A method and device for generating a communication network indicator parameter prediction model

Abstract

The present invention discloses a method and device for generating a communication network indicator parameter prediction model, which is used to solve the problem of difficulty in efficiently and accurately predicting communication network anomalies. This solution includes: obtaining historical parameters of multiple indicators of a target communication network within a historical period; classifying the historical parameters according to the parameter values ​​and corresponding timestamps to obtain classification labels corresponding to the parameter values; determining the associated characteristic values ​​of the multiple indicators based on the changing trends of the historical parameters of the multiple indicators within the historical period; using the classification labels corresponding to the parameter values ​​as training labels, using the associated characteristic values ​​as training characteristic values, and training a prediction model based on the historical parameters of the multiple indicators. This solution can achieve effective prediction before anomalies occur in the communication network and improve prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the field of communications, and in particular to a method and device for generating a communication network indicator parameter prediction model. Background Art

[0002] In the field of communications technology, there are numerous causes of network anomalies, and a wide variety of indicators reflect network status. After a network failure has occurred, technicians often infer the cause based on network connectivity and indicator parameters, then troubleshoot and repair the network. This method of repairing the network is time-consuming and often takes time, preventing users from using the network normally.

[0003] How to efficiently and accurately predict communication network anomalies is the technical problem to be solved by this application. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a method and device for generating a communication network indicator parameter prediction model to solve the problem of difficulty in efficiently and accurately predicting communication network anomalies.

[0005] In a first aspect, a method for generating a communication network indicator parameter prediction model is provided, comprising:

[0006] Acquire historical parameters of a plurality of indicators of a target communication network within a historical period, the historical parameters comprising a plurality of parameter values ​​and timestamps corresponding to the parameter values;

[0007] Classify the historical parameters according to the parameter values ​​and corresponding timestamps to obtain classification labels corresponding to the parameter values;

[0008] determining correlation characteristic values ​​of the multiple indicators based on the change trends of the historical parameters of the multiple indicators within the historical period, the correlation characteristic values ​​representing the correlation relationship between the multiple indicators, wherein the parameter of a first indicator among the multiple indicators is a superposition result of the parameters of multiple second indicators;

[0009] The classification label corresponding to the parameter value is used as the training label, and the associated characteristic value is used as the training characteristic value. The prediction model is trained based on the historical parameters of the multiple indicators. The trained prediction model is used to predict the parameters of the target indicator in the second time period based on the parameters of the input target indicator in the first time period. The second time period is the time period after the first time period.

[0010] In a second aspect, a device for generating a communication network indicator parameter prediction model is provided, comprising:

[0011] An acquisition module, configured to acquire historical parameters of a plurality of indicators of a target communication network within a historical period, wherein the historical parameters include a plurality of parameter values ​​and timestamps corresponding to the parameter values;

[0012] a classification module, classifying the historical parameters according to the parameter values ​​and corresponding timestamps, and obtaining classification labels corresponding to the parameter values;

[0013] a determination module, configured to determine correlation characteristic values ​​of the multiple indicators based on a change trend of historical parameters of the multiple indicators within the historical period, the correlation characteristic values ​​representing a correlation relationship between the multiple indicators, wherein a parameter of a first indicator among the multiple indicators is a superposition result of parameters of multiple second indicators;

[0014] A training module uses the classification label corresponding to the parameter value as a training label and the associated characteristic value as a training characteristic value, and trains a prediction model based on the historical parameters of the multiple indicators. The trained prediction model is used to predict the parameters of the target indicator in the second time period based on the parameters of the target indicator in the input first time period, and the second time period is the time period after the first time period.

[0015] In a third aspect, an electronic device is provided, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the method of the first aspect when executed by the processor.

[0016] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method of the first aspect are implemented.

[0017] In an embodiment of the present application, historical parameters of multiple indicators of a target communication network within a historical period are obtained, the historical parameters including multiple parameter values ​​and timestamps corresponding to the parameter values; the historical parameters are classified according to the parameter values ​​and the corresponding timestamps to obtain classification labels corresponding to the parameter values; the associated characteristic values ​​of the multiple indicators are determined according to the changing trends of the historical parameters of the multiple indicators within the historical period, the associated characteristic values ​​characterizing the association relationship between the multiple indicators, wherein the parameters of a first indicator among the multiple indicators are the superposition results of the parameters of multiple second indicators; the classification labels corresponding to the parameter values ​​are used as training labels, and the associated characteristic values ​​are used as training characteristic values, and a prediction model is trained based on the historical parameters of the multiple indicators. The trained prediction model is used to predict the parameters of the target indicator in a second period based on the parameters of the input target indicator in the first period, the second period being the period after the first period, so that effective prediction can be achieved before the anomaly occurs. Prediction of communication network anomalies based on the association relationship between multiple indicators can effectively improve prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0019] Figure 1 This is one of the flow charts of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0020] Figure 2 This is the second flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0021] Figure 3 This is the third flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0022] Figure 4 This is the fourth flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0023] Figure 5 This is the fifth flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0024] Figure 6 This is the sixth flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0025] Figure 7This is the seventh flow chart of a method for generating a communication network indicator parameter prediction model according to an embodiment of the present invention.

