A method, device, equipment and storage medium for monitoring data modality changes

Abstract

This application proposes a method, apparatus, device, and storage medium for monitoring data modal changes. The method includes: obtaining preset historical performance data; if the preset historical performance data is bi-periodic data, performing de-periodic preprocessing on the preset historical performance data; performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data; performing trend consistency determination on the change point based on the preset historical performance data in the left and right segments of each change point; wherein the preset historical performance data is segmented using the change point as a segmentation point; and filtering the change points that do not meet the trend consistency condition. This is used to accurately determine the modal changes of performance indicator data.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method, apparatus, device and storage medium for monitoring data modality changes. Background Art

[0002] As crucial infrastructure in the information age, wireless network systems must maintain continuous and stable operation to meet the communication needs of daily life, commerce, and public services. To this end, wireless communication equipment developers have designed a wide range of performance metrics to facilitate network operations and maintenance personnel in monitoring network components and network status. Currently, due to the slow trends in some performance data, it is difficult to effectively detect long-term changes using threshold methods. Using change point detection algorithms can be affected by trends and result in false detections. Summary of the Invention

[0003] The present application provides a method, apparatus, device and storage medium for monitoring data modal changes, so as to realize the monitoring of performance indicator modal changes.

[0004] The present invention provides a method for monitoring data modality changes, including:

[0005] Get preset historical performance data;

[0006] If the preset historical performance data is bi-periodic data, performing de-periodic preprocessing on the preset historical performance data;

[0007] performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data;

[0008] Performing trend consistency determination on each change point based on the preset historical performance data of the left and right segments of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point;

[0009] The change points that do not meet the trend consistency condition are filtered out.

[0010] An embodiment of the present application provides a device for monitoring data modality changes, including:

[0011] A data acquisition module is used to obtain preset historical performance data;

[0012] a de-periodicity processing module, configured to perform de-periodicity pre-processing on the preset historical performance data if the preset historical performance data is bi-periodic data;

[0013] a change point determination module, configured to perform change point detection on the preset historical performance data and determine a change point in the preset historical performance data;

[0014] a trend consistency determination module, configured to determine the trend consistency of each change point based on the preset historical performance data at the left and right ends of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point;

[0015] The change point filtering module is used to filter the change points that do not meet the trend consistency condition.

[0016] An embodiment of the present application provides a device, the device comprising a processor and a memory;

[0017] The processor is used to execute the program stored in the memory to implement any one of the methods in the embodiments of the present application.

[0018] An embodiment of the present application provides a storage medium storing a computer program. When the computer program is executed by a processor, any one of the methods in the embodiments of the present application is implemented.

[0019] The method, apparatus, device, and storage medium for monitoring data modal changes provided in the embodiments of the present application eliminate the influence of data periodicity and trends on data modal monitoring by de-periodic preprocessing of preset historical performance data, detecting change points, and identifying trends, thereby achieving accurate judgment of performance indicator modal changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the application scenario of the embodiment.

[0021] Figure 2 Flowchart of a method for monitoring data modality changes.

[0022] Figure 3 A schematic diagram of the structure of a device for monitoring data modality changes. DETAILED DESCRIPTION

[0023] To make the purpose, technical solutions and advantages of this application more clear, the embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other in any way.

[0024] Figure 1 FIG. 1 is a schematic diagram of an optional application scenario of an embodiment of the present invention, namely a Long Term Evolution (LTE) wireless network system. Figure 1As shown in the figure, a typical LTE wireless network system consists of multiple base stations, each of which is logically divided into several cells. Each cell hosts multiple mobile phone users, engaging in mobile communication activities such as making calls, browsing the internet, and watching videos. Key performance indicators (KPIs), such as the radio resource control (RRC) call drop rate, RRC connection establishment success rate, and average uplink and downlink traffic, reflect the operating status of the network system. Monitoring these performance indicators for long-term trends or modal data can help operations personnel manage the network.

