Method, apparatus and electronic device for determining abnormal timing segment

Abstract

This application discloses a method, apparatus, and electronic device for determining abnormal time series segments, belonging to the field of data processing technology. The method includes: acquiring a monitoring time series under a monitoring environment, the monitoring time series including multiple time-continuous monitoring data; performing data mutation detection on the monitoring time series to determine at least one mutation data in the monitoring time series; based on the at least one mutation data, dividing the monitoring time series into multiple first time series segments; inputting each of the first time series segments into a preset neural network prediction model to determine abnormal time series segments among the multiple first time series segments, the abnormal time series segment representing a time series segment including abnormal data.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and specifically relates to a method, apparatus and electronic device for determining abnormal time sequence segments. Background Technology

[0002] With the rapid development of data processing technology, it has been applied to various industries. People can acquire monitoring data from their respective industries, analyze the data, and obtain analytical results, which can then provide guidance for actual production and daily life. For example, in the field of geological disaster prevention and control, provinces and cities can set up early warning monitoring points. By using sensors at these points to acquire local geological monitoring data and analyzing it, early warnings of geological disasters can be achieved.

[0003] In the existing monitoring data collection process, abnormal data inevitably appears. Furthermore, in the analysis of monitoring data, abnormal data can affect the accuracy of the results. For example, abnormal data can trigger numerous alarms, thus requiring the identification of abnormal data from massive amounts of monitoring data. However, currently, the identification of abnormal data in monitoring data is usually done manually, resulting in low efficiency. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, and electronic device for determining abnormal time segments, which can solve the problem of low efficiency in identifying abnormal data.

[0005] In a first aspect, embodiments of this application provide a method for determining abnormal timing segments, including:

[0006] Acquire a monitoring time series under the monitoring environment, wherein the monitoring time series includes multiple monitoring data that are consecutive in time;

[0007] Data mutation detection is performed on the monitoring time series to identify at least one mutation data in the monitoring time series;

[0008] Based on the at least one mutation data, the monitoring time series is divided into multiple first time series segments;

[0009] Each of the first time-series segments is input into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments. The abnormal time-series segments represent time-series segments that include abnormal data.

[0010] In this embodiment, by performing data mutation detection on the monitoring time series acquired under the monitoring environment, at least one mutation data point in the monitoring time series is identified. Then, based on the at least one mutation data point, the monitoring time series is divided into multiple first time series segments. Finally, each of the multiple first time series segments is input into a preset neural network prediction model to identify abnormal time series segments among the multiple first time series segments. In this way, abnormal time series segments can be quickly identified from a monitoring time series including multiple monitoring data points. This allows for the identification of abnormal data in the monitoring time series by analyzing only abnormal time series segments, thereby reducing the complexity of identifying abnormal data from monitoring data and improving the efficiency of abnormal data identification.

[0011] In some implementations, the step of inputting each of the first time-series segments into a preset neural network prediction model to determine the anomalous time-series segments among the plurality of first time-series segments includes:

[0012] Each of the first time-series segments is concatenated with a preset time-series segment in a preset time-series sample to obtain a concatenated time-series segment associated with the first time-series segment. The preset time-series sample further includes a time-series segment identifier, which is used to indicate whether the preset time-series segment is an abnormal time-series segment.

[0013] The spliced ​​time series segments associated with each of the first time series segments are input into a preset neural network prediction model, and abnormal time series segments among the plurality of first time series segments are determined based on the time series segment identifier.

[0014] In this embodiment, by splicing each first time segment with a preset time segment in a preset time sample, and inputting the spliced ​​time segment into a preset neural network prediction model, and determining whether the first time segment is an abnormal time segment based on the time segment identifier in the preset time sample, the accuracy of identification can be improved.

