new center point). The above second and third steps can be repeated until the classification result no longer changes or the predetermined number of iterations is reached.

[0042] Optionally, any other method known in the art may be used to cluster the N first training curves.

[0043] In step 2, the representative curve of each category is screened out, and the M first training data corresponding to each representative curve is used as M (for example, 4 as shown in the figure) final training data. As shown in the figure, the most representative training curve (for example, the curve corresponding to base station 3, the curve corresponding to base station 8, the curve corresponding to base station 32, and the curve corresponding to base station 7) is selected from the 4 categories obtained by clustering, so that the training data corresponding to the curves of these 4 base stations are used as the final training data. This means that for each category, only one of the most representative training data is selected, thereby avoiding excessive training time caused by redundancy of the same type of training data.

[0044] For example, when the Kmeans clustering method as described above is used, for the final clustering result, the training curve closest to the final center point in each category can be determined as the representative curve of each category.

[0045] In step 3, the predictor is trained using the M (eg, 4) final training data selected in step 2.

[0046] exist Figure 2 In the example shown, it is assumed that there are 1000 first training data corresponding to 1000 base stations at the beginning, and it is assumed that the training time of each training data is Δ. In this case, if the predictor is trained directly with these 1000 first training data without screening, a total training time of 1000×Δ is required. In contrast, if Figure 2 The training method based on clustering and screening representative curves shown in the figure only needs 4 representative training data to train the predictor, and only 4xΔ training time is needed in total. In these two cases, the knowledge learned by the predictor is actually not much different, because there are a lot of similar and redundant data in the original 1000 first training data, but compared with the case where the training data is not clustered and screened, the training time after introducing clustering and screening is significantly reduced.

[0047] If you want the predictor to learn more knowledge, you can further expand the first training data, and then perform clustering and screening to learn more knowledge in less time.

[0048] Combination of the above Figure 1 and Figure 2 The present disclosure describes a first rule for streamlining training data based on clustering and screening processes and an example method for predicting base station traffic using a trained predictor.

[0049] In addition, in order to avoid knowledge confusion caused by conflicting waveforms or similar waveforms during the training process, and in order to more effectively locate the knowledge related to the historical data of the base station to be tested itself during the prediction process, the present disclosure also proposes another training rule, that is, adding labels to each training data to train the predictor during training, and using historical traffic data and corresponding labels as input data to predict future traffic data during prediction.

[0050] Figure 3 A schematic diagram of a second training method for a predictor for predicting traffic of a base station according to an embodiment of the present disclosure is shown. Figure 4 A schematic diagram of another example method for predicting the traffic of a base station according to an embodiment of the present disclosure is shown.

[0051] like Figure 3As shown, the predictor for predicting base station traffic in the present disclosure can also be trained according to the second rule, and training the predictor according to the second rule includes the following steps:

[0052] Step 1: Obtain K second training curves corresponding to the K second training data.

[0053] Step 2: Add K second training curves to the curve library, and set a label for each second training curve.

[0054] Step 3: Train the predictor using the K second training data and their labels.

[0055] Specifically, Figure 3 Four second training data corresponding to base station 1, base station 3, base station 4 and base station n are shown, that is, K = 4. And labels 1, 2, 3 and 4 are set for the four second training data corresponding to base station 1, base station 3, base station 4 and base station n respectively.

[0056] The label can be regarded as a unique ID for each training data, which is used to indicate the waveform shape corresponding to the training data, making it easier for the predictor to distinguish each training data during the training process. Therefore, when a label is added to each training data, the predictor can effectively distinguish different training data through the label during the training process, thereby helping the predictor learn more accurate knowledge and avoid knowledge confusion caused by conflicting waveforms or similar waveforms.

[0057] Accordingly, when using Figure 3 When the predictor is trained using the training method shown, the predictor can also more effectively locate the knowledge related to the historical data of the base station to be tested itself based on different labels during the prediction process, thereby predicting more accurate future traffic data.

[0058] For example, Figure 4 As shown, based on the historical traffic data, using the predictor to generate the future traffic data of the base station includes:

[0059] Step 1: Obtain a curve corresponding to the historical traffic data of the base station to be tested. For example, the curve of the historical traffic data of the base station to be tested is a graphical representation of the traffic values ​​in the historical traffic data changing over time.

