Data monitoring method, device and equipment based on dynamic threshold value and storage medium

Abstract

The embodiment of the invention provides a data monitoring method and device based on a dynamic threshold value, equipment and a storage medium. According to the technical scheme provided by the embodiment of the invention, the fluctuation information, the seasonal change information and the holiday and festival index information are determined according to the monitoring index observation data, and the fitting degree information and the periodic information in various statistical periods are determined according to the monitoring index observation data and the monitoring index prediction data; an index portrait can be constructed according to fluctuation information, seasonal change information, holiday index information, fitting degree information and periodic information, a dynamic threshold determination algorithm is determined according to the index portrait, and a target dynamic threshold is determined based on the dynamic threshold determination algorithm and monitoring index observation data for data monitoring. And an applicable dynamic threshold determination algorithm can be determined according to the change rule and trend of the monitoring index data, so that the workload of manual operation and maintenance and parameter adjustment of data monitoring is effectively reduced, and the accuracy and adaptive capacity of data monitoring are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a data monitoring method, apparatus, device and storage medium based on a dynamic threshold. Background Art

[0002] With the rapid expansion of the Internet industry, many Internet companies have built huge business architectures, and these business systems have generated a large amount of diverse indicator data. For these business indicators, performance indicators, infrastructure indicators and other data, effective data monitoring and anomaly detection are required to ensure service stability and improve operation and maintenance efficiency.

[0003] Data monitoring is generally based on dynamic thresholds. Data monitoring solutions based on dynamic thresholds can effectively adapt to the rapidly changing business environment and the growing amount of data. The monitoring thresholds are dynamically updated during the service operation process, and the data is monitored according to the dynamic thresholds. The determination of dynamic thresholds is generally based on statistical methods, machine learning methods, deep learning methods, etc. Statistical methods use statistical principles (such as mean, variance, standard deviation, etc.) to dynamically calculate thresholds, but this method is not sensitive enough to the recognition of abnormal patterns, and the determination of thresholds is not accurate enough for non-normally distributed data; machine learning methods predict the normal range of indicators by training machine learning models and set thresholds accordingly. This method requires a large amount of historical data for training, and may not be able to accurately learn when the amount of data is insufficient; the time and computing cost of training models are high. Deep learning methods use deep learning technology to process complex time series data, automatically learn the inherent laws of data, and dynamically adjust thresholds, but this method may require a large amount of data and computing resources, may overfit for small-scale data sets, and parameter adjustment and model selection are relatively complex. Data monitoring requires a lot of manual operation and maintenance and frequent parameter adjustment, and data monitoring accuracy and adaptability are poor. Summary of the invention

[0004] The embodiments of the present application provide a data monitoring method, device, equipment and storage medium based on dynamic thresholds to solve the technical problems in related technologies that data monitoring requires a large amount of manual operation and maintenance and frequent parameter adjustments, and the data monitoring accuracy and adaptability are poor. The dynamic threshold can be determined more accurately, and the workload of manual operation and maintenance and parameter adjustment of data monitoring can be effectively reduced, thereby improving the accuracy and adaptability of data monitoring.

[0005] In a first aspect, an embodiment of the present application provides a data monitoring method based on a dynamic threshold, comprising:

[0006] Acquire monitoring indicator observation data and monitoring indicator forecast data, and determine fluctuation information, seasonal change information and holiday index information based on the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods based on the monitoring indicator observation data and the monitoring indicator forecast data;

[0007] Constructing an indicator portrait according to the fluctuation information, the seasonal change information, the holiday index information, the fitting degree information and the periodicity information, and determining a dynamic threshold determination algorithm according to the indicator portrait;

[0008] Determine a target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data;

[0009] Data monitoring is performed based on the target dynamic threshold.

[0010] In a second aspect, an embodiment of the present application provides a data monitoring device based on a dynamic threshold, including a data processing module, an algorithm determination module, a threshold determination module and a monitoring execution module, wherein:

[0011] The data processing module is configured to obtain monitoring indicator observation data and monitoring indicator prediction data, and determine fluctuation information, seasonal change information and holiday index information according to the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data;

[0012] The algorithm determination module is configured to construct an indicator portrait according to the fluctuation information, the seasonal change information, the holiday index information, the fitting degree information and the periodicity information, and determine a dynamic threshold determination algorithm according to the indicator portrait;

[0013] The threshold determination module is configured to determine a target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data;

[0014] The monitoring execution module is configured to perform data monitoring based on the target dynamic threshold.

[0015] In a third aspect, an embodiment of the present application provides a data monitoring device based on a dynamic threshold, including: a memory and one or more processors;

[0016] The memory is used to store one or more programs;

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement the data monitoring method based on dynamic threshold as described in the first aspect.

[0018] In a fourth aspect, an embodiment of the present application provides a non-volatile storage medium storing computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, are used to execute the dynamic threshold-based data monitoring method as described in the first aspect.

[0019] In the fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program stored in a computer-readable storage medium, and at least one processor of the device reads and executes the computer program from the computer-readable storage medium, so that the device performs the dynamic threshold-based data monitoring method as described in the first aspect.

[0020] The embodiments of the present application determine fluctuation information, seasonal change information and holiday index information based on monitoring indicator observation data, and determine fitting degree information and periodic information under multiple statistical periods based on monitoring indicator observation data and monitoring indicator prediction data. An indicator portrait can be constructed based on the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodic information, and a dynamic threshold determination algorithm can be determined based on the indicator portrait. A target dynamic threshold can be determined based on the dynamic threshold determination algorithm and the monitoring indicator observation data. Data monitoring can be performed based on the target dynamic threshold, and an applicable dynamic threshold determination algorithm can be determined based on the changing pattern and trend of the monitoring indicator data. The dynamic threshold can be determined more accurately, effectively reducing the workload of manual operation and maintenance and parameter adjustment of data monitoring, and improving the accuracy and adaptability of data monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of a data monitoring method based on a dynamic threshold provided in an embodiment of the present application;

[0022] Figure 2 is a flow chart of another data monitoring method based on dynamic thresholds provided in an embodiment of the present application;

[0023] Figure 3 It is a schematic diagram of an indicator portrait;

[0024] Figure 4 is a structural schematic diagram of a data monitoring device based on a dynamic threshold provided in an embodiment of the present application;

[0025] Figure 5 It is a structural schematic diagram of a data monitoring device based on a dynamic threshold provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application clearer, the specific embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. It is understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for the convenience of description, only the part related to the present application but not all the contents are shown in the accompanying drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow chart describes each operation (or step) as a sequential process, many of the operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. The above process can be terminated when its operation is completed, but it can also have additional steps not included in the accompanying drawings. The above process can correspond to a method, a function, a procedure, a subroutine, a subprogram, etc.