[0025] Figure 2 A method for monitoring data modality changes according to an embodiment of the present application is shown, including:

[0026] S21. Obtaining preset historical performance data;

[0027] S22. If the preset historical performance data is bi-periodic data, perform de-periodic preprocessing on the preset historical performance data;

[0028] S23, performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data;

[0029] S24, determining trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point;

[0030] S25. Filter the change points that do not meet the trend consistency condition.

[0031] The preset historical performance data may be historical performance data reflecting the operating status of the wireless network system within a preset time period. These historical performance data are offline data, through which change points can be detected, and the recorded historical performance data often contain noise interference. The preset time period includes multiple weeks of historical performance data. Exemplarily, the preset time period is 30 days. In one implementation, obtaining the preset historical performance data includes: obtaining historical performance data within a preset time period; and performing denoising processing on the historical performance data using a preset filter to obtain the preset historical performance data. The historical performance data can be denoised using a preset filter. Optionally, a hamper filter can be used to perform denoising processing on the historical performance data to obtain the preset historical performance data.

[0032] In one implementation, if the preset historical performance data is bi-periodic data, performing de-periodic preprocessing on the preset historical performance data includes:

[0033] Performing a periodic data analysis on the preset historical performance data based on a discrete Fourier transform (DFT) algorithm to determine whether the preset historical performance data is bi-periodic data; wherein the bi-periodic data is data of a daily cycle and a weekly cycle;

[0034] If the preset historical performance data is bi-periodic data, calculating the average of the preset historical performance data on multiple specific days of the week;

[0035] The preset historical performance data on a specific day of each week is subtracted from the corresponding mean value.

[0036] The preset historical performance data is analyzed using a DFT algorithm for periodic data analysis to perform dual-cycle detection. If a dual-cycle (daily and weekly) signal is detected, the average of the preset historical performance data for multiple specific days of the week is calculated. The specific days of the week are Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, and Sunday. The average of the preset historical performance data for each day of the week is calculated. The preset historical performance data includes historical performance data for multiple weeks. That is, the average of the preset historical performance data for all Mondays, the average of the preset historical performance data for all Tuesdays, and so on is calculated. Then, the average of the preset historical performance data for each specific day of the week is subtracted from the preset historical performance data for the corresponding specific day of the week. That is, the average of the preset historical performance data for each Monday is subtracted from the preset historical performance data for each Monday, the average of the preset historical performance data for each Tuesday is subtracted from the preset historical performance data for each Tuesday, and so on, completing the operation of subtracting the average of the corresponding specific day of the week from all preset historical performance data. Because people's weekly work lives vary significantly between weekdays and weekends, and activity levels vary significantly at different times of the day, wireless network operations are affected, as reflected in the preset historical performance data. De-periodic preprocessing of this preset historical performance data can reduce the impact of weekend data on trend determination.

[0037] In one implementation, detecting a change point on the preset historical performance data to determine a change point in the preset historical performance data includes:

[0038] Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times;

[0039] If the number of steps of the change point detection reaches a preset number of steps, the number of change points detected at the same change point moment within the preset number of steps is compared with a first preset number of change points;

[0040] If the number of the change points at the same change point moment is greater than or equal to the first preset number of change points, the change point at the change point moment is retained, and the remaining change points are filtered.

[0041] Among them, the preset historical performance data can be scrolled and read with a preset time step S to detect change points, and the change point moments are recorded. After the preset number of steps T, the recorded change points are counted. Only when the change points detected within T steps meet a certain number condition, this change point is regarded as the real change point within the T-step time period, otherwise it is filtered out. In this implementation method, one day is set as the preset time step, and the preset number of steps is 7 steps, that is, the scroll detection is performed with 7 days as a cycle. When the 7-day cycle is met, the change points whose number of change points at the same change point moment is greater than or equal to the first preset number of change points are counted. In other words, if the number of change points at the same change point moment is less than the first preset number of change points, the change point at this change point moment will be filtered out. Exemplarily, the first preset number of change points can be 3 or 4.