[0026] Figure 8 It is a schematic diagram of the optimization process of an embodiment of the present invention based on a trained model in an application scenario.

[0027] Figure 9 It is a structural diagram of a device for generating a communication network indicator parameter prediction model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The figure numbers in this application are only used to distinguish the various steps in the scheme, and are not used to limit the execution order of the various steps. The specific execution order is subject to the description in the specification.

[0029] In the field of communications technology, historical data from communication indicators can be used to train prediction models. The resulting prediction models can then predict future communication data based on real-time communication data, enabling the prediction of communication anomalies in future time periods. However, various communication indicators are often interrelated, and the correlations between different indicators often vary under different conditions. If predictions are performed independently on multiple communication indicators, the resulting predictions are often inaccurate.

[0030] In order to solve the problems existing in the prior art, the embodiment of the present application provides a method for generating a communication network indicator parameter prediction model, such as Figure 1 Shown, including:

[0031] S11: Acquire historical parameters of a plurality of indicators of a target communication network within a historical period, where the historical parameters include a plurality of parameter values ​​and time stamps corresponding to the parameter values.

[0032] In this step, historical parameters of multiple indicators of the communication network within a historical period are obtained. The above indicators may include, for example, a call bill establishment success rate, communication traffic, etc., which may be related to the load of the communication network.

[0033] The length of the historical period may be, for example, two months or longer. A longer historical period can more clearly reflect the changing trends of the parameters of multiple indicators and the correlation between multiple indicators, thereby improving the prediction accuracy of the trained model.

[0034] The labels carried by the above-mentioned historical parameters can indicate whether the corresponding historical parameters are normal. For example, the historical parameters of a certain indicator can be compared with the preset parameter range corresponding to the indicator. If the historical parameter is within the preset parameter range, the label corresponding to the historical parameter is determined to be a label indicating normal data; otherwise, the corresponding label is a label indicating abnormal data. It should be understood that different indicators can correspond to different preset parameter ranges, and the preset parameter ranges can be manually pre-set or automatically generated based on historical parameters.

[0035] S12: Classify the historical parameters according to the parameter values ​​and corresponding timestamps to obtain classification labels corresponding to the parameter values;

[0036] In this step, the parameter values ​​are first sorted based on their magnitude, and the timestamps are then classified based on the sorting results to obtain various time periods. Subsequently, the classification labels corresponding to the parameter values ​​are determined based on the classified time periods. The classification labels correspond to the data sets containing the parameter values ​​and the timestamps corresponding to the parameter values.

[0037] S13: Determine the associated characteristic values ​​of the multiple indicators based on the changing trends of the historical parameters of the multiple indicators within the historical period, wherein the associated characteristic values ​​represent the association relationship between the multiple indicators, wherein the parameters of the first indicator among the multiple indicators are the superposition result of the parameters of the multiple second indicators.

[0038] There is often a correlation between multiple indicators in the communication network. For example, when the communication network is operating normally, the call establishment success rate of the wireless base station is correlated with the business status and equipment status of the base station itself. In the case of a failure in the bearer network (such as a break in the optical cable), the call establishment success rate is correlated with the indicator status of the bearer network. In this step, the correlation characteristic value that characterizes the correlation between these indicators is determined based on the changing trend of the historical parameters of multiple indicators of the communication network within the historical period. Among them, the parameter of the indicator of the call establishment success rate of the wireless base station is the superposition result of the parameters of the two indicators of the business status and equipment status of the base station itself.

[0039] This step can determine the associated characteristic values ​​of multiple indicators. The associated characteristic values ​​can characterize the changing relationship of the parameters of the associated indicators. In the subsequent steps, they can be used as feature training models to improve the accuracy of the trained model in predicting the parameters of indicators in future time periods.

[0040] S14: Using the classification label corresponding to the parameter value as the training label and the association relationship as the training feature value, a prediction model is trained based on the historical parameters of the multiple indicators. The trained prediction model is used to predict the parameters of the target indicator in the second time period based on the input parameters of the target indicator in the first time period, where the second time period is the time period after the first time period.

[0041] The prediction model used in this step can be determined based on actual needs. For example, xgboost (eXtreme Gradient Boosting), gdbt (Gradient Boosting Decision Tree), neural networks, deep learning, etc. can be used. Optionally, a genetic algorithm can be used to tune the model parameters to obtain a model with the minimum mean absolute error (MAE). After the prediction model is trained, the parameters of the communication network indicators monitored in real time can be input into the trained prediction model, and the output results can represent the predicted parameters of the communication network indicators in the future time period.