[0015] In some embodiments, the step of concatenating each of the first time-series segments with a preset time-series segment in a preset time-series sample to obtain a concatenated time-series segment associated with the first time-series segment includes:

[0016] The target time sequence segment is concatenated with a preset time sequence segment from a plurality of preset time sequence samples to obtain a plurality of first concatenated time sequence segments associated with the target time sequence segment, wherein the target time sequence segment is any one of the plurality of first time sequence segments;

[0017] The step of inputting the spliced ​​time series segments associated with each of the first time series segments into a preset neural network prediction model, and determining the abnormal time series segments among the plurality of first time series segments based on the determination of the time series segment identifier, includes:

[0018] The plurality of first spliced ​​time series segments are respectively input into a preset neural network prediction model to obtain a plurality of first prediction results for the plurality of first spliced ​​time series segments. Each first prediction result is used to indicate whether the first spliced ​​time series segment corresponding to it is an abnormal time series segment.

[0019] The plurality of first prediction results are compared with at least one second prediction result to obtain a comparison result. The at least one second prediction result is obtained by inputting at least one second spliced ​​time sequence segment into the neural network prediction model. The at least one second spliced ​​time sequence segment is obtained by splicing two preset time sequence segments from the plurality of preset time sequence samples.

[0020] If the comparison result indicates that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, then the target time series segment is determined to be an abnormal time series segment, so as to obtain the abnormal time series segment among the multiple first time series segments.

[0021] In this embodiment, by splicing the first time sequence segment with preset time sequence segments in multiple preset time sequence samples respectively, multiple first spliced ​​time sequence segments are obtained. Then, based on the first prediction results of the multiple first spliced ​​time sequence segments and the second prediction results of the second spliced ​​time sequence segments obtained by splicing the preset time sequence segments in multiple preset time sequence samples in pairs, it is determined whether the first time sequence segment is an abnormal time sequence segment, thereby further improving the recognition accuracy.

[0022] In some implementations, the step of inputting each of the first time-series segments into a preset neural network prediction model to determine the anomalous time-series segments among the plurality of first time-series segments includes:

[0023] Each of the first time segments is divided into multiple time sub-segments;

[0024] Multiple time-series sub-segments of each of the first time-series segments are respectively input into the neural network prediction model to obtain multiple prediction sub-results of the multiple time-series sub-segments. Each prediction sub-result is used to indicate whether its corresponding time-series sub-segment is an abnormal time-series segment.

[0025] Based on multiple prediction sub-results of the multiple time series sub-segments, it is determined whether the first time series segment corresponding to the multiple time series sub-segments is an abnormal time series segment, so as to obtain the abnormal time series segment among the multiple first time series segments.

[0026] In this embodiment, by dividing the first time series segment into multiple time series sub-segments and inputting each time series sub-segment into the neural network prediction model, multiple prediction sub-results of multiple time series sub-segments are obtained. Finally, based on the multiple prediction sub-results of multiple time series sub-segments, it is determined whether the first time series segment is an abnormal time series segment, thereby reducing the computational complexity in the prediction process and further improving the efficiency of abnormal data identification.

[0027] In some embodiments, the step of performing data mutation detection on the monitoring time series to determine at least one mutation data in the monitoring time series includes:

[0028] Based on a preset data length, the monitoring time series is divided into multiple second time series segments, and the data length of each second time series segment is less than or equal to the preset data length.

[0029] Data mutation detection is performed on each of the second time series segments to identify the mutation data in each of the second time series segments, thereby obtaining at least one mutation data in the monitored time series.

[0030] In this embodiment, by dividing the monitoring time series into multiple second time series segments and performing data mutation detection on each of the multiple second time series segments to determine the mutation data in each second time series segment, the mutation data in the monitoring time series can be obtained. This reduces the complexity of the time series being detected in data mutation detection, improves the efficiency of data mutation detection, and further improves the efficiency of abnormal data identification.

[0031] In some embodiments, before performing data mutation detection on the monitoring time series to determine at least one mutation data in the monitoring time series, the method further includes:

[0032] Obtain the variance of the monitoring time series;

[0033] The step of detecting data mutations in the monitored time series to determine at least one mutation data point in the monitored time series includes:

[0034] If the variance of the monitoring time series is greater than or equal to a preset threshold, data mutation detection is performed on the monitoring time series to identify at least one mutation data in the monitoring time series.

