Real-time data fluctuation detection method, device, equipment, medium and program product

Abstract

The application relates to the technical field of big data, and provides a real-time data fluctuation detection method, device, equipment, medium and program product. The real-time data fluctuation detection method comprises the following steps: determining historical data; performing data aggregation in a first time granularity and a second time granularity based on the historical data, to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and smaller than the second time granularity; fitting a local dynamic baseline corresponding to the time granularity of real-time data according to the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data is smaller than the first time granularity; and performing fluctuation detection on the real-time data based on the local dynamic baseline. The application is used for effectively and automatically capturing abnormal fluctuation of data at a high frequency, and false alarms caused by random disturbance are avoided as much as possible.

Description

Technical Field

[0001] This invention relates to the field of big data technology, and in particular to a method, apparatus, equipment, medium, and program product for real-time data fluctuation detection. Background Technology

[0002] To quickly gain insights into market conditions and seize market opportunities, there is an urgent need for accurate and rapid perception of massive amounts of data related to complex business operations such as marketing and business processing. Simultaneously, it is desirable to detect data fluctuations immediately.

[0003] In real-time data, the time granularity of data statistics has a significant impact on the characteristics presented in real time. Generally, a smaller time granularity results in more frequent changes in real-time data, with greater random disturbances and instability. However, sensitive data fluctuations contain more subtle fluctuation information, enabling timely detection of problems. A larger time granularity can better eliminate random disturbances, thus making the overall real-time data more stable; however, the trend changes are slower, resulting in a relatively delayed reflection of specific data conditions.

[0004] Therefore, how to effectively and automatically capture abnormal fluctuations in data at a high frequency, and minimize false alarms that may be caused by random disturbances, has become an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a real-time data fluctuation detection method, apparatus, device, medium, and program product, which can effectively and automatically capture abnormal data fluctuations at a high frequency, and minimize false alarms that may be caused by random disturbances.

[0006] In a first aspect, embodiments of this application provide a real-time data fluctuation detection method, including:

[0007] Determine historical data;

[0008] Based on the historical data, data aggregation is performed at a first time granularity and a second time granularity to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity.

[0009] Based on the first time granularity sequence and the second time granularity sequence, a local dynamic baseline corresponding to the time granularity of the real-time data is fitted and generated; the time granularity of the real-time data is smaller than the first time granularity.

[0010] Fluctuation detection is performed on the real-time data based on the local dynamic baseline.

[0011] In one embodiment, the step of fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence includes:

[0012] A first dynamic baseline is generated by fitting the first time-granularity sequence.

[0013] A second dynamic baseline is generated by fitting the second time-granularity sequence;

[0014] Based on the first dynamic baseline and the second dynamic baseline, a local dynamic baseline corresponding to the time granularity of the real-time data is fitted and generated.

[0015] In one embodiment, the step of fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first dynamic baseline and the second dynamic baseline includes:

[0016] The data at each corresponding time point on the first dynamic baseline and the second dynamic baseline are averaged to generate a local dynamic baseline with a time granularity corresponding to the real-time data.

[0017] In one embodiment, the step of generating a first dynamic baseline by fitting the first time-granularity sequence includes:

[0018] The first time-granularity sequence is fitted using the Holt-Winter method to generate the first dynamic baseline;

[0019] The step of generating a second dynamic baseline by fitting the second time-granularity sequence includes:

[0020] The second time-granularity sequence is fitted using the Holt-Winter method to generate the second dynamic baseline.

[0021] In one embodiment, the step of aggregating data at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence includes:

[0022] The historical data is averaged based on the first time granularity to obtain the first time granularity sequence.

[0023] The historical data is averaged based on the second time granularity to obtain the second time granularity sequence.

[0024] In one embodiment, the fluctuation detection of the real-time data based on the local dynamic baseline includes:

[0025] When the number of real-time data exceeding the local dynamic baseline within a preset time range reaches a set threshold, it is determined that the real-time data has fluctuated.