[0042] Through the solution provided in the embodiment of the present application, the correlation relationship between the various indicators is first determined before training the prediction model, and then the prediction model is trained based on the correlation relationship. This enables the trained model to predict the parameters of the indicators in the future time period based on the correlation relationship between the various indicators, thereby effectively improving the prediction accuracy.

[0043] Because the data of various indicators in the communication network influence each other, the correlation between the indicators often varies when the communication network is busy or idle. To further improve the accuracy of the trained prediction model, the solution provided in the embodiment of the application performs refined partitioning of historical parameters to perform more accurate correlation analysis on historical parameters of different time periods.

[0044] Based on the solution provided in the above embodiment, optionally, the historical parameter includes multiple parameter values ​​corresponding to multiple time points within the historical period;

[0045] Among them, such as Figure 2 As shown, the above step S12 includes:

[0046] S21: Sort the multiple parameter values ​​in the historical parameters according to their size to obtain a sorting result.

[0047] For example, the number of successful creations of wireless call records for a single cell may be ranked, wherein each parameter value in the ranking result corresponds to a time point within a historical period.

[0048] S22: Classify the multiple time points corresponding to the multiple parameter values ​​according to the sorting result.

[0049] Each parameter value in the sorting result can represent the communication network status at the corresponding time point. For example, if the number of successful wireless call bill creation is greater than the preset success number, it indicates that the communication network status at the time point corresponding to the successful number of wireless call bill creation is a busy state. If the number of successful wireless call bill creation is less than or equal to the preset success number, it indicates that the communication network status at the time point corresponding to the successful number of wireless call bill creation is an idle state.

[0050] S23: Divide the historical period into multiple historical sub-periods according to the classified multiple time points.

[0051] Typically, because the communication habits of users in different regions are generally common, the state of the communication network has a certain degree of continuity. For example, most users create wireless call bills after get off work, resulting in a higher number of successful wireless call bill creations between 7:00 PM and 9:00 PM. Most users go to bed before midnight, resulting in a lower number of successful wireless call bill creations between 12:00 AM and 6:00 AM. Because the communication habits of users in different regions vary, the solution provided in the embodiments of the present application classifies corresponding time points based on historical parameters, which can be used to identify the idle and busy times of the communication network. Further correlation analysis can then be performed on the historical parameters of the communication network in different states, thereby improving the prediction accuracy of the trained prediction model.

[0052] It should be understood that in the solution provided in the embodiment of the present application, multiple historical sub-periods can be divided according to actual needs. For example, three categories of historical sub-periods corresponding to busy hours, normal hours, and idle hours can be obtained, or more categories of historical sub-periods can be obtained to achieve fine division.

[0053] In addition, the above-mentioned historical sub-periods may include multiple discontinuous time periods. For example, the busy-hour historical sub-period may include 11:00 am to 1:00 pm and 7:00 pm to 9:00 pm. Furthermore, the multiple classified time points may be optimized based on continuity. For example, the parameter values ​​corresponding to most of the time points in the 7:00 to 8:00 period are greater than the preset success number, and only a very small number of time points correspond to parameter values ​​less than or equal to the preset success number. Then, 7:00 to 8:00 can be divided into busy-hour historical sub-periods. Among them, whether to perform the above-mentioned optimization can be determined based on the ratio of the parameter value greater than the preset success number to the parameter value less than or equal to the preset success number. For example, if the ratio of the parameter value greater than the preset success number to the parameter value less than or equal to the preset success number is greater than the preset ratio, the above-mentioned optimization is performed. The solution provided by the embodiment of the present application can improve the continuity of the historical sub-periods obtained by division and reduce the historical parameter processing load.

[0054] The above step S13 includes:

[0055] S24: Determine, according to the change trends of the historical parameters of the multiple indicators in the historical period, the associated characteristic values ​​of the multiple indicators in the multiple historical sub-periods.

[0056] Because the load of the communication network varies during busy and idle times, the correlation between various communication network indicators during busy and idle times is often different. The solution provided by the embodiments of the present application first classifies the corresponding time points according to the parameter values, and then determines the correlation characteristic values ​​of the indicators based on the parameter change trends within the historical sub-periods of different categories, which can improve the prediction accuracy of the subsequently trained prediction model.

[0057] It should be understood that the associated characteristic values ​​for an indicator can differ across different historical sub-periods. For example, the call record establishment success rate is closely related to traffic volume during off-peak historical sub-periods, but closely related to the indicator status of the bearer network during busy historical sub-periods. In this embodiment, an associated characteristic value corresponding to an indicator within one historical sub-period can have different associated characteristic values ​​for the same indicator within different historical sub-periods.

[0058] Based on the solution provided in the above embodiment, optionally, Figure 3 As shown, the above step S22 includes:

[0059] S31: Determine at least one quantile of the ranking result.

[0060] A quantile, also known as a quantile, refers to the value that divides the sorting result into several equal parts in the present embodiment. For example, the value that divides the sorting result into 4 equal parts is the quartile, and the value that divides the sorting result into 2 equal parts is the quintile. It can also be called the median. Taking the median as an example, the value in the middle of all parameter values ​​in the sorting result can be used as the median. If there are an even number of sorting results, the average of the two middle parameter values ​​can be taken as the median.