[0035] In this embodiment, before performing data mutation detection on the monitoring time series, the variance of the monitoring time series is also obtained. Data mutation detection is only performed on the monitoring time series if the variance is greater than or equal to a preset threshold. Since there will be mutated data when the variance of the monitoring time series is large, data mutation detection can be avoided on monitoring time series without mutated data, thereby reducing the computational overhead of electronic devices.

[0036] In some implementations, adjacent first timing segments among the plurality of first timing segments have overlapping timing windows.

[0037] In this embodiment, by dividing the monitoring time series into multiple first time series segments with overlapping time series windows based on at least one mutation data, the missing monitoring data of the monitoring time series in the data mutation detection can be avoided, thereby further improving the accuracy of abnormal data identification.

[0038] In some embodiments, before inputting each of the first time-series segments into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments, the method further includes:

[0039] Among the plurality of first time series segments, at least one time series segment to be predicted is selected, and the change magnitude of each time series segment to be predicted is greater than or equal to a preset magnitude.

[0040] The step of inputting each of the first time-series segments into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments includes:

[0041] Each of the time series segments to be predicted is input into a preset neural network prediction model to determine the abnormal time series segments in the at least one time series segment to be predicted.

[0042] In this embodiment, by selecting at least one time series segment to be predicted from multiple first time series segments, and inputting each segment of the at least one time series segment to be predicted into the neural network prediction model, the computational consumption of electronic devices can be reduced while ensuring accuracy, thereby further improving the efficiency of abnormal data identification.

[0043] In some implementations, selecting at least one time series segment to be predicted from the plurality of first time series segments includes:

[0044] The start and end data in each of the first time segments are compared to obtain the comparison results;

[0045] If the comparison result indicates that the change between the starting data and the ending data is greater than or equal to a preset change, the first time segment corresponding to the comparison result is determined as the time segment to be predicted.

[0046] In this embodiment, by determining the change in the starting and ending data in the first time series segment as the change magnitude, the method of filtering the time series segment to be predicted becomes more flexible.

[0047] In some implementations, the monitoring time series is a time series within a geological disaster monitoring environment.

[0048] In this embodiment, the efficiency of identifying abnormal data in time series within the geological disaster monitoring environment can be improved.

[0049] Secondly, embodiments of this application provide an apparatus for determining abnormal timing segments, comprising:

[0050] The time series acquisition module is used to acquire the monitoring time series under the monitoring environment, wherein the monitoring time series includes multiple monitoring data that are continuous in time;

[0051] A mutation data determination module is used to perform data mutation detection on the monitoring time series and determine at least one mutation data in the monitoring time series.

[0052] The first time series segment division module is used to divide the monitoring time series into multiple first time series segments based on the at least one mutation data.

[0053] An abnormal time series segment determination module is used to input each of the first time series segments into a preset neural network prediction model to determine the abnormal time series segments among the plurality of first time series segments, wherein the abnormal time series segment represents a time series segment that includes abnormal data.

[0054] In this embodiment, by performing data mutation detection on the monitoring time series acquired under the monitoring environment, at least one mutation data point in the monitoring time series is identified. Then, based on the at least one mutation data point, the monitoring time series is divided into multiple first time series segments. Finally, each of the multiple first time series segments is input into a preset neural network prediction model to identify abnormal time series segments among the multiple first time series segments. In this way, abnormal time series segments can be quickly identified from a monitoring time series including multiple monitoring data points. This allows for the identification of abnormal data in the monitoring time series by analyzing only abnormal time series segments, thereby reducing the complexity of identifying abnormal data from monitoring data and improving the efficiency of abnormal data identification.