[0060] Step 2: Determine the curve that best matches the curve of the historical traffic data of the base station to be tested from the curve library established during the training process, for example, the curve corresponding to label 4 as shown in the figure.

[0061] Step 3: Generate future traffic data of the base station to be tested based on the historical traffic data of the base station to be tested and a label (eg, label 4) corresponding to the most matching curve.

[0062] It should be noted that although the above is combined Figure 3 and 4 Describes the training and prediction methods for creating a library of curves and adding labels to the training data. This does not mean that the combination Figure 3 and 4 The training and prediction methods described must be independent of Figure 1 and 2 In fact, the above combination Figure 3 The training method described can also be used with Figure 2 For example, the N first training data may be clustered and screened to obtain M final training data. Then, the M final training data may be used as follows: Figure 3 As shown, create a curve library and train with labels, that is, Figure 2 The M final training data in is Figure 3 The K second training data in .

[0063] Specifically, for the M final training data, the corresponding curves can be added to the curve library and a corresponding label can be added to each curve, and then the base station can use the M final training data and their corresponding labels to train the predictor.

[0064] In this case, training the predictor using the M final training data may include the following steps:

[0065] First, add M representative curves corresponding to the M final training data to the curve library, and set a label for each representative curve. For example, return to reference Figure 2 , in such Figure 2 As shown, after selecting four representative curves (for example, the curves corresponding to base station 3, base station 8, base station 32 and base station 7) from the four categories obtained by clustering, these four representative curves can be added to the curve library, and labels 1, label 2, label 3 and label 4 can be set for these four representative curves respectively.

[0066] Then, the predictor may be trained using the M final training data (eg, data corresponding to base station 3, base station 8, base station 32, and base station 7) and their labels.

[0067] In the case of this superimposed training, it is equivalent to integrating the clustering step of training data, the screening step of representative curves, the creation step of curve library and the labeled training step, and finally only one training process is performed.

[0068] Accordingly, if the predictor creates a Figure 3 The curve library shown can be used in the prediction process Figure 4The prediction method with label matching processing shown is used to predict the traffic of the base station.

[0069] Combination of the above Figure 3 and Figure 4 The present disclosure describes a training method based on a curve library and labels and an example method for predicting base station traffic using a trained predictor.

[0070] In some cases, even if the training data screening method in the above embodiment and the labeled predictor are used to train, it cannot be guaranteed that all types of traffic curves are included in the training set, and thus it cannot be guaranteed that the predictor can learn comprehensive knowledge.

[0071] In addition, even if there is a traffic curve in the training set that is similar to the future traffic data of the base station to be tested, it is impossible to make the prediction result completely coincide with the similar curve. Sometimes small differences in details may lead to a decrease in the overall prediction performance.

[0072] To solve this problem, the present disclosure further proposes a method for further adjusting the prediction result when the historical traffic data of the base station to be tested and the curve in the curve library have a poor match.

[0073] Figure 5 A schematic diagram of another method for predicting the traffic of a base station according to an embodiment of the present disclosure is shown. Figure 6 A schematic diagram of an example method for adjusting a prediction result of base station traffic according to an embodiment of the present disclosure is shown.

[0074] like Figure 5 As shown, assuming that the matching degree between the curve of the historical traffic data of the base station to be tested and the most matching curve (for example, the curve corresponding to label 2) is lower than the first threshold, if the method is directly used as shown in Figure 4 The labeled prediction method shown will result in a large deviation between the predicted result (ie, future traffic data) and the true value ref, that is, the predicted result is inconsistent with the true value.

[0075] As shown in the figure, the first half of the result output by the predictor is relatively consistent with the true value ref. However, for the second half, there is a significant deviation between the two. This deviation is usually caused by the fact that the curves in the curve library fail to cover all types of curves. Due to the incompleteness of this knowledge coverage, the curve corresponding to the best label 2 matched in the prediction process cannot actually reflect the actual historical traffic data of the base station to be tested, which will further lead to deviations in the prediction results.

[0076] In response to the deviation problem of prediction results as mentioned above, the present disclosure also proposes to scale the generated future traffic data when the degree of match between the curve of the historical traffic data of the base station to be tested and the most matching curve (for example, the curve corresponding to label 2) is lower than a first threshold, that is, to adjust the waveform of the future traffic data.