[0061] S32: Classify the multiple time points corresponding to the multiple parameter values ​​according to the size relationship between the multiple parameter values ​​and the at least one quantile.

[0062] In this step, the multiple time points corresponding to the parameter values ​​are categorized based on at least one quantile determined in the previous step. For example, based on the relationship between the parameter values ​​and the quartiles in the sorted results, the parameter values ​​in the sorted results are divided into four equal parts, each containing 25% of the parameter values. The multiple time points corresponding to the parameter values ​​are then further divided into four equal parts. The categorized time point sets are then sorted by the parameter values ​​corresponding to the time points in the time point set. These four time point sets, from smallest to largest, represent, in order, idle periods, relatively idle periods, relatively busy periods, and finally busy periods of the communication network.

[0063] In practical applications, the classified time point sets can be further clustered. For example, given the four time point sets sorted from smallest to largest according to the above rules, the time period corresponding to the first time point set can be determined as the idle time period, the second time point set as the normal time period, and the third and fourth time point sets combined as the busy time period.

[0064] It should be noted that in actual applications, parameter values ​​may be equal to quantile values. In this case, parameter values ​​equal to quantile values ​​can be classified according to preset rules. For example, parameter values ​​equal to quantile values ​​can be classified into a category smaller than the quantile value. Alternatively, parameter values ​​equal to the quantile value can be divided into two equal parts, with the time points corresponding to one parameter value being classified into a category smaller than the quantile value, and the time points corresponding to the other parameter value being classified into a category greater than the quantile value. The above-mentioned division rules for time points corresponding to parameter values ​​equal to the quantile value can be pre-set based on actual conditions.

[0065] In addition, in addition to the solution of performing division based on quantiles used in this embodiment, time points may also be classified based on normal distribution or other statistical algorithms.

[0066] Through the solution provided in this application, the corresponding time points can be divided into multiple categories based on the size relationship of the parameter values, thereby realizing the division of different state time periods of the communication network, and then the correlation relationship of each indicator can be determined for different states of the communication network, thereby improving the prediction accuracy of the subsequently generated prediction model.

[0067] Based on the solution provided in the above embodiment, optionally, Figure 4 As shown, the above step S24 includes:

[0068] S41: Determine correlation characteristics of the multiple indicators in the first type of historical sub-period based on the change trends of the historical parameters of the multiple indicators in the first type of historical sub-period, where the correlation characteristics of the indicators represent the characteristics that the parameters of the first indicator change following the parameters of the second indicator associated with the first indicator.

[0069] In the solution provided in the embodiment of the present application, the associated characteristic values ​​of multiple indicators can be determined for each type of historical sub-period. Taking the first type of historical sub-period as an example, multiple indicators often show a certain trend of change within the first type of historical sub-period, such as gradually rising, gradually falling, increasing in a short period of time, decreasing in a short period of time, etc. When the parameters of a certain indicator change, one or more indicators associated with the indicator will often change under the influence of the change in the parameter of the indicator. In this step, when the parameters of the second indicator change, the parameters of the associated first indicator will also change accordingly with the parameters of the second indicator.

[0070] It should be understood that the phrase "the first indicator parameter changes following the second indicator parameter" in this step means that the change in the first indicator parameter is caused by the change in the second indicator parameter. The trend and magnitude of the change in the first indicator parameter can differ from the trend and magnitude of the change in the second indicator parameter. For example, if the second indicator parameter suddenly increases within a short period of time, the first indicator parameter may then slowly decrease after the second indicator parameter increases.

[0071] In this step, the correlation characteristics of the indicators can be determined based on the magnitude and duration of changes in the parameters of multiple indicators within the first historical sub-period. Different indicator parameters have different properties. Some indicators may experience sudden changes, such as those representing communication network connectivity. Other indicators tend to change slowly, such as network traffic and call volume. The correlation characteristics determined in this step can characterize the magnitude and duration of changes in the corresponding indicators, and can be used to distinguish between indicators with sudden changes and those with slowly changing parameters.

[0072] S42: Determine correlation characteristic values ​​of the multiple indicators in the first type of historical sub-period according to the correlation characteristics of the multiple indicators in the first type of historical sub-period.

[0073] Based on the correlation characteristics of the various indicators determined in the above steps, correlation characteristic values ​​for multiple indicators within the first historical sub-period are further determined. For example, these correlation characteristic values ​​can indicate which second indicator parameter(s) will be affected by the change in the first indicator's parameter, or they can indicate the degree to which each second indicator parameter affects the first indicator's parameter. These correlation characteristic values ​​are used as features for model training in subsequent steps.