[0055] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0056] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0057] Fifthly, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating an embodiment of the method for determining abnormal timing segments provided in this application;

[0059] Figure 2 This is a schematic diagram of the Bayesian theoretical framework in an embodiment of the method for determining abnormal time segments provided in this application;

[0060] Figure 3 This is a schematic diagram of timing segment division in an embodiment of the method for determining abnormal timing segments provided in this application;

[0061] Figure 4 This is a schematic diagram of an embodiment of the device for determining abnormal timing segments provided in this application;

[0062] Figure 5 This is a schematic diagram of the structure of an embodiment of the electronic device with abnormal timing segments provided in this application. Detailed Implementation

[0063] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0064] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0065] The method, apparatus, and electronic device for determining abnormal timing segments provided in this application will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios.

[0066] Please see Figure 1This is a flowchart illustrating a method for determining abnormal timing segments provided in an embodiment of this application. This method is applied to electronic devices, which may include servers, etc. Figure 1 As shown, the method for determining abnormal time series segments includes the following steps:

[0067] Step 101: Obtain the monitoring time series under the monitoring environment. The monitoring time series includes multiple monitoring data that are continuous in time.

[0068] Step 102: Perform data mutation detection on the monitoring time series to identify at least one mutation data in the monitoring time series;

[0069] Step 103: Based on at least one mutation data point, divide the monitoring time series into multiple first time series segments;

[0070] Step 104: Input each first time series segment into the preset neural network prediction model to identify abnormal time series segments among the multiple first time series segments. Abnormal time series segments refer to time series segments that include abnormal data.

[0071] In this embodiment, by performing data mutation detection on the monitoring time series acquired under the monitoring environment, at least one mutation data point in the monitoring time series is identified. Then, based on the at least one mutation data point, the monitoring time series is divided into multiple first time series segments. Finally, each of the multiple first time series segments is input into a preset neural network prediction model to identify abnormal time series segments among the multiple first time series segments. In this way, abnormal time series segments can be quickly identified from a monitoring time series including multiple monitoring data points. This allows for the identification of abnormal data in the monitoring time series by analyzing only abnormal time series segments, thereby reducing the complexity of identifying abnormal data from monitoring data and improving the efficiency of abnormal data identification.

[0072] In step 101 above, the electronic device can acquire the monitoring time series under the monitoring environment.

[0073] The aforementioned monitoring time series is monitoring data collected over a period of time by a monitoring device (such as at least one of sensors and cameras) set up in the monitoring environment. In other words, the monitoring time series includes multiple monitoring data that are consecutive in time.

[0074] The aforementioned monitoring time series includes multiple monitoring data points, which can include several, tens, hundreds, thousands, or even tens of thousands of data points, and there is no limitation here.

[0075] The time window length of the above monitoring time series can be set according to actual needs. For example, the time window length of the monitoring time series can be set to hours, days, weeks, months, or years, etc.

[0076] The aforementioned monitoring environment can be any environment where data monitoring and analysis are required. For example, the monitoring environment can be the atmospheric environment, and the aforementioned monitoring data can be at least one of atmospheric data such as temperature, humidity, and ultraviolet radiation intensity, so that weather can be predicted through the detection and analysis of atmospheric data.

[0077] In step 102 above, after acquiring the monitoring time series under the detection environment, the electronic device can perform data mutation detection on the monitoring time series to determine at least one mutation data in the monitoring time series.

[0078] The aforementioned data mutation detection of the monitored time series can be achieved through pre-configured data mutation detection rules or models in electronic devices. Specifically, it can be achieved through at least one mutation detection method pre-configured in the electronic device, such as Bayesian theory, Mann-Kendall mutation detection, sliding T-test, and Pettitt test.

[0079] For example, such as Figure 2 As shown, in Bayesian theory for detecting mutations in time series data, the posterior distribution is proportional to the product of the likelihood (i.e., the detection and monitoring time series) and the prior distribution. It is hoped that after continuously observing new samples, the prior and posterior distributions can maintain formal uniformity (for example, this can be achieved through conjugate priors, which is a property unique to exponential family functions, such as the prior of the mean of a Gaussian distribution being another Gaussian distribution). This allows for dynamic parameter updates, i.e., real-time online learning, thereby determining mutation data in the monitoring time series.