[0027] The data monitoring method based on dynamic thresholds provided in this application can be applied to the monitoring of indicator data and fault alarms. It aims to determine the applicable dynamic threshold determination algorithm according to the changing rules and trends of the monitored indicator data, determine the dynamic threshold more accurately, effectively reduce the manual operation and maintenance and parameter adjustment workload of data monitoring, and improve the accuracy and adaptability of data monitoring.

[0028] In the context of the rapid expansion of the Internet industry today, many Internet companies have built huge business architectures, and these business systems generate a large amount of diverse indicator data. For example, in live broadcast services, you may pay attention to key data such as the frequency of video freezes, the number of simultaneous online viewers, the CPU and memory usage of the server, network latency, and the number of request errors. For these business indicators, performance indicators, and infrastructure indicators, effective monitoring and anomaly detection mechanisms need to be implemented to ensure the stability of services, improve the efficiency of operation and maintenance, and prevent potential revenue losses. Dynamic thresholds are often used for data monitoring aggregation. Compared with traditional static thresholds, dynamic threshold algorithms can adapt to rapidly changing business environments and growing data volumes. Generally, machine learning algorithms are used to automatically adjust monitoring thresholds to reduce false positives and false negatives and improve operation and maintenance efficiency.

[0029] Existing data monitoring solutions based on dynamic thresholds are generally based on statistical methods, deep learning methods, game theory methods, etc. to determine dynamic thresholds. Statistical methods use statistical principles, such as mean, variance, standard deviation, etc., to dynamically calculate thresholds. This method is simple to implement, has a small amount of calculation, and can be quickly deployed, but may not be sensitive enough to identify abnormal patterns. For data with non-normal distribution, the threshold may not be accurate enough. Machine learning methods predict the normal range of indicators and set thresholds based on training machine learning models, such as SARIMA models and neural networks. This method can learn and predict future trends based on historical data, has strong adaptability and high accuracy, but requires a large amount of historical data for training, and may not be able to accurately learn when the amount of data is insufficient; the time and computational cost of training models are high. Deep learning methods use deep learning technologies, such as CNN, RNN, etc., to process complex time series data, automatically learn the inherent laws of data, and dynamically adjust thresholds. This method can process complex time series data, automatically learn the inherent laws of data, and is suitable for large-scale data sets, but requires a large amount of data and computing resources. It may overfit for small-scale data sets, and parameter adjustment and model selection are relatively complex. Game theory methods are often used to deal with complex network attacks, but their models are complex and computationally expensive. They are too complex for non-adversarial or non-competitive scenarios, and the determination of dynamic thresholds requires a lot of manual operation and maintenance and frequent parameter adjustments. At the same time, Internet companies usually have a large number of monitoring indicators, which requires dynamic threshold determination algorithms to efficiently process and analyze large-scale data sets. Model-based and machine learning algorithms consume a lot of computing power in calculations, resulting in a surge in computing power and electricity costs. Different monitoring indicators may have different trends and patterns. Some indicators may show obvious seasonality, while others may be affected by emergencies. A single algorithm cannot meet the data characteristics and behavioral characteristics of different indicators, resulting in the need for operation and maintenance personnel to make different sensitivity and algorithm parameter adjustments for different indicators, resulting in a lot of manpower costs. Even the same indicator may exhibit different behaviors in different time periods. For example, the number of visits to a website may be significantly different on weekdays, weekends, and holidays. The current dynamic threshold algorithm is difficult to capture this dynamic change in time series data and adjust the threshold accordingly, resulting in a high false alarm rate and poor data monitoring accuracy and adaptability. Based on this, a data monitoring method based on a dynamic threshold is provided in an embodiment of the present application to solve the technical problems that existing data monitoring requires a large amount of manual operation and maintenance and frequent parameter adjustments, and has poor data monitoring accuracy and adaptability.

[0030] Figure 1A flow chart of a data monitoring method based on a dynamic threshold provided in an embodiment of the present application is given. The data monitoring method based on a dynamic threshold provided in an embodiment of the present application can be executed by a data monitoring device based on a dynamic threshold. The data monitoring device based on a dynamic threshold can be implemented by hardware and / or software and integrated in a data monitoring device based on a dynamic threshold.

[0031] The following description is made by taking a data monitoring device based on a dynamic threshold to perform a data monitoring method based on a dynamic threshold as an example. Figure 1 , the data monitoring method based on dynamic threshold includes:

[0032] S110: Obtain monitoring indicator observation data and monitoring indicator forecast data, and determine fluctuation information, seasonal change information and holiday index information based on the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods based on the monitoring indicator observation data and monitoring indicator forecast data.

[0033] Exemplarily, the monitoring indicator observation data and monitoring indicator prediction data of the monitoring indicators that need to be monitored are obtained, wherein the monitoring indicator observation data can be understood as the real monitoring indicator data observed, monitored, and collected during the actual operation of the service, and the monitoring indicator prediction data can be understood as the monitoring indicator data predicted based on the monitoring indicator observation data. The monitoring indicator prediction data can be predicted based on the monitoring indicator observation data through a preset prediction algorithm or a trained machine learning model or deep learning model.

[0034] In one embodiment, the acquisition of monitoring indicator observation data and monitoring indicator prediction data can support different data sources and can also collect monitoring indicator observation data and monitoring indicator prediction data in a push and pull manner. For example, when the pull mode is adopted, the data collector and the external system are in a question-and-answer interaction mode, and the data collector sends a query request to the external system to obtain data. When the push mode is adopted, through one registration and one information return, the external system can continuously push data to the data collector according to the collection cycle specified by the registration, which can better adapt to the reporting of massive Internet indicators and reduce the system consumption caused by indicator reporting. Optionally, a reporting channel for the acquisition of monitoring indicator observation data and monitoring indicator prediction data can be provided, and users can push data to data monitoring devices, and the data is finally stored through time series databases such as clickhouse and victoriaMetrics. Or the user can give the path of the indicator, and the data monitoring device collects data regularly. Optionally, when collecting relevant indicator data, the data can be cleaned and preprocessed to ensure data quality.