[0042] In one implementation, determining the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point includes:

[0043] Segmenting the preset historical performance data based on the change point as a segmentation point;

[0044] Calculate the moving average of each segment of the preset historical performance data, where the period is the number of daily periodic data;

[0045] Performing linear regression on the moving mean of the left segment of each change point, the moving mean of the right segment of the change point, and the mean line of the moving mean of the left and right segments across the change point, respectively, and calculating the corresponding linear regression mean square error per unit length statistic;

[0046] If the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point are both less than the first threshold, and the mean square error of the linear regression of unit length statistics of the left and right segments of the change point is less than twice the largest of the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point, it is determined that the change point meets the trend consistency condition.

[0047] Among them, the preset historical performance data is segmented with the change point as the segmentation point, and thus segmented preset historical performance data is obtained. The preset historical performance data on the left segment of the current change point is the preset historical performance data between the adjacent change point on the left and the current change point, and the preset historical performance data on the right segment of the current change point is the preset historical performance data between the adjacent change point on the right and the current change point. For each change point, linear regression is respectively performed on the mean lines of the moving average of the left segment of the change point, the moving average of the right segment of the change point, and the moving average across the left and right segments of the change point, and the mean square error (MSE) lr_mse of the linear regression (Liner regression, LR) of the unit length statistic is calculated:

[0048]

[0049] Among them, y

[0048] , , , , <00001...is the value on the mean line, is the value on the regression fitting line, and len(bkp2 - bkp1) is the distance of the horizontal axis of the current fitting line. Specifically, when calculating lr_mse_left of the left segment of the change point, bkp1 is the current change point, and bkp2 is the first change point starting from the left of the current change point; when calculating lr_mse_rigiht of the right segment of the change point, bkp2 is the current change point, and bkp1 is the first change point starting from the right of the current change point; when calculating lr_mse_cross across the left and right segments of the change point, bkp2 is the first change point starting from the current change point, and bkp...

[0050] The trend consistency of the above-mentioned change points is judged as follows:

[0051] When lr_mse_left < TH and lr_mse_rigiht < TH and lr_mse_cross < 2 * MAX(lr_mse_left, lr_mse_rigiht) are satisfied, it is judged that this change point is a pseudo change point caused by trend influence and is filtered. Among them, the first threshold TH is a threshold set empirically, such as 0.01.

[0052] Calculate the moving average of each segment of the preset historical performance data, and judge the trend consistency of the change point according to the mean square error of the linear regression of the unit length statistic of the moving average of the left segment of the change point, the moving average of the right segment of the change point, and the moving average across the left and right segments of the change point. For the situation where the amplitude of the data change trend continuously increases, the effect of filtering the change point is better, and the judgment of the data modality change is more accurate.

[0053] In one implementation, detecting a change point on the preset historical performance data to determine a change point in the preset historical performance data includes:

[0054] Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times;

[0055] If the number of steps for detecting the change point reaches a preset number of steps, comparing the number of consecutive change points detected at the same change point within the preset number of steps with a second preset number of change points;

[0056] If the number of the continuous change points at the same change point moment is greater than or equal to the second preset number of change points, the change point at this change point moment is retained, and the remaining change points are filtered.

[0057] Among them, the preset historical performance data can be read in a rolling manner with a preset time step S to detect change points and record the time of the change points. After the preset number of steps T, the recorded change points are counted. Only when the change points detected within T steps meet a certain number condition, this change point is regarded as the real change point within the T-step time period, otherwise it is filtered out. In this implementation method, the preset time step is set to one day, and the preset number of steps is set to 7 steps, that is, the rolling detection is performed in a cycle of 7 days. When the 7-day cycle is met, the change points whose number of change points at the same change point moment is greater than or equal to the second preset number of change points are counted. In other words, it is determined whether the change points at the same change point moment appear continuously, and the number of consecutive appearances is greater than or equal to the second preset number of change points. If they are, they are determined to be change points, otherwise they are filtered out. Exemplarily, the second preset number of change points can be 3 or 4. If change points are detected to appear continuously at the same change point moment within the preset number of steps, and the number of continuous change points is greater than or equal to the second preset number of change points, the change points that appear continuously at this change point moment are retained; if change points are not detected to appear continuously at the same change point moment within the preset number of steps, the change points that appear discontinuously at this change point moment are filtered out; or, if change points are detected to appear continuously at the same change point moment within the preset number of steps, but the number of continuous change points is less than the second preset number of change points, the change points that appear continuously at this change point moment are filtered out.