[0080] In step 103 above, after determining at least one mutation data in the monitoring time series, the electronic device can divide the monitoring time series into multiple first time series segments based on the at least one mutation data.

[0081] The above method of dividing the monitoring time series into multiple first time series segments based on at least one mutation data can be achieved by determining the first data, at least one mutation data, and the last data in the monitoring time series, and taking the time series between two adjacent data in the first data, at least one mutation data, and the last data as the first time series segment.

[0082] It should be noted that the multiple first time series segments after the above division may not contain the aforementioned mutation data; or, at least one of the starting and ending data of each first time series segment may be mutation data, thereby ensuring data integrity.

[0083] For example, suppose the above monitoring time series is [A1, A2, ..., A... i A i+1, ..., A j A j+1 , ..., A k A k+1 , ..., A n ], and A i A j and A k Given this monitoring time series, the electronic device can divide the monitoring time series into 4 time segments (i.e., the first time segment), namely [A1, A2, ..., A...]. i ], [A i+1 , ..., A j ], [A j+1 , ..., A k ], and [A k+1 , ..., A n ].

[0084] In step 104 above, after the monitoring time series is divided into multiple first time series segments, the electronic device can input each of the multiple first time series segments into a preset neural network prediction model, and predict whether each first time series segment is an abnormal time series segment through the neural network prediction model, so as to determine the abnormal time series segments among the multiple first time series segments.

[0085] The aforementioned neural network prediction model can be pre-configured in an electronic device, and it can be trained using a training dataset to predict whether the input time segment is an anomalous time segment.

[0086] The training dataset mentioned above can be a dataset that includes a large number of training samples, and each training sample can include a historical time series segment and a time series segment identifier for that historical time series segment. The time series segment identifier is used to indicate whether the corresponding historical time series segment is an abnormal time series segment, and the label can be manually assigned.

[0087] The neural network prediction model obtained by training the training dataset can be achieved by inputting training samples into the initial neural network model, iterating the initial neural network model continuously (i.e. adjusting the network parameters of the initial neural network model) based on the output results of the initial neural network model and the time segment identifiers of the training samples, until the accuracy of the output results of the initial neural network model reaches the preset accuracy, and then stopping the iteration of the initial neural network model to obtain the aforementioned neural network prediction model.

[0088] The first time series segment identified as an anomalous time series segment by the aforementioned neural network prediction model indicates that the first time series segment contains anomalous data, and anomalous data can be further filtered out from each first time series segment identified as an anomalous time series segment by human or electronic means.

[0089] In this embodiment of the application, the above-mentioned inputting each first time segment into a preset neural network prediction model to determine whether each first time segment is an abnormal time segment can be done by directly inputting each first time segment into the neural network prediction model without any processing.

[0090] In some implementations, the above-mentioned inputting each first time segment into a preset neural network prediction model to determine abnormal time segments among multiple first time segments includes:

[0091] Each first time sequence segment is concatenated with a preset time sequence segment in a preset time sequence sample to obtain a concatenated time sequence segment associated with the first time sequence segment. The preset time sequence sample also includes a time sequence segment identifier, which is used to indicate whether the preset time sequence segment is an abnormal time sequence segment.

[0092] The spliced ​​time segments associated with each first time segment are input into a preset neural network prediction model, and abnormal time segments in multiple first time segments are identified based on the time segment identifier.

[0093] In this embodiment, by splicing each first time segment with a preset time segment in a preset time sample, and inputting the spliced ​​time segment into a preset neural network prediction model, and determining whether the first time segment is an abnormal time segment based on the time segment identifier in the preset time sample, the accuracy of identification can be improved.

[0094] The aforementioned preset time series sample may include the aforementioned preset time series segment and a time series segment identifier for the preset time series segment. The time series segment identifier is used to indicate whether the preset time series segment is an abnormal time series segment. The preset time series segment may include a time series segment comprising multiple historical monitoring data.