[0074] The solution provided by the embodiments of this application can determine the correlation characteristics of each indicator based on the trend of parameter changes of each indicator. Furthermore, based on the correlation characteristics of the indicators, the correlation characteristic values ​​of each indicator can be further determined, thereby improving the accuracy of the correlation characteristic values ​​in representing the characteristics of the indicators. Furthermore, by using these correlation characteristic values ​​as training feature inputs into the model, the prediction accuracy of the trained prediction model can be improved.

[0075] Based on the solution provided in the above embodiment, optionally, Figure 5 As shown, the correlation characteristics include linear correlation characteristics and mutation correlation characteristics.

[0076] The above step S41 includes:

[0077] S51: If the change trend of the historical parameter of the first indicator in the first historical sub-period is positively correlated or negatively correlated with the change trend of the historical parameter of the associated second indicator in the first historical sub-period, determine that the correlation characteristic of the first indicator is a linear correlation characteristic.

[0078] In this embodiment, the parameters of indicators with linear correlation characteristics tend to change slowly, and the changes in parameter values ​​have a certain degree of continuity. In this step, the changing trend of the historical parameters of the first indicator within the first historical sub-period refers to the overall changing trend of the historical parameters of the first indicator over a period of time, which may include a few fluctuation points that do not conform to the overall changing trend. For example, the historical parameters of the first indicator may increase overall over a certain period of time, but a very small number of parameter points may cause the historical parameters to decrease within a short period of time, resulting in parameter value fluctuations. In this case, the changing trend of the historical parameters of the first indicator can still be determined to be an upward trend.

[0079] In this step, whether the historical parameter of the first indicator is positively or negatively correlated with the historical parameter of the second indicator within the first historical sub-period can be determined based on the slope of the change in the historical parameter of the first indicator and the slope of the change in the historical parameter of the second indicator within the first historical sub-period. For example, if the historical parameter of the second indicator gradually increases, and the increase becomes increasingly larger. If the first indicator changes with the second indicator, and the historical parameter of the first indicator gradually decreases within the corresponding time period, and the decrease also becomes increasingly larger, then the correlation characteristic of the first indicator can be determined to be a linear correlation characteristic.

[0080] S52: If the first indicator does not have the linear correlation characteristic, and the mutation period of the historical parameter of the first indicator within the first type of historical sub-period is the same as the mutation period of the historical parameter of the associated second indicator within the first type of historical sub-period, then the correlation characteristic of the first indicator is determined to be a mutation correlation characteristic, wherein the difference between the maximum and minimum values ​​of the historical parameters within the mutation period is greater than the preset difference.

[0081] The duration of the mutation period described in this step is no longer than a preset time length, which can be manually set in advance. The indicators with linear correlation characteristics determined in the above steps are usually related to the service level of the communication network, while the parameters of the indicators with mutation correlation characteristics determined in this step do not change significantly when the parameters of other indicators increase or decrease slightly, but will mutate accordingly when other indicators mutate. For example, assuming that the first indicator is the call establishment success rate, when the communication network is normal, there is no obvious positive or negative correlation between the traffic volume of the communication network and the call establishment success rate. However, when a communication network failure occurs, the traffic volume of the communication network mutates, and the call establishment success rate also mutates. Through this step, it can be determined that the correlation characteristic of the call establishment success rate indicator is a mutation correlation characteristic.

[0082] The solutions provided by the embodiments of this application can further distinguish the correlation characteristics of various indicators in the communication network, and then determine correlation feature values ​​for indicators with different correlation characteristics. This allows the correlation feature values ​​to more accurately represent the corresponding indicators, and further, a prediction model trained using the correlation feature values ​​as training features can more accurately predict the parameters of the indicators.

[0083] Based on the solution provided in the above embodiment, optionally, Figure 6 As shown, the above step S42 includes:

[0084] S61: If the first indicator in the first historical sub-period has a linear correlation characteristic or a sudden correlation characteristic, determine the correlation characteristic value of the first indicator according to historical parameters of the first indicator and the associated second indicator in the first historical sub-period.

[0085] In this step, the corresponding correlation characteristic value is further determined for indicators with linear or sudden correlation characteristics. Specifically, it can be determined whether the trend of the first indicator parameter following the second indicator parameter within the first historical sub-period matches the correlation characteristic value of the first indicator. In other words, the consistency of the correlation characteristic value of the first indicator within the first historical sub-period is determined, and then the correlation characteristic value of the first indicator is further determined.

[0086] If the parameters of the first indicator show linear correlation characteristics in part of the first historical sub-period, and show mutation correlation characteristics in another part of the period, it means that the correlation characteristics of the indicator do not have obvious characteristics in the first historical sub-period, and it is difficult to determine a correlation characteristic value to characterize the change characteristics of the first indicator in this sub-period, and then the trained prediction model will find it difficult to achieve accurate prediction.

[0087] Therefore, in the solution provided in the embodiments of the present application, a correlation feature value is further determined for a first indicator having uniform correlation characteristics within a first type of historical sub-period. This determined correlation feature value accurately represents the variation characteristics of the first indicator within that historical sub-period. Furthermore, a prediction model trained using the correlation feature value as a training feature can more accurately predict the parameters of the first indicator.