[0026] Secondly, embodiments of this application provide a real-time data fluctuation detection device, comprising:

[0027] The data determination module is used to determine historical data;

[0028] The time series acquisition module is used to perform data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity.

[0029] A dynamic baseline generation module is used to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data is smaller than the first time granularity.

[0030] The data fluctuation detection module is used to detect fluctuations in the real-time data based on the local dynamic baseline.

[0031] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the real-time data fluctuation detection method as described above.

[0032] Fourthly, embodiments of this application also provide a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the real-time data fluctuation detection method as described above.

[0033] Fifthly, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the real-time data fluctuation detection method as described above.

[0034] The real-time data fluctuation detection method, apparatus, device, medium, and program product provided by this invention aggregates historical data at a first time granularity and a second time granularity to obtain a first time granularity sequence and a second time granularity sequence. Based on the first and second time granularity sequences, a local dynamic baseline corresponding to the time granularity of the real-time data is fitted and generated. Fluctuation detection of the real-time data is then performed based on this local dynamic baseline. Since the first time granularity is larger than the time granularity of the historical data but smaller than the second time granularity, the first time granularity sequence contains more nuanced data fluctuation information; the second time granularity sequence effectively eliminates random disturbances. Therefore, by using the local dynamic baseline obtained from the first and second time granularity sequences to perform fluctuation detection on the real-time data, abnormal data fluctuations can be effectively and automatically captured at a high frequency, while minimizing false alarms that may be caused by random disturbances. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0036] Figure 1 This is one of the flowcharts of the real-time data fluctuation detection method provided by the present invention;

[0037] Figure 2 This is the second flowchart of the real-time data fluctuation detection method provided by the present invention;

[0038] Figure 3 This represents the time-series data trend of a certain KPI data based on a frequency of one minute in this embodiment of the invention;

[0039] Figure 4 This diagram illustrates the structure of the early warning baseline prediction results of the present invention for sequences aggregated at 30 minutes and sequences aggregated at 15 minutes, respectively.

[0040] Figure 5 This indicates the data fluctuation detection results of the new sequence obtained by aggregating 30-minute frequencies in this embodiment of the invention;

[0041] Figure 6 This indicates the data fluctuation detection results of the new sequence obtained by aggregating 15-minute frequencies in this embodiment of the invention;

[0042] Figure 7 This indicates the data fluctuation detection results based on local dynamic baselines in this embodiment of the invention.

[0043] Figure 8 This is a schematic diagram of the structure of the real-time data fluctuation detection device provided by the present invention;

[0044] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0046] Current technologies face several challenges:

[0047] 1. Data fluctuation detection at a smaller time granularity has strong random perturbations, which weakens the regularity of data changes and makes it more difficult to fit accurately.

[0048] 2. Detecting data fluctuations at a smaller time granularity means that high-frequency monitoring and high-frequency prediction of a large number of indicators are required, which requires a lot of computing power.

[0049] 3. Detecting data fluctuations at a smaller time granularity results in more severe periodic drift characteristics and even loses the basic real-time characteristics.

[0050] In real-time data, the time granularity of data statistics has a significant impact on the characteristics presented in real time. Generally, a smaller time granularity means that real-time data changes more frequently, exhibiting greater random disturbances and instability; however, these sensitive data fluctuations contain more nuanced information, allowing for timely detection of problems. A larger time granularity can better eliminate random disturbances, resulting in greater overall stability in real-time data; however, the trend changes more slowly, leading to a relative lag in reflecting specific data conditions.

[0051] In view of this, embodiments of this application propose a real-time data fluctuation detection method, apparatus, device, medium, and program product, which aims to achieve effective and automatic capture of abnormal data fluctuations at a high frequency, and to minimize false alarms that may be caused by random disturbances.