[0035] Among them, there are corresponding monitoring indicator observation data and monitoring indicator prediction data, that is, at a time or time period, a monitoring indicator data has real monitoring indicator observation data and predicted monitoring indicator prediction data, for example, based on the previously collected monitoring indicator observation data, the monitoring indicator observation data corresponding to the monitoring indicator observation data will be observed, monitored, and collected at the corresponding time later. The monitoring indicator prediction data can be the same as the corresponding monitoring indicator observation data, or it can be different.

[0036] The monitoring indicator data (including monitoring indicator observation data and monitoring indicator prediction data) provided by this solution can correspond to one or more different monitoring indicators, such as the video freeze frequency of the monitoring indicator live broadcast service, the number of simultaneous online viewers, the CPU and memory usage of the server, network latency, and the number of request errors, etc. The monitoring indicator can be determined according to the object to be monitored. This solution can obtain the monitoring indicator observation data and monitoring indicator prediction data corresponding to the monitoring indicator for different monitoring indicators, and use the data monitoring method based on dynamic thresholds to determine the corresponding target dynamic threshold, and perform data monitoring of the corresponding monitoring indicator based on the target dynamic threshold.

[0037] In one embodiment, after obtaining the monitoring indicator observation data and the monitoring indicator prediction data, the fluctuation information, seasonal change information and holiday index information can be determined based on the monitoring indicator observation data, and the fitting degree information and the periodicity information under multiple statistical periods can be determined based on the corresponding monitoring indicator observation data and the monitoring indicator prediction data.

[0038] Among them, the fluctuation information can be used to reflect the overall stability and / or fluctuation of the monitoring indicator observation data. If the fluctuation of the monitoring indicator is too large and shows irregularity, it is not appropriate to apply the dynamic threshold; when the volatility of the monitoring indicator is small, it is more suitable to generate the threshold by simple statistical methods; when the monitoring indicator is at a medium level, the change of the data shows regularity, and it is suitable to use the dynamic threshold of the machine learning method. Seasonal change information can be used to reflect the fluctuation of the monitoring indicator observation data following the season, and the holiday index information can be used to reflect the fluctuation of the monitoring indicator observation data following the holidays. The degree of fit information can be used to reflect the overall degree of fit and similarity of the monitoring indicator prediction data and the monitoring indicator observation data, and the periodicity information can be used to reflect the periodicity of the monitoring indicator prediction data and the monitoring indicator observation data values ​​under the corresponding statistical period. Optionally, the statistical period can be month, week, day, etc.

[0039] S120: Construct an indicator portrait according to the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and determine a dynamic threshold determination algorithm according to the indicator portrait.

[0040] Exemplarily, an indicator portrait is constructed based on the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodic information determined above. The indicator portrait can reflect the impact of data fluctuation information, data seasonal changes, holidays, data fitting degree and data periodic fluctuations on monitoring indicators.

[0041] In one embodiment, after constructing the indicator portrait, a dynamic threshold determination algorithm can be determined based on the indicator portrait. Optionally, the indicator portrait can record fluctuation information, seasonal change information, holiday index information, fitting degree information, and periodicity information, and can also determine corresponding evaluation indexes according to fluctuation information, seasonal change information, holiday index information, fitting degree information, and periodicity information, and record the corresponding evaluation index as the indicator portrait. The indicator portrait can be recorded in a table or in a polygon, for example, the fluctuation information, seasonal change information, holiday index information, fitting degree information, and periodicity information are normalized, or the evaluation index is normalized, and the corresponding information or index is recorded at different corners in the polygon. For example, the corresponding information or index is normalized to 0-1, and the corresponding value changes from 0 to 1 from the center of the polygon to the corner.

[0042] In one embodiment, different indicator portraits of this solution correspond to different dynamic threshold determination algorithms. For example, the dynamic threshold determination algorithm can be a statistical method, a machine learning method, a deep learning method, a game theory method, a moving average method, an exponentially weighted moving average method, etc. The corresponding dynamic threshold determination algorithm can be configured in advance for different indicator portraits, indicator portrait ranges, or the ranking and proportion of different monitoring indicators in the indicator portrait, and the correspondence between the indicator portrait and the dynamic threshold determination algorithm can be determined. After the indicator portrait is constructed, the dynamic threshold determination algorithm corresponding to the current indicator portrait can be determined according to the correspondence.

[0043] S130: Determine a target dynamic threshold based on a dynamic threshold determination algorithm and monitoring indicator observation data.

[0044] Exemplarily, after determining the dynamic threshold determination algorithm, the monitoring indicator observation data (the latest monitoring indicator observation data corresponding to a preset time length or data volume) is analyzed and processed based on the determined dynamic threshold determination algorithm to obtain the target dynamic threshold, and the dynamic threshold of the monitoring indicator is generated in advance and reliably based on historical data and current data and by selecting a more appropriate dynamic threshold determination algorithm (statistical algorithm, machine learning algorithm, deep learning algorithm, etc.) to more accurately monitor and alarm the data.

[0045] S140: Perform data monitoring based on the target dynamic threshold.

[0046] Exemplarily, after determining the target dynamic threshold, data monitoring of the corresponding monitoring indicator can be performed based on the target dynamic threshold. Optionally, after abnormal data is found, an alarm can be issued based on the abnormal data. For example, when the monitoring indicator observation data collected in real time meets the alarm condition (for example, the monitoring indicator observation data exceeds or falls below the target dynamic threshold), it is determined that the corresponding monitoring indicator is abnormal, and relevant alarm operations can be performed.