[0077] For example, the matching degree between the curve of the historical flow data of the base station to be tested and the best matching curve can be calculated by the difference between the flow values ​​at each time point between the two curves. For example, if the curve of the historical flow data of the base station to be tested and the best matching curve both have 100 corresponding flow values ​​at 100 time points, if the difference between the 100 corresponding flow values ​​between the two curves is all less than the error tolerance, the matching degree between the two can be determined as 100%. If the difference between only 50 corresponding flow values ​​between the two curves is less than the error tolerance, the matching degree can be determined as 50%, and so on.

[0078] For example, in this case, the first threshold may be set to 80%, that is, when the difference between less than 80 corresponding flow values ​​is smaller than the error tolerance, the generated future flow data is scaled.

[0079] In some examples, the waveform of the generated future traffic data can be adjusted based on the reduction process or the amplification process. Among them, the ratio used for the reduction process and the amplification process can be calculated by, for example, a formula obtained by summarizing experience, or can be obtained by finding a certain law using other statistical methods. For example, it can be generally considered that the base station is in a low-traffic state in the early morning hours and in a high-traffic state during the morning working peak hours. In this case, the traffic value corresponding to the early morning time of the generated future traffic data can be reduced by a certain proportion, and the traffic value corresponding to the morning working peak time of the generated future traffic data can be amplified by a certain proportion.

[0080] In another example, the generated future traffic data may also be scaled based on the ratio between the curve of the historical traffic data of the base station and the best matching curve. Figure 6 A schematic diagram of an example method for adjusting a prediction result of base station traffic according to an embodiment of the present disclosure is shown.

[0081] like Figure 6As shown, the corresponding ratio value A / A' between the curve of the historical traffic data of the base station to be tested and the most matching curve (for example, the curve corresponding to label 2) at each time point t can be calculated point by point to generate a scaling ratio dictionary including the corresponding ratio values ​​at each time point. Where A represents the traffic value of the curve of the historical traffic data of the base station to be tested at time t, and A' represents the traffic value of the most matching curve (for example, the curve corresponding to label 2) at the same time point t. For example, Figure 6 As shown, assuming that the curve of the historical flow data of the base station to be tested and the most matching curve (for example, the curve corresponding to label 2) both include 24 flow values ​​from 0 to 23 o'clock, the generated scaling ratio dictionary can include the respective ratio values ​​corresponding to 0 o'clock, 1 o'clock, ... 23 o'clock. Then, referring to the generated scaling ratio dictionary, the future flow data of the base station to be tested generated by the predictor can be multiplied point by point by the ratio value at the corresponding time point to generate scaled future flow data.

[0082] Combination of the above Figure 6 An embodiment of scaling the generated future traffic data is described when the matching degree between the curve of the historical traffic data of the base station to be tested and the best matching curve is not high. This is actually a post-processing method of correcting the prediction result when the matching degree is not high.

[0083] In some cases, if the matching degree between the curve of the historical traffic data of the base station to be tested and the best matching curve is not high, a segmented prediction method of segmentally matching the historical traffic data of the base station to be tested may also be considered.

[0084] Figure 7 A schematic diagram of an example method for performing segmented prediction on base station traffic according to an embodiment of the present disclosure is shown.

[0085] For example, Figure 7 As shown, when the matching degree between the curve of the historical flow data of the base station to be tested and the most matching curve is lower than the second threshold, the curve of the historical flow data can be first divided into multiple partial curves, that is, divided into multiple segments. For ease of description, Figure 7 An example of dividing the historical traffic data of the base station to be tested into two sections is shown. That is, if the historical traffic data of the base station to be tested includes the traffic values ​​of the past 24 hours, the curve of the first 12 hours (i.e. corresponding to 0-11 o'clock) is used as the first part of the curve, and the curve of the last 12 hours (i.e. corresponding to 12-23 o'clock) is used as the second part of the curve.

[0086] After the first and second partial curves are divided as shown in the figure, the curve that best matches each partial curve is determined from the curve library for each partial curve. For example, for the first partial curve of the first 12 hours (i.e., corresponding to 0-11 hours), the curve that best matches the label 1 is determined, and for the second partial curve of the last 12 hours (i.e., corresponding to 12-23 hours), the curve that best matches the label 2 is determined.