[0088] Based on the solution provided in the above embodiment, optionally, Figure 7 As shown, before S61, it also includes:

[0089] S71: Determine data characteristics of the first indicator, where the data characteristics represent numerical distribution characteristics of the indicator data.

[0090] The data features described in this step are used to characterize the numerical distribution characteristics of the parameters of the first indicator, which may include continuous characteristics and classification characteristics. For example, the parameters of the indicators characterizing the number of users and the amount of business have a certain continuity, that is, they have continuous properties. The parameters of the indicators characterizing the on / off status of the communication network can be classified, for example, 0 represents that the network is unavailable and 1 represents that the network is available. Among them, the parameters of the indicators with classification properties can also be understood as indicators with discontinuous parameter values. For example, if the parameter value range of the indicator is less than or equal to 30 and greater than or equal to 50, then this indicator has a classification property.

[0091] The above step S61 includes:

[0092] S72: Analyze historical parameters of the first indicator within the first type of historical sub-period using a preset correlation analysis method corresponding to the data characteristics of the first indicator.

[0093] For indicators with different data characteristics, appropriate analysis methods can be used to improve the accuracy of the analysis results. The parameters of a first indicator may be the sum of the parameters of multiple second indicators. In this step, a pairwise correlation analysis can be performed on the multiple related indicators. The analysis results can indicate the strength of the correlation between the indicators.

[0094] Optionally, if the relationships between devices in the communication network are complex, the analysis in this step can be performed based on the topology of the communication network. For example, based on the data characteristics of the second indicator Y and the first indicator X, the analysis method shown in the following table can be used to perform the analysis in this step:

[0095]

[0096] This step can yield an analysis result that characterizes the strength of the correlation between indicators. Optionally, indicators with a correlation strength less than a preset strength are removed, leaving only indicators with strong correlations to improve the prediction accuracy of the training model.

[0097] S73: Determine the correlation weight of the first indicator according to the analysis result, where the correlation weight represents the strength of the correlation between the indicators.

[0098] Based on the analysis results obtained in the above steps, this step further determines the association weight of the first indicator, which is used as a training feature value to characterize the strength of the association between the first indicator and the associated second indicator. Typically, the association weight can be greater than 0 and less than or equal to 1. Indicators with sudden association characteristics generally have larger association weights. When an indicator with sudden association characteristics undergoes a sudden change, the associated indicators often also undergo a sudden change. For example, after a communication network failure, a large number of indicators related to communication data will undergo a sudden change.

[0099] The above step S13 includes:

[0100] S74: Using the labels carried by the historical parameters as training labels, using the associated weights of the multiple indicators as training feature values, and training a prediction model based on the historical parameters of the multiple indicators.

[0101] The correlation weights determined in the above steps can characterize the correlation between the associated indicators. The model trained with the correlation weights as feature values ​​can predict the indicator parameters based on the correlation strength of multiple associated indicators, effectively improving the prediction accuracy.

[0102] Based on the solution provided by the above embodiment, it is assumed that the trained model is xgboost, which uses a genetic algorithm to perform tuning on the model parameters. In actual application, the parameters of the indicators of the communication network in the first time period monitored in real time can be input into the trained model, and the model predicts the parameters of the indicators in the second time period in the future based on the input parameters. Subsequently, the prediction results output by the model are compared with the preset indicator interval. If the predicted parameters are within the preset indicator interval, it indicates that it is normal. If the predicted parameters exceed the preset indicator interval, it indicates that the indicator may be abnormal in the second time period, and then an alarm corresponding to the indicator that may be abnormal in the second time period can be issued, so that technical personnel can take preventive measures in advance for possible abnormalities and effectively avoid abnormalities.

[0103] Figure 8 This is a schematic diagram of an optimization process based on a trained model in an application scenario according to an embodiment of the present invention. Figure 8, further describing the solution provided by this embodiment. Based on the prediction model trained by this solution, in practical applications, performance indicators of the communication network can be collected in real time. Data preprocessing and performance indicator calculation are then performed on the collected real-time performance indicators, and the relevant parameters obtained from the processing and calculation are input into the trained prediction model. Dynamic threshold self-learning for outlier detection is then performed based on the trained prediction model. Model optimization is achieved by updating the anomaly detection model, and performance anomaly detection services are executed based on the model output results to determine which indicators may experience anomalies in future periods. If an indicator is likely to experience an anomaly, an anomaly alarm message is output to the alarm center. The alarm center can send feedback data to the model based on the alarm results to trigger model optimization. The solution provided by the embodiment of this application makes it possible to predict future indicator data based on the prediction model, issue timely alarms for indicators that may experience anomalies, and optimize the prediction model based on the feedback results, effectively achieving automatic prediction and alarming of communication network indicator parameters. Optionally, based on historical indicator data, AI algorithms such as statistical judgment, unsupervised and supervised classification can be used to automatically learn and model, integrating multiple models for comprehensive anomaly detection.