[0047] For example, the data information flow (i.e., the monitoring indicator observation data collected in real time) can be consumed in real time, and judgment can be made by combining the data with the target dynamic threshold. Through pre-configured suppression rules, alarm rules, etc., the data that meets the conditions can be alarmed to timely discover data anomalies so that measures can be taken quickly to solve the problem. Optionally, the alarm can be sent through email, SMS, instant messaging, etc. Optionally, the user's processing action on the alarm and the user's feedback on the alarm can also be collected, such as whether it is a false alarm, whether it is too sensitive, whether the algorithm needs to be switched, whether the user manually selects the algorithm, etc., and the subsequent dynamic threshold algorithm can be adjusted based on the feedback, so as to correct the inappropriate dynamic threshold generation in time.

[0048] As described above, by determining fluctuation information, seasonal change information and holiday index information based on monitoring indicator observation data, and determining fitting degree information and periodicity information under multiple statistical periods based on monitoring indicator observation data and monitoring indicator prediction data, an indicator portrait can be constructed based on fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and a dynamic threshold determination algorithm can be determined based on the indicator portrait, and the target dynamic threshold can be determined based on the dynamic threshold determination algorithm and the monitoring indicator observation data, data monitoring can be performed based on the target dynamic threshold, and an applicable dynamic threshold determination algorithm can be determined based on the changing rules and trends of the monitoring indicator data, so as to more accurately determine the dynamic threshold, effectively reduce the workload of manual operation and maintenance and parameter adjustment of data monitoring, and improve the accuracy and adaptability of data monitoring.

[0049] Based on the above embodiments, Figure 2 A flowchart of another data monitoring method based on a dynamic threshold provided in an embodiment of the present application is given. The data monitoring method based on a dynamic threshold is a specific implementation of the above-mentioned data monitoring method based on a dynamic threshold. Figure 2 , the data monitoring method based on dynamic threshold includes:

[0050] S210: Obtain monitoring indicator observation data and monitoring indicator prediction data, and determine fluctuation information, seasonal change information and holiday index information based on the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods based on the monitoring indicator observation data and monitoring indicator prediction data.

[0051] In a possible embodiment, the data monitoring method based on dynamic threshold provided by the present solution determines fluctuation information according to the monitoring indicator observation data, which may be: determining the standard deviation or coefficient of variation of the monitoring indicator observation data, and determining the standard deviation or coefficient of variation as fluctuation information.

[0052] Exemplarily, the fluctuation information provided by the present solution can be expressed according to the standard deviation (SD) or coefficient of variation (CV) of the monitoring indicator observation data. For example, when the standard deviation of the monitoring indicator observation data is used as the fluctuation information, the standard deviation can be determined by the following formula:

[0053]

[0054] Among them, SD is the standard deviation, x i is the i-th monitoring indicator observation data, n is the data volume of the monitoring indicator observation data, and u is the average value of the monitoring indicator observation data. Among them, a higher standard deviation indicates that the data points are more widely distributed relative to the average value, that is, the fluctuation of the monitoring indicator observation data is larger; a lower standard deviation indicates that the data points are more concentrated near the average value, that is, the fluctuation of the monitoring indicator observation data is smaller.

[0055] In one embodiment, the coefficient of variation can be expressed as the ratio of the standard deviation to the mean value, and the coefficient of variation can be used to measure relative volatility (especially when comparing data sets of different magnitudes or units). When the coefficient of variation of the monitoring indicator observation data is used as the volatility information, the volatility information can be determined by the following formula:

[0056]

[0057] Among them, CV is the coefficient of variation, SD is the standard deviation of the monitoring indicator observation data, and u is the average value of the monitoring indicator observation data. A higher coefficient of variation indicates that the monitoring indicator observation data fluctuates more relative to the average value, and a lower coefficient of variation indicates that the monitoring indicator observation data fluctuates less. The coefficient of variation is dimensionless, so the coefficient of variation can be used to compare the volatility of data sets of different units or magnitudes.

[0058] This solution uses the standard deviation or coefficient of variation of the monitoring indicator observation data as fluctuation information to accurately reflect the overall stability and / or fluctuation of the monitoring indicator observation data, more accurately determine the dynamic threshold determination algorithm, and improve the accuracy and adaptability of data monitoring.

[0059] In one possible embodiment, the dynamic threshold-based data monitoring method provided by the present solution determines seasonal change information based on monitoring indicator observation data, including: determining the periodic average observation data of the monitoring indicator observation data in multiple preset seasonal cycles, and the total average observation data of the monitoring indicator observation data; determining a first ratio of the periodic average observation data and the total average observation data, and determining the first ratio as the seasonal change information of the monitoring indicator in the corresponding preset seasonal cycle.

[0060] In one embodiment, in addition to data fluctuations, the trend of the monitoring indicator is also affected by the season, and the seasonal change of the monitoring indicator can be measured by the seasonal index method. The seasonal index method can be understood as a method of calculating the seasonal change index that describes the change based on the time series characteristics containing seasonal periodic changes. The future status of the prediction target can be predicted based on the regularity of seasonal changes presented by the time series.

[0061] Optionally, the seasonal index can calculate the relative number of seasonal influences in each period within the cycle by the averaging method, and multiple preset seasonal cycles can be preset, for example, a year is divided into multiple preset seasonal cycles (for example, divided into 4 preset seasonal cycles according to the four seasons, or divided into 12 preset seasonal cycles according to the months, and different seasonal granularities can be set by dividing different numbers of preset seasonal cycles), and each preset seasonal cycle corresponds to multiple monitoring indicator observation data. Exemplarily, the periodic average observation data of the monitoring indicator observation data in multiple preset seasonal cycles and the total average observation data of the monitoring indicator observation data are calculated, and the first ratio of the periodic average observation data and the total average observation data is further calculated, and the first ratio can be determined as the seasonal change information of the monitoring indicator in the corresponding preset seasonal cycle.

[0062] Optionally, the period average observation data provided by this scheme can be determined by the following formula:

[0063]

[0064] in, is the average observation data of the kth preset seasonal cycle, m is the total number of preset seasonal cycles, n is the amount of monitoring indicator observation data for each preset seasonal cycle, x ik The observation data of the ith monitoring indicator in the kth preset seasonal cycle.