[0058] In one implementation, determining the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point includes:

[0059] Segmenting the preset historical performance data based on the change point as a segmentation point;

[0060] For each of the above-mentioned change points, perform linear regression on the left-segment data of the change point, the right-segment data of the change point, and the data line across the left and right segments of the change point respectively, and calculate the corresponding linear regression mean square error of the unit length statistic;

[0061] If the linear regression mean square error of the unit length statistic of the left segment of the change point and the linear regression mean square error of the unit length statistic of the right segment of the change point are both less than the second threshold, and the linear regression mean square error of the unit length statistic across the left and right segments of the change point is less than twice the maximum of the linear regression mean square error of the unit length statistic of the left segment of the change point and the linear regression mean square error of the unit length statistic of the right segment of the change point, it is determined that the change point satisfies the trend consistency condition.

[0062] For each change point, find the linear regression line for the left-segment data of the change point, the right-segment data of the change point, and the data across the left and right segments of the change point respectively, and calculate the linear regression mean square error lr_mse of the unit length statistic:

[0063]

[0064] where y j is the original data value, is the value on the regression fitting line, and len(bkp2 - bkp1) is the distance of the horizontal axis of the current fitting line. Specifically, when calculating lr_mse_left of the left segment of the change point, bkp1 is the current change point and bkp2 is the first change point starting from the left of the current change point; when calculating lr_mse_rigiht of the right segment of the change point, bkp2 is the current change point and bkp1 is the first change point starting from the right of the current change point; when calculating lr_mse_cross across the left and right segments of the change point, bkp2 is the first change point starting from the left of the current change point and bkp1 is the first change point starting from the right of the current change point.

[0065] Determine the trend consistency of the above-mentioned change points as follows:

[0066] When lr_mse_left < TH and lr_mse_rigiht < TH and lr_mse_cross < 2 * MAX(lr_mse_left, lr_mse_rigiht) are satisfied, it is determined that the change point is a pseudo-change point caused by trend influence and is filtered. Here, the second threshold TH is an empirically set threshold, such as 0.02.

[0067] Judge the trend consistency of the change point according to the linear regression mean square error of the unit length statistic of the left-segment data, the right-segment data, and the data across the left and right segments of the change point. For the case where the amplitude of the data change trend remains stable, the effect of filtering the change point is better and the judgment of data modality change is more accurate.

[0068] Figure 3 A device for monitoring data modality changes is shown, comprising:

[0069] A data acquisition module 31 is used to acquire preset historical performance data;

[0070] a de-periodicity processing module 32 for performing de-periodicity pre-processing on the preset historical performance data if the preset historical performance data is bi-periodic data;

[0071] a change point determination module 33, configured to perform change point detection on the preset historical performance data and determine a change point in the preset historical performance data;

[0072] a trend consistency determination module 34 for determining the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point;

[0073] The change point filtering module 35 is configured to filter the change points that do not meet the trend consistency condition.

[0074] In one embodiment, the data acquisition module 31 includes:

[0075] A historical performance data acquisition unit, configured to acquire historical performance data within a preset time period;

[0076] The noise filtering unit is used to perform noise removal processing on the performance history data using a preset filter to obtain the preset historical performance data.

[0077] In one embodiment, the de-periodicity processing module 32 includes:

[0078] a periodic data determination unit, configured to perform periodic data analysis on the preset historical performance data based on a discrete Fourier transform (DFT) algorithm to determine whether the preset historical performance data is bi-periodic data; wherein the bi-periodic data is data of a daily cycle and a weekly cycle;

[0079] a weekly specific day data average calculation unit, configured to calculate an average of the preset historical performance data on multiple weekly specific days if the preset historical performance data is bi-periodic data;

[0080] The data de-periodization unit is used to subtract the corresponding mean value from the preset historical performance data on a specific day of each week.