[0104] In order to solve the problems existing in the prior art, the embodiment of the present application also provides a communication network indicator parameter prediction model generation device 90, such as Figure 9 Shown, including:

[0105] An acquisition module 91 is configured to acquire historical parameters of a plurality of indicators of a target communication network within a historical period, wherein the historical parameters include a plurality of parameter values ​​and timestamps corresponding to the parameter values;

[0106] a classification module 92, classifying the historical parameters according to the parameter values ​​and corresponding timestamps, and obtaining classification labels corresponding to the parameter values;

[0107] a determination module 93 for determining correlation characteristic values ​​of the multiple indicators based on the change trends of the historical parameters of the multiple indicators within the historical period, wherein the correlation characteristic values ​​represent the correlation relationship between the multiple indicators, wherein the parameter of a first indicator among the multiple indicators is the superposition result of the parameters of multiple second indicators;

[0108] The training module 94 uses the classification label corresponding to the parameter value as the training label and the associated characteristic value as the training characteristic value, and trains a prediction model based on the historical parameters of the multiple indicators. The trained prediction model is used to predict the parameters of the target indicator in the second time period based on the parameters of the target indicator in the input first time period, and the second time period is the time period after the first time period.

[0109] The device provided by the embodiment of the present application obtains historical parameters of multiple indicators of a target communication network within a historical period, wherein the historical parameters include multiple parameter values ​​and timestamps corresponding to the parameter values; classifies the historical parameters according to the parameter values ​​and the corresponding timestamps to obtain classification labels corresponding to the parameter values; determines associated characteristic values ​​of the multiple indicators according to the changing trends of the historical parameters of the multiple indicators within the historical period, wherein the associated characteristic values ​​characterize the association relationship between the multiple indicators, wherein the parameters of a first indicator among the multiple indicators are the superposition result of the parameters of multiple second indicators; uses the classification labels corresponding to the parameter values ​​as training labels, and uses the associated characteristic values ​​as training characteristic values, and trains a prediction model based on the historical parameters of the multiple indicators, wherein the trained prediction model is used to predict the parameters of the target indicator in a second period based on the parameters of the input target indicator in the first period, wherein the second period is a period after the first period, thereby enabling effective prediction before the anomaly occurs. Prediction of communication network anomalies based on the association relationship between multiple indicators can effectively improve prediction accuracy.

[0110] Preferably, an embodiment of the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and runnable on the processor. When the computer program is executed by the processor, the various processes of the embodiment of the method for generating a communication network indicator parameter prediction model described above are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0111] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the various processes of the embodiment of the method for generating a communication network indicator parameter prediction model described above are implemented, and the same technical effects are achieved. To avoid repetition, the details are not described here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0112] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0113] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0114] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A method for generating a communication network indicator parameter prediction model, characterized in that: include: Acquire historical parameters of a plurality of indicators of a target communication network within a historical period, the historical parameters including a plurality of parameter values ​​and timestamps corresponding to the parameter values; Classify the historical parameters according to the parameter values ​​and corresponding timestamps to obtain classification labels corresponding to the historical parameters; determining correlation characteristic values ​​of the multiple indicators based on the change trends of the historical parameters of the multiple indicators within the historical period, wherein the correlation characteristic values ​​represent the correlation relationship between the multiple indicators, wherein the parameter of a first indicator among the multiple indicators is the superposition result of the parameters of multiple second indicators; Using the classification label corresponding to the parameter value as a training label and the associated feature value as a training feature value, a prediction model is trained based on the historical parameters of the multiple indicators, and the trained prediction model is used to predict the parameters of the target indicator in a second time period based on the input parameters of the target indicator in the first time period, where the second time period is a time period after the first time period; The historical parameter includes a plurality of parameter values ​​corresponding to a plurality of time points within the historical period; The historical parameters are classified according to the parameter values ​​and corresponding timestamps to obtain classification labels corresponding to the parameter values, including: Sorting multiple parameter values ​​in the historical parameters according to size to obtain a sorting result; classifying the multiple time points corresponding to the multiple parameter values ​​according to the sorting result; Dividing the historical period into multiple historical sub-periods according to the classified multiple time points; The determining of the associated characteristic values ​​of the multiple indicators according to the change trends of the historical parameters of the multiple indicators within the historical period includes: According to the change trends of the historical parameters of the multiple indicators in the historical period, the associated characteristic values ​​of the multiple indicators in the multiple types of historical sub-periods are respectively determined.

2. The method according to claim 1, wherein Classifying the multiple time points corresponding to the multiple parameter values ​​according to the sorting result includes: determining at least one quantile of the ranking result; The multiple time points corresponding to the multiple parameter values ​​are classified according to a size relationship between the multiple parameter values ​​and the at least one quantile.