[0087] Then, for each partial curve, partial future flow data of the base station to be tested can be generated based on the historical flow data of the base station to be tested and the label corresponding to the most matching curve. That is, as shown in the figure, the historical flow data of the base station to be tested and label 1 can be input into the predictor to produce the first future flow data; and the historical flow data of the base station to be tested and label 2 can be input into the predictor to produce the second future flow data.

[0088] Then, a part corresponding to the time period of the first partial curve (0-11 o'clock) can be intercepted from the first future flow data as the first part of the future flow data generated for the first part of the curve; and a part corresponding to the time period of the second partial curve (12-23 o'clock) can be intercepted from the second future flow data as the second part of the future flow data generated for the second part of the curve.

[0089] Finally, multiple parts of future traffic data corresponding to multiple partial curves can be spliced ​​to generate future traffic data of the base station to be tested. Figure 7 In the example shown, the first part of future traffic data (0-11 o'clock) and the second part of future traffic data (12-23 o'clock) are spliced ​​together to form the final prediction result of the base station to be tested.

[0090] For example, Figure 7 As shown, when segmented prediction is used and concatenated to obtain the final prediction result, the final prediction result will be highly consistent with the true value ref.

[0091] It should be noted that for the convenience of description, Figure 7 The example of dividing the curve of the historical traffic data of the base station to be tested into two segments is shown in the example, which is only exemplary and not limiting. For example, according to actual conditions, the curve of the historical traffic data of the base station to be tested can be evenly or unevenly divided into more segments, such as 3 segments, 4 segments, 5 segments, etc., so as to achieve more accurate prediction.

[0092] In addition to the above-mentioned method of segmented prediction plus result splicing, in order to improve the prediction accuracy, it is also possible to consider preprocessing the input data (i.e., the historical traffic data of the base station to be tested) to make it more closely matched with the representative curve in the curve library. This is because even if the best matching curve is found in the curve library, there may still be a problem of low matching. For example, since only part of the historical traffic data may be similar to the best matching curve, while the other part is not similar, this may cause errors in the final prediction results.

[0093] To solve the above problem, in some embodiments, for example, if the matching degree between the curve of the historical flow data of the base station to be tested and the best matching curve is lower than the third threshold, at least one segment of the curve of the historical flow data can be identified, and the matching degree between the at least one segment and the corresponding segment of the best matching curve is lower than the fourth threshold. For example, the fourth threshold can be the same as or different from the third threshold.

[0094] After identifying at least one segment with a lower matching degree, each segment of at least one segment of the curve of the historical traffic data may be replaced with a corresponding segment of the best matching curve to generate corrected historical traffic data.

[0095] Then, you can refer to Figure 4 ,The corrected historical traffic data and the label of the best matching curve are used as input, and the future traffic data of the base station to be tested is generated.

[0096] Figure 8 A schematic diagram of an example method for preprocessing historical traffic data of a base station according to an embodiment of the present disclosure is shown.

[0097] For example, as shown in the figure, only the first half of the curve of the historical traffic data of the base station to be tested is similar to the most matching label 1 curve, and the second half is very different. In this case, the second half of the curve of the historical traffic data of the base station to be tested can be replaced with the second half of the curve of the most matching label 1, that is, the first half of the curve of the historical traffic data of the base station to be tested and the second half of the curve of the most matching label 1 are spliced ​​to obtain the corrected historical traffic data.

[0098] By splicing, we can obtain Figure 8 After the corrected historical traffic data is shown, compared with the original historical traffic data of the base station to be tested, the corrected historical traffic data is obviously closer to the most matching label 1 curve. After that, the corrected historical traffic data and label 1 can be input into the predictor together to obtain the prediction result of the future traffic data of the base station to be tested.

[0099] It should be noted that in Figure 8In the example, since the input data is corrected, that is, only the first half of the final input data is the real historical traffic data, in fact, only the corresponding first half of the prediction results output by the predictor are close to the real value. For the second half of the prediction results, since the corresponding part of the input data is corrected, there may be a significant difference between it and the real value. Therefore, in practical applications, it is usually necessary to extract the first half of the prediction results output by the predictor for use, while discarding the second half.