[0065] Optionally, the total average observation data provided by this scheme can be determined by the following formula:

[0066]

[0067] in, is the total average observation data. Correspondingly, the seasonal variation information provided by this scheme can be determined by the following formula:

[0068]

[0069] Among them, S k is the seasonal variation information of the kth preset seasonal cycle. Among them, the seasonal variation information reflects the relatively stable relationship between the season and the total average value. If the seasonal variation information is greater than 1, it can be determined that the monitoring indicator observation data of the season is often higher than the total average value. If the seasonal variation information is less than 1, it can be determined that the monitoring indicator observation data of the season is often lower than the total average value. If the seasonal variation information of each preset seasonal cycle is approximately 1, it means that the monitoring indicator has no obvious seasonal effect. This scheme uses the first ratio of the periodic average observation data of the monitoring indicator observation data in multiple preset seasonal cycles and the total average observation data of the monitoring indicator observation data as the seasonal variation information of the monitoring indicator in the corresponding preset seasonal cycle, accurately reflects the fluctuation of the monitoring indicator observation data following the season, more accurately determines the dynamic threshold determination algorithm, and improves the accuracy and adaptability of data monitoring.

[0070] In one possible embodiment, the dynamic threshold-based data monitoring method provided by the present scheme determines holiday index information based on monitoring indicator observation data, including: determining the average value of the first monitoring indicator observation data of holidays in a preset period, and the average value of the second monitoring indicator observation data of non-holidays; determining the absolute value of the difference between the average value of the first monitoring indicator observation data and the average value of the second monitoring indicator observation data, determining a second ratio of the absolute value to the average value of the second monitoring indicator observation data, and determining the second ratio as the holiday index information.

[0071] Exemplarily, a first average value of the monitoring indicator observation data during holidays in a preset period is determined, and the first average value is used as the first monitoring indicator observation data average value, and a second average value of the monitoring indicator observation data during non-holidays in the preset period is determined, and the second average value is used as the second monitoring indicator observation data average value.

[0072] In one embodiment, after determining the average value of the first monitoring indicator observation data and the average value of the second monitoring indicator observation data, the absolute value of the first monitoring indicator observation data and the average value of the second monitoring indicator observation data are calculated, and the second ratio of the absolute value and the average value of the second monitoring indicator observation data is calculated. The second ratio can be determined as holiday index information. Optionally, holidays can be determined by an internationally used holiday event table or specified by the user. The preset period can be a combination of one or more of one day, one week, and one month. Optionally, the holiday index information provided by this solution can be determined by the following formula:

[0073] MAPE = abs(y holiday -y origin ) / y origin

[0074] Among them, MAPE is the holiday index information, abs(y holiday -y origin ) is the average value of the first monitoring indicator observation data y holiday and the average value y of the second monitoring indicator observation data origin This solution accurately calculates the holiday index information based on the average value of the first monitoring indicator observation data of holidays in the preset period and the average value of the second monitoring indicator observation data of non-holidays, accurately reflects the fluctuation of the monitoring indicator observation data following the holidays, more accurately determines the dynamic threshold determination algorithm, and improves the accuracy and adaptability of data monitoring.

[0075] In one possible embodiment, after determining the target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data, it can be determined whether the current time is in a preset holiday. If the current time is in a preset holiday, the target dynamic threshold can be corrected based on a preset offset value, and then data monitoring can be performed based on the corrected target dynamic threshold, thereby improving the accuracy and adaptability of data monitoring during holidays.

[0076] Optionally, the target dynamic threshold is corrected based on the preset offset value, which may be by multiplying the first offset value and the target dynamic threshold, adding the product to the second offset value, and using the added result as the corrected target dynamic threshold. For example, the target dynamic threshold may be corrected by the following formula:

[0077] y predict =a*x predict +b

[0078] Among them, y predict is the corrected target dynamic threshold, x predict is the target dynamic threshold before correction, a is the first offset value, and b is the second offset value. The first offset value and the second offset value can be determined based on the following formula:

[0079]

[0080] in, is the pre-collected target dynamic threshold before correction, The i-th corrected target dynamic threshold collected in advance.

[0081] In one possible embodiment, the similarity between the observed value and the predicted value of the monitoring indicator can be measured based on R-squared. R-squared can effectively explain the proportion of data variability in the total variability. Based on this, the data monitoring method based on dynamic threshold provided by the present scheme determines the degree of fit information according to the monitoring indicator observation data and the monitoring indicator prediction data, including: determining the data residual sum of squares and the data total sum of squares according to the monitoring indicator observation data and the monitoring indicator prediction data; determining a second ratio of the data residual sum of squares to the data total sum of squares, and determining the difference between the first preset coefficient and the second ratio as the degree of fit information.

[0082] Exemplarily, the residual sum of squares of the monitoring indicator observation data and the monitoring indicator prediction data, as well as the total sum of squares of the monitoring indicator observation data and the monitoring indicator prediction data are calculated, and a second ratio of the data residual sum of squares to the data total sum of squares is calculated, and the difference between the first preset coefficient and the second ratio can be determined as the degree of fit information. Optionally, the degree of fit information provided by this solution can be determined by the following formula:

[0083]

[0084] Among them, R 2 is the fitting degree information, y i is the observed data of the i-th monitoring indicator, for y i The corresponding monitoring indicator prediction data, is the average value of the monitoring indicator observation data, a is the first preset coefficient, for example, a=1, is the residual sum of squares of the data, This solution determines the data residual sum of squares and the total sum of squares of the data according to the monitoring indicator observation data and the monitoring indicator prediction data to determine the degree of fit information, accurately reflecting the overall degree of fit and similarity of the monitoring indicator prediction data and the monitoring indicator observation data, and can more accurately determine the dynamic threshold determination algorithm, thereby improving the accuracy and adaptability of data monitoring.

[0085] In a possible embodiment, the data monitoring method based on dynamic threshold provided by the present solution determines the periodic information of the monitoring indicator under multiple statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data, including:

[0086] Determine the period residual sum of squares and the period total sum of squares based on the monitoring indicator observation data and monitoring indicator prediction data under the statistical period; determine the third ratio of the period residual sum of squares to the period total sum of squares, and determine the difference between the first preset coefficient and the third ratio as the periodic information under the statistical period.