[0095] It should be noted that the predicted time series sample can also be a positive sample. For example, the time series segment is identified as 1, which means that the time series segment is not an abnormal time series segment.

[0096] The above-mentioned splicing of the first time sequence segment and the preset time sequence segment can be done by connecting the first time sequence segment and the preset time sequence segment end to end to form a time sequence segment.

[0097] For example, suppose the above-mentioned preset timing segment is [B1, B2, ..., B m ], and the time segment identifier of the preset time segment is 1, when the above monitoring time series is divided into [A1, A2, ..., A i ], [A i+1 , ..., A j ], [A j+1 , ..., Ak ], and [A k+1 , ..., A n In the case of a total of 4 time segments, [B1, B2, ..., B] can be respectively... m [A1, A2, ..., A] are concatenated with these four time series segments to obtain four concatenated segments [A1, A2, ..., A]. i B1, B2, ..., B m ], [A i+1 , ..., A j B1, B2, ..., B m ], [A j+1 , ..., A k B1, B2, ..., B m ], and [A k+1 , ..., A n B1, B2, ..., B m ].

[0098] The above-mentioned input of the spliced ​​time segment associated with the first time segment into the preset neural network prediction model, and prediction of whether the first time segment is an abnormal segment based on the time segment identifier, can be achieved by comparing the output of the neural network prediction model with the time segment identifier to obtain a comparison result of whether the output result is consistent with the time segment identifier, and finally determining whether the first time segment is an abnormal segment based on the comparison result.

[0099] For example, when the preset time series sample is a positive sample, i.e., the time series segment is identified as 1, if the spliced ​​segments [A1, A2, ..., A...] are... i B1, B2, ..., B m If the input is given to the above neural network prediction model and the output is 1, then the time segment [A1, A2, ..., A] is determined. i [A1, A2, ..., A] is not an abnormal timing segment; if the output result is 0, then the timing segment [A1, A2, ..., A] is determined. i [This is an abnormal timing segment.]

[0100] The above-mentioned splicing of the first timing segment with a preset timing segment in a preset timing sample can be done by splicing only the first timing segment with a preset timing segment.

[0101] In some implementations, each first time sequence segment is concatenated with a preset time sequence segment in a preset time sequence sample to obtain a concatenated time sequence segment associated with the first time sequence segment, including:

[0102] The target time sequence segment is concatenated with a preset time sequence segment from multiple preset time sequence samples to obtain multiple first concatenated time sequence segments associated with the target time sequence segment. The target time sequence segment is any one of the multiple first time sequence segments.

[0103] The spliced ​​time series segments associated with each first time series segment are input into a preset neural network prediction model, and based on the determined time series segment identifiers, abnormal time series segments among multiple first time series segments are identified, which may include:

[0104] Multiple first spliced ​​time segments are input into a preset neural network prediction model to obtain multiple first prediction results for multiple first spliced ​​time segments. Each first prediction result is used to indicate whether the first spliced ​​time segment corresponding to it is an abnormal time segment.

[0105] Multiple first prediction results are compared with at least one second prediction result to obtain a comparison result. The at least one second prediction result is obtained by inputting at least one second spliced ​​time sequence segment into the neural network prediction model. The at least one second spliced ​​time sequence segment is obtained by splicing two preset time sequence segments from multiple preset time sequence samples.

[0106] If the comparison results indicate that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, each target time series segment is identified as an abnormal time series segment, so as to obtain the abnormal time series segments among the multiple first time series segments.

[0107] In this embodiment, by splicing the first time sequence segment with preset time sequence segments in multiple preset time sequence samples respectively, multiple first spliced ​​time sequence segments are obtained. Then, based on the first prediction results of the multiple first spliced ​​time sequence segments and the second prediction results of the second spliced ​​time sequence segments obtained by splicing the preset time sequence segments in multiple preset time sequence samples in pairs, it is determined whether the first time sequence segment is an abnormal time sequence segment, thereby further improving the recognition accuracy.