[0052] The core idea of ​​this invention is to map historical data with a smaller time granularity to a larger time granularity for dynamic baseline prediction, and then perform high-frequency data fluctuation detection on the real data at the smaller time granularity. After aggregating historical data at different time granularities, different new sequences are obtained; then, dynamic baseline prediction is performed on these new sequences to obtain a dynamic baseline within a certain future time range; finally, these dynamic baselines are mapped onto the time axis of real-time data with a smaller time granularity, resulting in a local dynamic baseline that measures the real-time data. Subsequently, real-time monitoring and anomaly fluctuation judgment are performed on the real-time data based on the local dynamic baseline, achieving real-time monitoring of high-frequency data.

[0053] The following is combined Figures 1-2 This invention describes the real-time data fluctuation detection.

[0054] Please refer to Figure 1 Real-time data fluctuation detection includes:

[0055] Step 100: Determine historical data;

[0056] Historical data is acquired through electronic devices. This historical data can be network performance KPI (Key Performance Indicator) test data with a smaller time granularity (such as seconds or minutes). Specifically, KPI test data represents key performance parameters. These may include: call success rate (for voice services), activation success rate (for packet-switched services), network latency, PS (Packet Switch) throughput, PS ping packet latency, ping packet success rate, etc.

[0057] For example, historical data could include network latency data per minute for an hour, from 10 a.m. to 11 a.m.

[0058] Step 200: Based on the historical data, perform data aggregation at a first time granularity and a second time granularity respectively to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity.

[0059] The electronic device performs data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence.

[0060] Directly detecting fluctuations in small-granularity historical data can introduce significant random disturbances. However, since small-granularity historical data contains rich information about data fluctuations, as much detail of these fluctuations as possible should be preserved. Therefore, data aggregation at a first time granularity is performed on the historical data to obtain a first time granularity sequence. The first time granularity is larger than the time granularity of the historical data.

[0061] For example, historical data has a time granularity of 1 minute. The time granularity of the first time granularity sequence can be 10 minutes, 15 minutes, or 20 minutes.

[0062] Furthermore, high-frequency real-time data is more sensitive to external interference. By mapping high-frequency real-time data to low-frequency real-time data, random disturbances are weakened, resulting in a sequence with better real-time characteristics. Data aggregation at a second time granularity is performed based on the historical data to obtain a second time-granularity sequence. The second time granularity is larger than the first time granularity.

[0063] For example, the time granularity of the first time granularity sequence can be 10 minutes, 15 minutes, or 20 minutes; the time granularity of the second time granularity sequence can be 25 minutes, 30 minutes, or 35 minutes, etc.

[0064] Specifically, in this embodiment of the application, step 200, which involves performing data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence, includes:

[0065] Step 210: Perform mean processing on the historical data based on the first time granularity to obtain the first time granularity sequence;

[0066] The electronic device performs averaging on the historical data based on the first time granularity to obtain the first time granularity sequence. The electronic device then converts high-frequency historical data into a lower-frequency first time granularity sequence based on the first time granularity.

[0067] For example, electronic devices can convert historical data at a frequency of one minute into an average value over a 15-minute period as the value of a new sequence. Specifically, the average of the sums of KPI test data per minute within a 1-15 minute period can be used as the data value of the first time-granularity sequence per minute within a 1-15 minute period.

[0068] Step 220: Perform mean processing on the historical data based on the second time granularity to obtain the second time granularity sequence.

[0069] The electronic device performs averaging on the historical data based on the second time granularity to obtain a second time granularity sequence. The electronic device then converts high-frequency historical data into a lower-frequency second time granularity sequence based on the second time granularity.

[0070] For example, electronic devices can convert historical data at a frequency of one minute into an average value over a 30-minute period as the value of a new sequence. Specifically, the average of the sums of KPI test data per minute within a 1-30 minute period can be used as the data value of a second time-granularity sequence per minute within a 1-30 minute period.

[0071] Step 300: Based on the first time granularity sequence and the second time granularity sequence, fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data; the time granularity of the real-time data is smaller than the first time granularity.

[0072] The electronic device uses the first and second time granularity sequences to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data. The real-time data can also be the aforementioned key performance parameters, such as call success rate (for voice services), activation success rate (for packet-switched services), network latency, PS (Packet Switch) throughput, PS ping packet latency, ping packet success rate, etc.