[0100] For the second half, a similar method can be used to obtain a more accurate prediction result. For example, a curve that best matches the second half of the curve of the historical traffic data of the base station to be tested (for example, the label 2 curve, not shown) can be found in the curve library. Assuming that only the second half of the historical traffic data of the base station to be tested is similar to the corresponding part of the label 2 curve, while the first half is not similar, the same method can be used to replace the first half of the curve of the historical traffic data of the base station to be tested with the corresponding first half of the label 2 curve to obtain the corrected historical traffic data. Afterwards, the corrected historical traffic data and label 2 can be input into the predictor, and then the second half of the prediction result can be extracted as the final prediction result of the second half.

[0101] Optionally, the first half and the second half extracted as above may be concatenated to obtain a complete prediction result.

[0102] It should be noted that Figure 8 The curve of the historical traffic data shown does not match the curve of label 1 in only the second half, which is merely illustrative. In practice, there may be a situation where multiple discrete segments in the curve of the historical traffic data do not match the curve of label 1. In this case, the multiple non-matching discrete segments can be replaced with the corresponding segments of the curve of label 1 by referring to the above method, thereby obtaining the corrected historical traffic data.

[0103] In the case of multiple mismatched segments, the prediction results can be extracted and spliced ​​according to similar methods as described above, so as to obtain a complete prediction result of the future traffic data of the base station to be tested.

[0104] Combination of the above Figures 1 to 8 Embodiments of various methods for predicting base station traffic of the present disclosure are described. It should be noted that the various training methods or prediction methods described above can be used in combination unless they are obviously inapplicable.

[0105] After the future traffic data of the base station is predicted according to the above various embodiments, power saving management of the base station can be performed based on the future traffic data. Fig. 9 A schematic diagram is shown of performing power saving management on a base station according to predicted future traffic data of the base station according to an embodiment of the present disclosure.

[0106] like Fig. 9 As shown, the power saving level of the base station, i.e., the ES level, can be determined according to the predicted future traffic data of the base station. Specifically, for example, the ES level of the base station can be determined from the provided ES level table by looking up the table. For example, if the predicted future traffic of the base station at time t is 20,000, it can be determined that the power saving level at time t is ES2, and then the base station can be switched on and off based on ES2 and the related policies set, so as to achieve the purpose of power saving.

[0107] The present disclosure further provides, in another aspect, a device for predicting traffic of a base station. Fig.10 A schematic diagram of an apparatus 1000 for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0108] like Fig.10 As shown, the apparatus 1000 includes a historical traffic data acquisition unit 1010 and a prediction unit 1020. The historical traffic data acquisition unit 1010 is used to acquire historical traffic data of a base station, and the prediction unit 1020 is used to generate future traffic data of the base station based on the historical traffic data.

[0109] The prediction unit 1010 is trained according to the first rule. Specifically, the first rule includes: clustering N first training curves into M categories based on similarity, wherein the N first training curves are graphical representations of flow values ​​of N first training data changing over time, and M is less than or equal to N; screening out representative curves of each category, and using the M first training data corresponding to each representative curve as M final training data; and training the predictor using the M final training data.

[0110] The first rule mentioned above is consistent with the Figure 2 The first rule in the method for training a predictor is similar.

[0111] In addition, the prediction unit 1010 can also Figure 3 The second rule described is used for training.

[0112] In addition, regarding Figures 1 to 8 The features of the various embodiments of the method for predicting the traffic of a base station are also applicable to the apparatus 1000 for predicting the traffic of a base station, unless obviously not applicable.

[0113] The present disclosure further provides, in another aspect, a non-transitory computer-readable storage medium for predicting traffic of a base station. Fig.11 A schematic diagram of a non-transitory computer-readable storage medium 1100 for predicting traffic of a base station according to an embodiment of the present disclosure is shown.

[0114] The non-transitory computer-readable storage medium 1100 stores computer program instructions 1110, wherein the computer program instructions 1110 are executed by the processor to perform the embodiments of the present disclosure. Figures 1 to 9 The various methods for predicting the traffic of a base station and the various methods for managing a base station described in the specification are not repeated here.

[0115] Those skilled in the art will appreciate that various aspects of the present disclosure may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks," "modules," "engines," "units," "components," or "systems." In addition, various aspects of the present application may be implemented as a computer product located in one or more computer-readable media, the product including computer-readable program code.