[0108] The differences between the aforementioned multiple first prediction results and at least one second prediction result can be obtained through a pre-defined method for determining significant differences. For example, a T-test can be used to determine whether there are significant differences between the means μ1 and μ2 of the multiple first prediction results and at least one second prediction result. This can be achieved by setting the null hypothesis H0: μ1 = μ2, the alternative hypothesis H1: μ1 ≠ μ2, and the significance level α = 0.01, thus obtaining the p-value of the difference between the two.

[0109] If the comparison results indicate that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, the electronic device can determine that the target timing segment is an abnormal timing segment; otherwise, the target timing segment is not an abnormal timing segment.

[0110] For example, suppose an electronic device has multiple preset positive samples, each positive sample including a preset timing segment and a timing segment identifier, and the target timing segment can be any one of the four timing segments mentioned above, then:

[0111] Multiple positive samples can be concatenated in pairs and input into the neural network prediction model mentioned above to obtain a set of output results X (i.e., at least one second prediction result);

[0112] These multiple positive samples can be concatenated with the target time segment and input into the above neural network prediction model to obtain a set of output results Y (i.e. multiple first prediction results);

[0113] Then, the T-test can be used to determine the p-value of the difference between set X and set Y. If the p-value is greater than the preset difference threshold α, the test sample is an abnormal time segment; otherwise, the target time segment is not an abnormal time segment.

[0114] In this embodiment of the application, the above-mentioned input of the first time segment into the preset neural network prediction model can also be that the first time segment is input into the neural network prediction model as a whole.

[0115] In some implementations, each first time-series segment is input into a preset neural network prediction model to identify anomalous time-series segments among the multiple first time-series segments, including:

[0116] Each first time segment is divided into multiple time sub-segments;

[0117] Multiple time-series sub-segments of each first time-series segment are input into the neural network prediction model to obtain multiple prediction sub-results for multiple time-series sub-segments. Each prediction sub-result is used to indicate whether its corresponding time-series sub-segment is an abnormal time-series segment.

[0118] Based on multiple prediction sub-results of multiple time series sub-segments, it is determined whether the first time series segment corresponding to the multiple time series sub-segments is an abnormal time series segment, so as to obtain the abnormal time series segments among the multiple first time series segments.

[0119] In this embodiment, by dividing the first time series segment into multiple time series sub-segments and inputting each time series sub-segment into the neural network prediction model, multiple prediction sub-results of multiple time series sub-segments are obtained. Finally, based on the multiple prediction sub-results of multiple time series sub-segments, it is determined whether the first time series segment is an abnormal time series segment, thereby reducing the computational complexity in the prediction process and further improving the efficiency of abnormal data identification.

[0120] The above-mentioned determination of whether a first time series segment corresponding to multiple time series segments is an abnormal time series segment based on multiple prediction sub-results of multiple time series segments can be made in the following ways: if at least one prediction sub-result among the multiple prediction sub-results indicates that its corresponding time series segment is an abnormal time series segment, then the first time series segment is determined to be an abnormal time series segment; or, it can be made by determining the proportion of prediction sub-results that indicate that the corresponding time series segment is an abnormal time series segment among the multiple prediction sub-results, and if the proportion exceeds a preset proportion, then the first time series segment is determined to be an abnormal time series segment.

[0121] The above-mentioned division of each first time segment into multiple time sub-segments can be achieved by the electronic device having a preset length threshold for the time segment, and the length of each time sub-segment obtained by the electronic device being less than or equal to the length threshold. Alternatively, the first time segment can be divided according to a preset data quantity.

[0122] For example, if the first time segment contains 350 monitoring data points and the preset data quantity is 100, then the electronic device can divide the monitoring data points 1 to 100 in the first time segment into time segment 1, the monitoring data points 101 to 200 in the first time segment into time segment 2, the monitoring data points 201 to 300 in the first time segment into time segment 3, and the monitoring data points 301 to 350 in the first time segment into time segment 4, and so on.