[0073] It should be noted that the time granularity of the real-time data is smaller than that of the first time granularity. In this embodiment, the time granularity of the real-time data is the same as that of the historical data.

[0074] Because the first time granularity of the first time granularity sequence is larger than the time granularity of the historical data and smaller than the second time granularity of the second time granularity sequence, the first time granularity sequence retains richer and more detailed data fluctuation information. The second time granularity sequence, on the other hand, removes random disturbances caused by external factors from the high-frequency real-time data. Therefore, by fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first and second time granularity sequences, rapid data fluctuation detection of high-frequency real-time KPIs can be achieved, while reducing random disturbances in high-frequency real-time data and ensuring the capture of local data fluctuations within the long-term trend of the sequence.

[0075] Specifically, in the embodiments of this application, please refer to Figure 2 Step 300, the step of fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence, includes:

[0076] Step 310: Generate a first dynamic baseline by fitting the first time granularity sequence.

[0077] The electronic device generates a first dynamic baseline by fitting the first time granularity sequence.

[0078] Specifically, step 310, generating a first dynamic baseline based on the first time-granularity sequence, includes:

[0079] Step 311: Fit the first time-granularity sequence using the Holt-Winter method to generate the first dynamic baseline. The obtained first dynamic baseline can be used for fluctuation detection of low-frequency real-time data.

[0080] Step 320: Generate a second dynamic baseline by fitting the second time granularity sequence;

[0081] The electronic device generates a second dynamic baseline by fitting the second time-granularity sequence.

[0082] Specifically, step 320, generating a second dynamic baseline based on the second time-granularity sequence, includes:

[0083] Step 321: Fit the second time-granularity sequence using the Holt-Went method to generate the second dynamic baseline. The obtained second dynamic baseline can be used for fluctuation detection in low-frequency real-time data.

[0084] By mapping high-frequency historical data detection onto low-frequency data for learning, and then mapping the prediction results back onto high-frequency real-time data for comparison, the model can be fitted to data sequences with better characteristics. This reduces the optimization time consumed by the model during high-frequency data fluctuation detection, thereby improving the model's detection efficiency. Furthermore, baseline fitting at a lower-frequency time granularity allows for better automatic acquisition of data fluctuation detection results, serving as auxiliary labels for data fluctuation anomalies in different time periods, and providing label information for subsequent algorithm learning during high-frequency real-time data fluctuation detection.

[0085] Step 330: Based on the first dynamic baseline and the second dynamic baseline, fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data.

[0086] The electronic device uses the first dynamic baseline and the second dynamic baseline to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data. By combining the first dynamic baseline and the second dynamic baseline to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data, it is possible to achieve rapid data fluctuation detection of high-frequency real-time KPI data, reduce random disturbances in high-frequency real-time data, and ensure the capture of local data fluctuations under the long-term trend of the sequence.

[0087] Specifically, step 330 involves fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first dynamic baseline and the second dynamic baseline, including:

[0088] Step 331: Average the data at each corresponding time point on the first dynamic baseline and the second dynamic baseline to generate a local dynamic baseline corresponding to the time granularity of the real-time data.

[0089] The electronic device performs averaging on the data at each corresponding time point on the first dynamic baseline and the second dynamic baseline to generate a local dynamic baseline with a time granularity corresponding to the real-time data.

[0090] Specifically, the prediction results on the first dynamic baseline and the second dynamic baseline can be filled into their corresponding time windows, and then the average value at each time point on the first dynamic baseline and the second dynamic baseline can be calculated as the local dynamic baseline, that is, the high-frequency dynamic baseline corresponding to the time granularity of the real-time data.

[0091] Step 400: Detect fluctuations in the real-time data based on the local dynamic baseline.

[0092] The electronic device performs fluctuation detection on the real-time data based on the local dynamic baseline.