[0116] The present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0117] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined in this document.

[0118] The above is an explanation of the present disclosure and should not be considered as limiting it. Although several exemplary embodiments of the present disclosure are described, it will be easily understood by those skilled in the art that many modifications can be made to the exemplary embodiments without departing from the teachings and advantages of the present disclosure. Therefore, all these modifications are included in the scope of the present disclosure. It should be understood that the above is an explanation of the present disclosure and should not be considered as being limited to the specific embodiments disclosed.

Claims

1. A method for predicting the flow of a base station, comprising: Obtaining historical traffic data of the base station; as well as Based on the historical traffic data, using a predictor to generate future traffic data of the base station, wherein the predictor is trained according to a first rule, The first rule includes: Clustering N first training curves into M categories based on similarity, wherein the N first training curves are graphical representations of flow values ​​of the N first training data changing over time, and M is less than or equal to N; Screening out representative curves of each category, and using M first training data corresponding to each representative curve as M final training data; and The predictor is trained using the M final training data.

2. The method of claim 1, wherein: Training the predictor using the M final training data includes: Adding M representative curves corresponding to the M final training data to the curve library, and setting a label for each representative curve; and The predictor is trained using the M final training data and their labels.

3. The method of claim 2, wherein: Based on the historical traffic data, using a predictor to generate future traffic data of the base station includes: Determining from the curve library a curve that best matches the curve of the historical flow data, the curve of the historical flow data being a graphical representation of flow values ​​in the historical flow data changing over time; and Future traffic data of the base station is generated based on the historical traffic data and a label corresponding to the best matching curve.

4. The method of claim 3, wherein: In a case where the matching degree between the curve of the historical traffic data and the best matching curve is lower than a first threshold, the method further includes: Scaling is performed on the generated future traffic data.

5. The method of claim 4, wherein: Scaling the generated future traffic data includes: Calculating a ratio value corresponding to each time point based on the curve of the historical traffic data and the best matching curve; and The generated future traffic data is multiplied point by point by the scale value of the corresponding time point to generate scaled future traffic data.

6. The method of claim 3, wherein: In a case where the matching degree between the curve of the historical flow data and the best matching curve is lower than a second threshold, the method further includes: Dividing the curve of the historical flow data into a plurality of partial curves; For each of the plurality of partial curves: Determining a curve that best matches the partial curve from the curve library; and generating partial future traffic data of the base station based on the historical traffic data and the label corresponding to the best matching curve, and The multiple parts of future traffic data corresponding to the multiple partial curves are spliced ​​to generate the future traffic data of the base station.

7. The method of claim 6, wherein: Generating part of the future traffic data of the base station based on the historical traffic data and the label corresponding to the best matching curve includes: generating first future traffic data for the base station based on the historical traffic data and a label corresponding to the best matching curve; and A portion corresponding to the time period of the partial curve is intercepted from the first future flow data as the partial future flow data.

8. The method of claim 3, wherein: In a case where the matching degree between the curve of the historical flow data and the best matching curve is lower than a third threshold, the method further includes: identifying at least one segment of the curve of the historical flow data, wherein the degree of matching between the at least one segment and a corresponding segment of the best matching curve is lower than the fourth threshold; Replacing each of the at least one segment with the corresponding segment of the best matching curve to generate corrected historical flow data; The future traffic data of the base station is generated based on the corrected historical traffic data and the label corresponding to the best matching curve.

9. A device for predicting the flow of a base station, comprising: A historical traffic data acquisition unit, used to acquire the historical traffic data of the base station; as well as a prediction unit, configured to generate future traffic data of the base station based on the historical traffic data, The prediction unit is trained according to the first rule. The first rule includes: Clustering N first training curves into M categories based on similarity, wherein the N first training curves are graphical representations of flow values ​​of the N first training data changing over time, and M is less than or equal to N; Screening out representative curves of each category, and using M first training data corresponding to each representative curve as M final training data; and The predictor is trained using the M final training data.

10. A non-transitory computer-readable storage medium having computer instructions stored thereon, wherein: When the computer instructions are executed by a processor, the method for predicting the traffic of a base station according to any one of claims 1 to 8 is performed.

Images