[0081] In one implementation, the change point determination module 33 is specifically configured to:

[0082] Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times;

[0083] If the number of steps of the change point detection reaches a preset number of steps, the number of change points detected at the same change point moment within the preset number of steps is compared with a first preset number of change points;

[0084] If the number of the change points at the same change point moment is greater than or equal to the first preset number of change points, the change point at the change point moment is retained, and the remaining change points are filtered.

[0085] In one implementation, the trend consistency determination module 34 is specifically configured to:

[0086] Segmenting the preset historical performance data based on the change point as a segmentation point;

[0087] Calculate the moving average of each segment of the preset historical performance data, where the period is the number of daily periodic data;

[0088] Performing linear regression on the moving mean of the left segment of each change point, the moving mean of the right segment of the change point, and the mean line of the moving mean of the left and right segments across the change point, respectively, and calculating the corresponding linear regression mean square error per unit length statistic;

[0089] If the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point are both less than the first threshold, and the mean square error of the linear regression of unit length statistics of the left and right segments of the change point is less than twice the largest of the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point, it is determined that the change point meets the trend consistency condition.

[0090] In one implementation, the change point determination module 33 is specifically configured to:

[0091] Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times;

[0092] If the number of steps for detecting the change point reaches a preset number of steps, comparing the number of consecutive change points detected at the same change point within the preset number of steps with a second preset number of change points;

[0093] If the number of the continuous change points at the same change point moment is greater than or equal to the second preset number of change points, the change point at this change point moment is retained, and the remaining change points are filtered.

[0094] In one implementation, the trend consistency determination module 34 is specifically configured to:

[0095] Segmenting the preset historical performance data based on the change point as a segmentation point;

[0096] Performing linear regression on the left segment data of each change point, the right segment data of the change point, and the data line spanning the left and right segments data of each change point, and calculating the corresponding linear regression mean square error per unit length statistic;

[0097] If the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point are both less than the second threshold, and the mean square error of the linear regression of unit length statistics of the left and right segments of the change point is less than twice the largest of the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point, it is determined that the change point meets the trend consistency condition.

[0098] An embodiment of the present application provides a device, the device comprising a processor and a memory;

[0099] The processor is used to execute the program stored in the memory to implement any one of the methods in the embodiments of the present application.

[0100] An embodiment of the present application provides a storage medium storing a computer program. When the computer program is executed by a processor, any one of the methods in the embodiments of the present application is implemented.

[0101] The above description is merely an exemplary embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0102] It will be appreciated by those skilled in the art that the term user terminal covers any suitable type of wireless user equipment, such as a mobile phone, a portable data processing device, a portable web browser or a vehicle-mounted mobile station.

[0103] In general, various embodiments of the present application may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. For example, some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although the present application is not limited thereto.

[0104] Embodiments of the present application may be implemented by executing computer program instructions by a data processor of a mobile device, for example, in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages.

[0105] The block diagram of any logical flow in the accompanying drawings of the present application can represent program steps, or can represent interconnected logical circuits, modules and functions, or can represent a combination of program steps and logical circuits, modules and functions. The computer program can be stored on a memory. The memory can have any type suitable for the local technical environment and can be implemented using any suitable data storage technology, such as but not limited to read-only memory (ROM), random access memory (RAM), optical memory device and system (digital versatile disc DVD or CD optical disc) etc. Computer-readable media can include non-transient storage media. The data processor can be any type suitable for the local technical environment, such as but not limited to a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a programmable logic device (FPGA) and a processor based on a multi-core processor architecture.

[0106] The above description of exemplary embodiments of the present application has been provided by way of exemplary and non-limiting examples. However, various modifications and adaptations to the above embodiments will be apparent to those skilled in the art, when considered in conjunction with the accompanying drawings and claims, without departing from the scope of the present invention. Therefore, the proper scope of the present invention will be determined by reference to the claims.

Claims

1. A method for monitoring data modality changes, characterized in that: include: Acquire preset historical performance data; wherein the preset historical performance data is historical performance data reflecting the operating status of the wireless network system within a preset time period; If the preset historical performance data is bi-periodic data, performing de-periodic preprocessing on the preset historical performance data; wherein the bi-periodic data is data of daily period and weekly period; performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data; Performing trend consistency determination on each change point based on the preset historical performance data of the left and right segments of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point; The change points that do not meet the trend consistency condition are filtered out.