[0087] Exemplarily, the monitoring indicator observation data and the monitoring indicator prediction data under the preset statistical period are determined, the monitoring indicator observation data and the monitoring indicator prediction data under the statistical period are calculated to determine the period residual sum of squares and the period total sum of squares, the third ratio of the period residual sum of squares to the period total sum of squares is calculated, and the difference between the first preset coefficient and the third ratio is determined as the periodicity information under the statistical period. This solution accurately determines the periodicity information under the statistical period by determining the period residual sum of squares and the period total sum of squares according to the monitoring indicator observation data and the monitoring indicator prediction data under the statistical period, accurately reflects the periodicity of the monitoring indicator prediction data and the monitoring indicator observation data values ​​corresponding to the statistical period, can more accurately determine the dynamic threshold determination algorithm, and improve the accuracy and adaptability of data monitoring.

[0088] In one embodiment, the statistical period provided by the present solution can be configured as one or more, for example, a daily statistical period and a weekly statistical period can be configured, and the daily periodicity information under the daily statistical period and the weekly periodicity information under the weekly statistical period are calculated respectively.

[0089] For example, daily periodicity information can be determined by the following formula:

[0090]

[0091] Among them, DAY_R 2 For daily periodic information, is the i-th monitoring indicator observation data of the day before the current time, for The corresponding monitoring indicator prediction data, is the average value of the monitoring indicator observation data of the day before the current time, a is the first preset coefficient, for example, a=1, is the square of the daily period residual, is the total sum of squares of the daily cycle.

[0092] Optionally, weekly periodicity information may be determined by the following formula:

[0093]

[0094] Among them, WEEK_R 2 For weekly periodic information, is the i-th monitoring indicator observation data of the week before the current time, for The corresponding monitoring indicator prediction data, is the average value of the monitoring indicator observation data of the week before the current time, a is the first preset coefficient, for example, a=1, is the square of the periodic residual, is the total sum of squares of the cycles.

[0095] S220: Normalize the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information to obtain fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information.

[0096] Exemplarily, after determining the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information can be normalized (for example, normalized to the range of 0-1), and the corresponding normalized results can be used as fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information.

[0097] In a possible embodiment, the data monitoring method based on dynamic threshold provided by the present solution performs normalization processing on fluctuation information, seasonal variation information, holiday index information, fitting degree information and periodicity information, which may be:

[0098] S221: When the intermediate indicator information is greater than the second preset coefficient, determine the third preset coefficient as the indicator evaluation information.

[0099] S222: When the intermediate indicator information is less than or equal to the second preset coefficient, determine the difference between the second preset coefficient and the intermediate indicator information, and use the difference as the indicator evaluation information.

[0100] Exemplarily, for each indicator intermediate information, determine whether the indicator intermediate information is greater than a second preset coefficient (for example, 1). When the indicator intermediate information is greater than the second preset coefficient, a third preset coefficient (for example, 0) can be determined as the indicator evaluation information. When the indicator intermediate information is less than or equal to the second preset coefficient, the difference between the second preset coefficient and the indicator intermediate information can be calculated, and the difference can be used as the indicator evaluation information.

[0101] The intermediate indicator information provided by this solution includes fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and the indicator evaluation information includes fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information corresponding to the intermediate indicator information. For example, the indicator evaluation information provided by this solution can be determined by the following formula:

[0102]

[0103] Among them, apdex is the indicator evaluation information, y is the indicator intermediate information, b is the second preset coefficient, and c is the third preset coefficient. Among them, the larger the indicator evaluation information, the higher the satisfaction evaluation of the indicator intermediate information, that is, the higher the fit between the corresponding monitoring indicator and the prediction algorithm. This solution accurately determines the indicator evaluation information by normalizing the comparison result between the indicator intermediate information and the second preset coefficient. The indicator portrait can more accurately reflect the fluctuation trend of each monitoring indicator, accurately determine the appropriate dynamic threshold determination algorithm, and improve the accuracy and adaptability of data monitoring.

[0104] S230: construct an evaluation polygon according to the fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodic evaluation information, and use the evaluation polygon as an indicator portrait.

[0105] Exemplarily, an evaluation polygon is constructed based on the fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodic evaluation information obtained by the above normalization, and the evaluation polygon can be used as an indicator portrait.

[0106] For example, assuming that the periodic evaluation information includes daily periodic evaluation information and weekly periodic evaluation information, the evaluation polygon is an evaluation hexagon, and assuming that the corresponding values ​​of fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information, daily periodic evaluation information and periodic evaluation information are 0.9, 0.7, 0.1, 0.8, 0.5 and 0.5 respectively, then the indicator portrait is as follows: Figure 3 As shown in the provided indicator portrait diagram, the fuller the hexagon of the monitoring indicator, the more suitable the corresponding algorithm is for predicting and alerting the monitoring indicator. This solution accurately constructs the evaluation polygon by normalizing the fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information, and periodic evaluation information, and uses the evaluation polygon as the indicator portrait. According to the indicator portrait, the impact of different evaluation indicators on the fluctuation trend of the monitoring indicator can be accurately determined, and the appropriate dynamic threshold determination algorithm can be accurately determined to improve the accuracy and adaptability of data monitoring.

[0107] S240: Determine a dynamic threshold determination algorithm based on the indicator portrait.

[0108] S250: Determine a target dynamic threshold based on a dynamic threshold determination algorithm and monitoring indicator observation data.

[0109] S260: Perform data monitoring based on the target dynamic threshold.

[0110] As described above, by determining the fluctuation information, seasonal change information and holiday index information according to the monitoring indicator observation data, and determining the fitting degree information and the periodicity information under various statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data, the indicator portrait can be constructed according to the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and the dynamic threshold determination algorithm can be determined according to the indicator portrait, and the target dynamic threshold can be determined based on the dynamic threshold determination algorithm and the monitoring indicator observation data, and data monitoring can be performed based on the target dynamic threshold, and the applicable dynamic threshold determination algorithm can be determined according to the change law and trend of the monitoring indicator data, and the dynamic threshold can be determined more accurately, effectively reducing the manual operation and maintenance and parameter adjustment workload of data monitoring, and improving the accuracy and adaptability of data monitoring. At the same time, the fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information obtained by normalization processing can accurately construct the evaluation polygon, and the evaluation polygon can be used as the indicator portrait, and the influence of different evaluation indicators on the fluctuation trend of the monitoring indicator can be accurately determined according to the indicator portrait, and the appropriate dynamic threshold determination algorithm can be accurately determined, so as to improve the accuracy and adaptability of data monitoring.