[0093] It should be noted that the dynamic baseline value at each moment of high-frequency time series obtained from historical data represents a local average level and cannot be simply used as an absolute standard for judging whether there are abnormal fluctuations in high-frequency real-time data. Therefore, it is necessary to determine whether abnormal fluctuations exist within a smaller local range.

[0094] Specifically, step 400, the fluctuation detection of the real-time data based on the local dynamic baseline, includes:

[0095] Step 410: When the number of real-time data exceeding the local dynamic baseline within a preset time range reaches a set threshold, it is determined that the real-time data is fluctuating.

[0096] Specifically, in this embodiment of the invention, the actual value within a preset time range exceeds a set proportion of the local dynamic baseline using sliding statistics. When the value exceeds a given threshold, the range is considered to be abnormally fluctuating; otherwise, it is considered normal.

[0097] For example, fluctuation is defined as the proportion of true values ​​exceeding the local dynamic baseline by more than 30% within a 10-minute timeframe. That is, if 3 out of 10 data points within a 10-minute timeframe have true values ​​exceeding the local dynamic baseline, then data fluctuation is considered to have occurred within that 10-minute timeframe.

[0098] By aggregating historical data at first and second time granularities, a first time granularity sequence and a second time granularity sequence are obtained. Based on these sequences, a local dynamic baseline corresponding to the time granularity of the real-time data is generated. Fluctuation detection of the real-time data is then performed based on this local dynamic baseline. Since the first time granularity is larger than the time granularity of the historical data but smaller than the second time granularity, the first time granularity sequence contains more nuanced data fluctuation information; the second time granularity sequence effectively eliminates random disturbances. Therefore, using the local dynamic baseline obtained from the first and second time granularity sequences to perform fluctuation detection on the real-time data can effectively and automatically capture abnormal data fluctuations at a high frequency, while minimizing false alarms that may be caused by random disturbances.

[0099] The following is an illustration using a typical example.

[0100] Electronic devices obtain time-series data of a certain business KPI at a frequency of one minute, such as... Figure 3 As shown. From Figure 3 As can be seen, the KPI data is relatively low most of the time, but during operation, there are sudden and significant fluctuations of varying magnitudes. Assuming a traditional fixed baseline approach, if the threshold is set to 100, a large number of false alarms will be generated; if it is set to greater than 200, while generating false alarms, most of the abnormal data fluctuations will not be detected.

[0101] In this embodiment, Figure 3 The time-series data was mapped to 15-minute and 30-minute frequencies and aggregated to obtain two new sequences. Then, early warning baseline predictions were performed on these two new sequences, such as... Figure 4 As shown, predictions for a certain step length are obtained from two new sequences aggregated at different time frequencies. The prediction results are assigned to the minute granularity within the corresponding time range, and the warning baseline value corresponding to each minute is calculated to obtain the local dynamic baseline. The detailed calculation process is shown in Table 1.

[0102] Table 1: Schematic diagram of generating local dynamic baselines

[0103]

[0104] In this embodiment of online data fluctuation detection, fluctuation is defined as the proportion of the true value exceeding the baseline by more than 30% within a 10-minute timeframe; other situations are considered normal. The final detection result output is as follows: Figures 5-7 .

[0105] from Figures 5-7As can be seen, based on the local dynamic baseline of this embodiment, compared with the data fluctuation detection results obtained by 30-minute frequency aggregation and 15-minute frequency aggregation, this embodiment can effectively and automatically capture abnormal data fluctuations at a higher frequency, and can minimize false alarms that may be caused by random disturbances.

[0106] Therefore, the real-time data fluctuation detection method of this application embodiment has the following effects:

[0107] 1. Solve the problem of rapid data fluctuation detection for high-frequency real-time KPI data, improving the algorithm's detection performance. Achieve high-performance unsupervised data fluctuation identification without requiring extensive data labeling work.

[0108] 2. It reduces high-frequency real-time random disturbances, ensures the capture of local data fluctuations under the long-term trend of the sequence, and has strong applicability and promotion in most application scenarios.