3. The method according to claim 2, wherein Determining, according to the changing trends of the historical parameters of the multiple indicators in the historical period, the associated characteristic values ​​of the multiple indicators in the multiple historical sub-periods, respectively, includes: Determining, based on the changing trends of the historical parameters of the multiple indicators within the first historical sub-period, correlation characteristics of the multiple indicators within the first historical sub-period, the correlation characteristics of the indicators representing characteristics of changes in parameters of a first indicator following changes in parameters of a second indicator associated with the first indicator; Determine correlation characteristic values ​​of the multiple indicators in the first type of historical sub-period according to correlation characteristics of the multiple indicators in the first type of historical sub-period.

4. The method according to claim 3, wherein The correlation characteristics include linear correlation characteristics and mutation correlation characteristics; Determining the correlation characteristics of the multiple indicators in the first historical sub-period based on the change trends of the historical parameters of the multiple indicators in the first historical sub-period includes: If the change trend of the historical parameter of the first indicator in the first historical sub-period is positively correlated or negatively correlated with the change trend of the historical parameter of the associated second indicator in the first historical sub-period, then the correlation characteristic of the first indicator is determined to be a linear correlation characteristic; If the first indicator does not have the linear correlation characteristic, and the mutation period of the historical parameter of the first indicator within the first type of historical sub-period is the same as the mutation period of the historical parameter of the associated second indicator within the first type of historical sub-period, then the correlation characteristic of the first indicator is determined to be a mutation correlation characteristic, wherein the difference between the maximum and minimum values ​​of the historical parameters within the mutation period is greater than the preset difference.

5. The method according to claim 4, wherein Determining, according to the correlation characteristics of the multiple indicators in the first type of historical sub-period, the correlation characteristic values ​​of the multiple indicators in the first type of historical sub-period includes: If the first indicator in the first historical sub-period has a linear correlation characteristic or a sudden change correlation characteristic, the correlation characteristic value of the first indicator is determined according to historical parameters of the first indicator and the associated second indicator in the first historical sub-period.

6. The method according to claim 5, wherein Before determining the associated characteristic value of the first indicator according to the historical parameters of the first indicator and the associated second indicator in the first type of historical sub-period, the method further includes: Determining a data characteristic of the first indicator, wherein the data characteristic represents a numerical distribution characteristic of the data of the indicator; Determining the associated characteristic value of the first indicator according to historical parameters of the first indicator and the associated second indicator in the first type of historical sub-period includes: Analyzing historical parameters of the first indicator within the first type of historical sub-period using a preset correlation analysis method corresponding to the data characteristics of the first indicator; Determining a correlation weight of the first indicator based on the analysis result, wherein the correlation weight represents the strength of the correlation between the indicators; The method of using the classification labels corresponding to the parameter values ​​as training labels and the associated characteristic values ​​as training characteristic values, and training a prediction model based on the historical parameters of the multiple indicators, includes: The classification labels corresponding to the parameter values ​​are used as training labels, the association weights of the multiple indicators are used as training feature values, and a prediction model is trained based on the historical parameters of the multiple indicators.

7. A device for generating a communication network indicator parameter prediction model, characterized in that: include: An acquisition module, configured to acquire historical parameters of a plurality of indicators of a target communication network within a historical period, wherein the historical parameters include a plurality of parameter values ​​and timestamps corresponding to the parameter values; a classification module, classifying the historical parameters according to the parameter values ​​and corresponding timestamps, and obtaining classification labels corresponding to the parameter values; a determination module, configured to determine correlation characteristic values ​​of the multiple indicators based on a change trend of historical parameters of the multiple indicators within the historical period, the correlation characteristic values ​​representing a correlation relationship between the multiple indicators, wherein a parameter of a first indicator among the multiple indicators is a superposition result of parameters of multiple second indicators; a training module, using the classification labels corresponding to the parameter values ​​as training labels and the associated feature values ​​as training feature values, and training a prediction model based on the historical parameters of the multiple indicators, wherein the trained prediction model is used to predict the parameters of the target indicator in a second time period based on the input parameters of the target indicator in the first time period, where the second time period is a time period after the first time period; The historical parameter includes a plurality of parameter values ​​corresponding to a plurality of time points within the historical period; In the classification module, the historical parameters are classified according to the parameter values ​​and the corresponding timestamps to obtain classification labels corresponding to the parameter values, including: Sorting multiple parameter values ​​in the historical parameters according to size to obtain a sorting result; classifying the multiple time points corresponding to the multiple parameter values ​​according to the sorting result; Dividing the historical period into multiple historical sub-periods according to the classified multiple time points; In the determination module, determining the associated characteristic values ​​of the multiple indicators according to the change trends of the historical parameters of the multiple indicators within the historical period includes: According to the change trends of the historical parameters of the multiple indicators in the historical period, the associated characteristic values ​​of the multiple indicators in the multiple types of historical sub-periods are respectively determined.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the method according to any one of claims 1 to 6 when executed by the processor.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images