[0111] Figure 4 is a structural diagram of a data monitoring device based on a dynamic threshold provided by an embodiment of the present application. Figure 4 The data monitoring device based on dynamic threshold includes a data processing module 41, an algorithm determination module 42, a threshold determination module 43 and a monitoring execution module 44.

[0112] Among them, the data processing module 41 is configured to obtain monitoring indicator observation data and monitoring indicator prediction data, and determine fluctuation information, seasonal change information and holiday index information based on the monitoring indicator observation data, and determine the degree of fit information and periodicity information under multiple statistical periods based on the monitoring indicator observation data and the monitoring indicator prediction data; the algorithm determination module 42 is configured to construct an indicator portrait based on fluctuation information, seasonal change information, holiday index information, degree of fit information and periodicity information, and determine the dynamic threshold determination algorithm based on the indicator portrait; the threshold determination module 43 is configured to determine the target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data; the monitoring execution module 44 is configured to perform data monitoring based on the target dynamic threshold.

[0113] As described above, by determining fluctuation information, seasonal change information and holiday index information based on monitoring indicator observation data, and determining fitting degree information and periodicity information under multiple statistical periods based on monitoring indicator observation data and monitoring indicator prediction data, an indicator portrait can be constructed based on fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and a dynamic threshold determination algorithm can be determined based on the indicator portrait, and the target dynamic threshold can be determined based on the dynamic threshold determination algorithm and the monitoring indicator observation data, data monitoring can be performed based on the target dynamic threshold, and an applicable dynamic threshold determination algorithm can be determined based on the changing rules and trends of the monitoring indicator data, so as to more accurately determine the dynamic threshold, effectively reduce the workload of manual operation and maintenance and parameter adjustment of data monitoring, and improve the accuracy and adaptability of data monitoring.

[0114] In a possible embodiment, the data processing module 41 determines the fluctuation information according to the monitoring indicator observation data, and is configured to: determine the standard deviation or coefficient of variation of the monitoring indicator observation data, and determine the standard deviation or coefficient of variation as the fluctuation information.

[0115] In a possible embodiment, the data processing module 41 determines seasonal variation information according to the monitoring index observation data, and is set as follows:

[0116] Determine the periodic average observation data of the monitoring indicator observation data in a plurality of preset seasonal periods, and the total average observation data of the monitoring indicator observation data;

[0117] A first ratio of the period average observation data to the total average observation data is determined, and the first ratio is determined as seasonal variation information of the monitoring indicator in the corresponding preset seasonal period.

[0118] In a possible embodiment, the data processing module 41 determines holiday index information according to the monitoring index observation data, and is set as:

[0119] Determine an average value of the first monitoring indicator observation data during holidays in a preset period, and an average value of the second monitoring indicator observation data during non-holidays;

[0120] Determine the absolute value of the difference between the average value of the first monitoring indicator observation data and the average value of the second monitoring indicator observation data, determine a second ratio of the absolute value to the average value of the second monitoring indicator observation data, and determine the second ratio as holiday index information.

[0121] In a possible embodiment, the data processing module 41 determines the fitting degree information according to the monitoring indicator observation data and the monitoring indicator prediction data, and is set as:

[0122] Determine the data residual sum of squares and the data total sum of squares based on the monitoring indicator observation data and the monitoring indicator prediction data;

[0123] A second ratio of the data residual sum of squares to the data total sum of squares is determined, and a difference between the first preset coefficient and the second ratio is determined as fitting degree information.

[0124] In a possible embodiment, the data processing module 41 determines the periodic information of the monitoring indicator under various statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data, and is set as follows:

[0125] Determine the period residual sum of squares and the period total sum of squares according to the monitoring indicator observation data and monitoring indicator forecast data in the statistical period;

[0126] A third ratio of the period residual sum of squares to the period total sum of squares is determined, and a difference between the first preset coefficient and the third ratio is determined as periodic information under the statistical period.

[0127] In a possible embodiment, the algorithm determination module 42 constructs an indicator portrait according to the fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and is set as follows:

[0128] Normalizing the fluctuation information, seasonal variation information, holiday index information, fitting degree information and periodicity information to obtain fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information;

[0129] An evaluation polygon is constructed based on fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodic evaluation information, and the evaluation polygon is used as an indicator portrait.

[0130] In a possible embodiment, the algorithm determination module 42 normalizes the fluctuation information, seasonal variation information, holiday index information, fitting degree information, and periodicity information, and is set to:

[0131] When the intermediate information of the indicator is greater than the second preset coefficient, the third preset coefficient is determined as the indicator evaluation information, the intermediate information of the indicator includes fluctuation information, seasonal change information, holiday index information, fitting degree information and periodicity information, and the indicator evaluation information includes fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information and periodicity evaluation information;

[0132] When the intermediate indicator information is less than or equal to the second preset coefficient, the difference between the second preset coefficient and the intermediate indicator information is determined, and the difference is used as the indicator evaluation information.

[0133] It is worth noting that in the above-mentioned embodiment of the data monitoring device based on dynamic thresholds, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the embodiments of the present application.

[0134] The embodiment of the present application also provides a data monitoring device based on a dynamic threshold, and the data monitoring device based on a dynamic threshold can be integrated with the data monitoring apparatus based on a dynamic threshold provided in the embodiment of the present application. Figure 5 is a structural diagram of a data monitoring device based on a dynamic threshold provided by an embodiment of the present application. Figure 5 The data monitoring device based on dynamic thresholds includes: an input device 53, an output device 54, a memory 52, and one or more processors 51; the memory 52 is used to store one or more programs; when one or more programs are executed by one or more processors 51, the one or more processors 51 implement the data monitoring method based on dynamic thresholds as provided in the above embodiments. The data monitoring device, equipment, and computer based on dynamic thresholds provided above can be used to execute the data monitoring method based on dynamic thresholds provided in any of the above embodiments, and have corresponding functions and beneficial effects.