[0109] This invention learns sequence characteristics by converting historical data to a larger time granularity, and then maps it to a smaller time granularity of real-time data for data fluctuation identification. This significantly reduces computational requirements and is suitable for monitoring scenarios involving large-scale data fluctuations.

[0110] The real-time data fluctuation detection device provided by the present invention is described below. The real-time data fluctuation detection device described below can be referred to in correspondence with the real-time data fluctuation detection method described above.

[0111] Please refer to Figure 8 This application provides a real-time data fluctuation detection device, comprising:

[0112] Data determination module 201 is used to determine historical data;

[0113] The time series acquisition module 202 is used to perform data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity.

[0114] The dynamic baseline generation module 203 is used to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data is smaller than the first time granularity.

[0115] The data fluctuation detection module 204 is used to perform fluctuation detection on the real-time data based on the local dynamic baseline.

[0116] The real-time data fluctuation detection device in this embodiment aggregates historical data at a first time granularity and a second time granularity to obtain a first time granularity sequence and a second time granularity sequence. Based on the first and second time granularity sequences, a local dynamic baseline corresponding to the time granularity of the real-time data is generated. Fluctuation detection of the real-time data is then performed based on this local dynamic baseline. Since the first time granularity is larger than the time granularity of the historical data but smaller than the second time granularity, the first time granularity sequence contains more nuanced data fluctuation information; the second time granularity sequence effectively eliminates random disturbances. Therefore, by using the local dynamic baseline obtained from the first and second time granularity sequences to detect fluctuations in the real-time data, abnormal data fluctuations can be captured effectively and automatically at a high frequency, while minimizing false alarms that may be caused by random disturbances.

[0117] Based on the above embodiments, as an optional embodiment, the dynamic baseline generation module specifically includes:

[0118] The first dynamic baseline generation module is used to generate a first dynamic baseline by fitting the first time granularity sequence.

[0119] The second dynamic baseline generation module is used to fit and generate a second dynamic baseline based on the second time granularity sequence.

[0120] The local dynamic baseline generation module is used to fit and generate a local dynamic baseline corresponding to the time granularity of the real-time data based on the first dynamic baseline and the second dynamic baseline.

[0121] Based on the above embodiments, as an optional embodiment, the local dynamic baseline is specifically used for:

[0122] The data at each corresponding time point on the first dynamic baseline and the second dynamic baseline are averaged to generate a local dynamic baseline with a time granularity corresponding to the real-time data.

[0123] Based on the above embodiments, as an optional embodiment, the first dynamic baseline generation module is specifically used for:

[0124] The first time-granularity sequence is fitted using the Holt-Winter method to generate the first dynamic baseline;

[0125] The second dynamic baseline generation module is specifically used for:

[0126] The second time-granularity sequence is fitted using the Holt-Winter method to generate the second dynamic baseline.

[0127] Based on the above embodiments, as an optional embodiment, the time series acquisition module includes:

[0128] The first time-granularity sequence acquisition module is used to perform mean processing on the historical data based on the first time granularity to obtain the first time-granularity sequence;

[0129] The second time granularity sequence acquisition module is used to perform mean processing on the historical data based on the second time granularity to obtain the second time granularity sequence.

[0130] Based on the above embodiments, as an optional embodiment, the data fluctuation detection module is specifically used for:

[0131] When the number of real-time data exceeding the local dynamic baseline within a preset time range reaches a set threshold, it is determined that the real-time data has fluctuated.

[0132] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 9 As shown, the electronic device may include a processor 910, a communications interface 920, a memory 930, and a communication bus 940, wherein the processor 910, the communications interface 920, and the memory 930 communicate with each other via the communication bus 940. The processor 910 can call logical instructions in the memory 930 to execute a real-time data fluctuation detection method. This method includes: determining historical data; performing data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity; fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data is less than the first time granularity; and performing fluctuation detection on the real-time data based on the local dynamic baseline.

[0133] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0134] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the real-time data fluctuation detection method provided by the above methods. The method includes: determining historical data; performing data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity; fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data according to the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data is less than the first time granularity; and performing fluctuation detection on the real-time data based on the local dynamic baseline.