2. The method according to claim 1, characterized in that The obtaining of preset historical performance data includes: Get historical performance data within a preset time period; A preset filter is used to perform denoising processing on the performance history data to obtain the preset historical performance data.

3. The method according to claim 1, characterized in that If the preset historical performance data is bi-periodic data, performing de-periodic preprocessing on the preset historical performance data includes: Performing periodic data analysis on the preset historical performance data based on a discrete Fourier transform (DFT) algorithm to determine whether the preset historical performance data is bi-periodic data; If the preset historical performance data is bi-periodic data, calculating the average of the preset historical performance data on multiple specific days of the week; The preset historical performance data on a specific day of each week is subtracted from the corresponding mean value.

4. The method according to claim 1, wherein The performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data includes: Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times; If the number of steps of the change point detection reaches a preset number of steps, the number of change points detected at the same change point moment within the preset number of steps is compared with a first preset number of change points; If the number of the change points at the same change point moment is greater than or equal to the first preset number of change points, the change point at the change point moment is retained, and the remaining change points are filtered.

5. The method according to claim 4, characterized in that The step of determining the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point includes: Segmenting the preset historical performance data based on the change point as a segmentation point; Calculate the moving average of each segment of the preset historical performance data, where the period is the number of daily periodic data; Performing linear regression on the moving mean of the left segment of each change point, the moving mean of the right segment of the change point, and the mean line of the moving mean of the left and right segments across the change point, respectively, and calculating the corresponding linear regression mean square error per unit length statistic; If the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point are both less than the first threshold, and the mean square error of the linear regression of unit length statistics of the left and right segments of the change point is less than twice the largest of the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point, it is determined that the change point meets the trend consistency condition.

6. The method according to claim 1, characterized in that The performing change point detection on the preset historical performance data to determine the change point in the preset historical performance data includes: Reading the preset historical performance data in a rolling manner at a preset time step, gradually detecting change points of the preset historical performance data, and recording the detected change points and corresponding change point times; If the number of steps for detecting the change point reaches a preset number of steps, comparing the number of consecutive change points detected at the same change point within the preset number of steps with a second preset number of change points; If the number of the continuous change points at the same change point moment is greater than or equal to the second preset number of change points, the change point at this change point moment is retained, and the remaining change points are filtered.

7. The method according to claim 6, characterized in that The step of determining the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point includes: Segmenting the preset historical performance data based on the change point as a segmentation point; Performing linear regression on the left segment data of each change point, the right segment data of the change point, and the data line spanning the left and right segments data of each change point, and calculating the corresponding linear regression mean square error per unit length statistic; If the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point are both less than the second threshold, and the mean square error of the linear regression of unit length statistics of the left and right segments of the change point is less than twice the largest of the mean square error of the linear regression of unit length statistics of the left segment of the change point and the mean square error of the linear regression of unit length statistics of the right segment of the change point, it is determined that the change point meets the trend consistency condition.

8. A device for monitoring data modality changes, characterized in that: include: A data acquisition module is used to acquire preset historical performance data; wherein the preset historical performance data is historical performance data reflecting the operating status of the wireless network system within a preset time period; a de-periodicity processing module, configured to perform de-periodicity pre-processing on the preset historical performance data if the preset historical performance data is bi-periodicity data; wherein the bi-periodicity data is data of a daily cycle and a weekly cycle; a change point determination module, configured to perform change point detection on the preset historical performance data and determine a change point in the preset historical performance data; a trend consistency determination module, configured to determine the trend consistency of each change point based on the preset historical performance data of the left and right segments of each change point; wherein the preset historical performance data is segmented with the change point as a segmentation point; The change point filtering module is used to filter the change points that do not meet the trend consistency condition.

9. A device, characterized in that The device includes a processor and a memory; The processor is configured to execute a program stored in the memory to implement the method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images