[0135] The embodiments of the present application also provide a non-volatile storage medium storing computer executable instructions, which are used to execute the data monitoring method based on dynamic thresholds as provided in the above embodiments when executed by a computer processor. Of course, the non-volatile storage medium storing computer executable instructions provided in the embodiments of the present application, whose computer executable instructions are not limited to the data monitoring method based on dynamic thresholds provided above, can also execute the related operations in the data monitoring method based on dynamic thresholds provided in any embodiment of the present application. The data monitoring apparatus, device and storage medium based on dynamic thresholds provided in the above embodiments can execute the data monitoring method based on dynamic thresholds provided in any embodiment of the present application. For technical details not described in detail in the above embodiments, please refer to the data monitoring method based on dynamic thresholds provided in any embodiment of the present application.

[0136] On the basis of the above embodiments, the embodiments of the present application also provide a computer program product. The essence of the technical solution of the present application or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer program product is stored in a storage medium, including a number of instructions for enabling a computer device, a mobile terminal or a processor therein to execute all or part of the steps of the dynamic threshold-based data monitoring method provided in each embodiment of the present application.

Claims

1. A data monitoring method based on dynamic threshold, characterized in that: include: Acquire monitoring indicator observation data and monitoring indicator forecast data, and determine fluctuation information, seasonal change information and holiday index information based on the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods based on the monitoring indicator observation data and the monitoring indicator forecast data; Constructing an indicator portrait according to the fluctuation information, the seasonal change information, the holiday index information, the fitting degree information and the periodicity information, and determining a dynamic threshold determination algorithm according to the indicator portrait; Determine a target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data; Data monitoring is performed based on the target dynamic threshold.

2. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: The determining of fluctuation information according to the monitoring indicator observation data includes: Determine the standard deviation or coefficient of variation of the monitoring indicator observation data, and determine the standard deviation or the coefficient of variation as fluctuation information.

3. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: Determining seasonal variation information based on the monitoring indicator observation data includes: Determine the periodic average observation data of the monitoring indicator observation data in a plurality of preset seasonal periods, and the total average observation data of the monitoring indicator observation data; A first ratio of the period average observation data to the total average observation data is determined, and the first ratio is determined as seasonal variation information of the monitoring indicator in the corresponding preset seasonal period.

4. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: Determining holiday index information according to the monitoring indicator observation data includes: Determine an average value of the first monitoring indicator observation data during holidays in a preset period, and an average value of the second monitoring indicator observation data during non-holidays; Determine the absolute value of the difference between the average value of the first monitoring indicator observation data and the average value of the second monitoring indicator observation data, determine a second ratio of the absolute value to the average value of the second monitoring indicator observation data, and determine the second ratio as holiday index information.

5. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: The determining of the degree of fit information according to the monitoring indicator observation data and the monitoring indicator prediction data includes: Determine the data residual sum of squares and the data total sum of squares according to the monitoring indicator observation data and the monitoring indicator prediction data; A second ratio of the data residual sum of squares to the data total sum of squares is determined, and a difference between the first preset coefficient and the second ratio is determined as fitting degree information.

6. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: Determining periodic information of the monitoring indicator under multiple statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data includes: Determine the period residual sum of squares and the period total sum of squares according to the monitoring indicator observation data and the monitoring indicator prediction data in the statistical period; A third ratio of the period residual sum of squares to the period total sum of squares is determined, and a difference between the first preset coefficient and the third ratio is determined as periodicity information under the statistical period.

7. The data monitoring method based on dynamic threshold according to claim 1, characterized in that: The constructing of an indicator portrait according to the fluctuation information, the seasonal change information, the holiday index information, the fitting degree information and the periodicity information includes: Normalizing the fluctuation information, the seasonal variation information, the holiday index information, the fitting degree information, and the periodicity information to obtain fluctuation evaluation information, seasonal evaluation information, holiday evaluation information, fitting degree evaluation information, and periodicity evaluation information; An evaluation polygon is constructed according to the fluctuation evaluation information, the seasonal evaluation information, the holiday evaluation information, the fitting degree evaluation information and the periodic evaluation information, and the evaluation polygon is used as an indicator portrait.

8. The data monitoring method based on dynamic threshold according to claim 7, characterized in that: The normalizing of the fluctuation information, the seasonal variation information, the holiday index information, the fitting degree information and the periodicity information includes: In the case where the intermediate indicator information is greater than the second preset coefficient, the third preset coefficient is determined as the indicator evaluation information, the intermediate indicator information includes the fluctuation information, the seasonal variation information, the holiday index information, the fitting degree information and the periodicity information, and the indicator evaluation information includes the fluctuation evaluation information, the seasonal evaluation information, the holiday evaluation information, the fitting degree evaluation information and the periodicity evaluation information; When the indicator intermediate information is less than or equal to the second preset coefficient, a difference between the second preset coefficient and the indicator intermediate information is determined, and the difference is used as the indicator evaluation information.

9. A data monitoring device based on dynamic threshold, characterized in that: It includes a data processing module, an algorithm determination module, a threshold determination module and a monitoring execution module, wherein: The data processing module is configured to obtain monitoring indicator observation data and monitoring indicator prediction data, and determine fluctuation information, seasonal change information and holiday index information according to the monitoring indicator observation data, and determine fitting degree information and periodicity information under multiple statistical periods according to the monitoring indicator observation data and the monitoring indicator prediction data; The algorithm determination module is configured to construct an indicator portrait according to the fluctuation information, the seasonal change information, the holiday index information, the fitting degree information and the periodicity information, and determine a dynamic threshold determination algorithm according to the indicator portrait; The threshold determination module is configured to determine a target dynamic threshold based on the dynamic threshold determination algorithm and the monitoring indicator observation data; The monitoring execution module is configured to perform data monitoring based on the target dynamic threshold.

10. A data monitoring device based on dynamic threshold, characterized in that: include: memory and one or more processors; The memory is used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the data monitoring method based on dynamic thresholds as described in any one of claims 1 to 8.

11. A non-volatile storage medium storing computer executable instructions, characterized in that: When the computer executable instructions are executed by a computer processor, they are used to perform the data monitoring method based on dynamic thresholds as described in any one of claims 1 to 8.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the data monitoring method based on dynamic threshold value described in any one of claims 1 to 8 is implemented.

Images