[0135] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a real-time data fluctuation detection method provided by the methods described above. This method includes: determining historical data; performing data aggregation at a first time granularity and a second time granularity based on the historical data to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity being larger than the time granularity of the historical data and smaller than the second time granularity; fitting and generating a local dynamic baseline corresponding to the time granularity of the real-time data based on the first time granularity sequence and the second time granularity sequence; the time granularity of the real-time data being smaller than the first time granularity; and performing fluctuation detection on the real-time data based on the local dynamic baseline.

[0136] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0137] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time data fluctuation detection, the method comprising: The method comprises the following steps: determining historical data; performing data aggregation of the historical data based on a first time granularity and a second time granularity respectively to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity; generating a local dynamic baseline corresponding to the time granularity of real-time data according to the first time granularity sequence and the second time granularity sequence, comprising: generating a first dynamic baseline according to the first time granularity sequence; generating a second dynamic baseline according to the second time granularity sequence; generating a local dynamic baseline corresponding to the time granularity of real-time data according to the first dynamic baseline and the second dynamic baseline; the time granularity of the real-time data is the same as the time granularity of the historical data; performing fluctuation detection on the real-time data based on the local dynamic baseline.

2. The real-time data fluctuation detection method of claim 1, wherein, The method of generating a local dynamic baseline corresponding to the time granularity of real-time data according to the first dynamic baseline and the second dynamic baseline comprises: performing mean value processing on the data of each corresponding time point on the first dynamic baseline and the second dynamic baseline to generate a local dynamic baseline corresponding to the time granularity of real-time data.

3. The real-time data fluctuation detection method of claim 1, wherein, The method of generating a first dynamic baseline according to the first time granularity sequence comprises: performing fitting on the first time granularity sequence based on the Holt-Winters method to generate the first dynamic baseline; The method of generating a second dynamic baseline according to the second time granularity sequence comprises: performing fitting on the second time granularity sequence based on the Holt-Winters method to generate the second dynamic baseline.

4. The real-time data fluctuation detection method of claim 1, wherein, The method of performing data aggregation of the historical data based on a first time granularity and a second time granularity respectively to obtain a first time granularity sequence and a second time granularity sequence comprises: performing mean value processing on the historical data based on the first time granularity to obtain the first time granularity sequence; performing mean value processing on the historical data based on the second time granularity to obtain the second time granularity sequence.

5. The real-time data fluctuation detection method according to claim 1, wherein The method of performing fluctuation detection on the real-time data based on the local dynamic baseline comprises: when the number of real-time data exceeding the local dynamic baseline in a preset time range reaches a set threshold, it is determined that the real-time data fluctuates.

6. A real-time data fluctuation detection apparatus characterized by comprising: The method comprises the following steps: a data determination module for determining historical data; a time sequence acquisition module for performing data aggregation of the historical data based on a first time granularity and a second time granularity respectively to obtain a first time granularity sequence and a second time granularity sequence; the first time granularity is greater than the time granularity of the historical data and less than the second time granularity; The dynamic baseline generation module is configured to generate a local dynamic baseline corresponding to the time granularity of real-time data according to the first time granularity sequence and the second time granularity sequence, including: generating a first dynamic baseline according to the first time granularity sequence; generating a second dynamic baseline according to the second time granularity sequence; and generating a local dynamic baseline corresponding to the time granularity of real-time data according to the first dynamic baseline and the second dynamic baseline; the time granularity of the real-time data is the same as the time granularity of the historical data; The data fluctuation detection module is configured to detect fluctuations in the real-time data based on the local dynamic baseline.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the real-time data fluctuation detection method according to any one of claims 1 to 5 when executing the program.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program implements the real-time data fluctuation detection method according to any one of claims 1 to 5 when executed by the processor.

9. A computer program product comprising a computer program, characterized in that, The computer program implements the real-time data fluctuation detection method according to any one of claims 1 to 5 when executed by the processor.

Images