[0123] The above-mentioned data mutation detection of the monitoring time series, to identify at least one mutation data in the monitoring time series, can be performed by treating the monitoring time series as a whole for data mutation detection.

[0124] In some implementations, the above-described data mutation detection of the monitoring time series to identify at least one mutation data in the monitoring time series includes:

[0125] Based on the preset data length, the monitoring time series is divided into multiple second time series segments, and the data length of each second time series segment is less than or equal to the preset data length;

[0126] Data mutation detection is performed on each second time segment to identify the mutation data in each second time segment, thereby obtaining at least one mutation data in the monitored time series.

[0127] In this embodiment, by dividing the monitoring time series into multiple second time series segments and performing data mutation detection on each of the multiple second time series segments to determine the mutation data in each second time series segment, the mutation data in the monitoring time series can be obtained. This reduces the complexity of the time series being detected in data mutation detection, improves the efficiency of data mutation detection, and further improves the efficiency of abnormal data identification.

[0128] The preset data length mentioned above can be the length of the time window or the data length, i.e., the number of data points; there is no limitation here.

[0129] For example, if the aforementioned monitoring time series includes 10,000 monitoring data points, the preset data length can be set to 2,000 data points. Then, the electronic device can divide the 10,000 monitoring data points into five second time-series segments, each containing 2,000 consecutive monitoring data points. In this way, the electronic device can perform data mutation detection on each of the five second time-series segments to obtain the mutation data in each segment.

[0130] In some implementations, before performing data mutation detection on the monitoring time series to identify at least one mutation in the monitoring time series, the method further includes:

[0131] Obtain the variance of the monitoring time series;

[0132] Data mutation detection is performed on the monitored time series to identify at least one mutation in the monitored time series, including:

[0133] If the variance of the monitored time series is greater than or equal to a preset threshold, data mutation detection is performed on the monitored time series to identify at least one mutation data in the monitored time series.

[0134] In this embodiment, before performing data mutation detection on the monitoring time series, the variance of the monitoring time series is also obtained. Data mutation detection is only performed on the monitoring time series if the variance is greater than or equal to a preset threshold. Since there will be mutated data when the variance of the monitoring time series is large, data mutation detection can be avoided on monitoring time series without mutated data, thereby reducing the computational overhead of electronic devices.

[0135] Of course, if the variance of the above-mentioned monitoring time series is less than the preset threshold, the electronic device can directly mark the monitoring time series as a non-abnormal time series segment.

[0136] The multiple first time series segments obtained by dividing at least one mutation data mentioned above may be two adjacent first time series segments without data overlap.

[0137] In some implementations, adjacent timing first time segments in a plurality of first time segments have overlapping timing windows.

[0138] In this embodiment, by dividing the monitoring time series into multiple first time series segments with overlapping time series windows based on at least one mutation data, the missing monitoring data of the monitoring time series in the data mutation detection can be avoided, thereby further improving the accuracy of abnormal data identification.

[0139] For example, in such Figure 3 In the monitoring time series shown, assuming that the 1000th and 2000th monitoring data are abrupt changes, the electronic device can divide the monitoring time series into first time segments including 0-1000, 500-1000, and 1000-2000, etc.

[0140] The above-mentioned inputting each first time-series segment into a preset neural network prediction model to determine the abnormal time-series segments among multiple first time-series segments can be achieved by inputting all segments from multiple first time-series segments into the neural network prediction model.

[0141] In some implementations, before inputting each first time segment into a preset neural network prediction model to determine the abnormal time segments among the multiple first time segments, the method further includes:

[0142] Among multiple first time series segments, at least one time series segment to be predicted is selected, and the change amplitude of each time series segment to be predicted is greater than or equal to a preset amplitude.

[0143] Each first time series segment is input into a preset neural network prediction model to identify anomalous time series segments among the multiple first time series segments, including:

[0144] Each time series segment to be predicted is input into a preset neural network prediction model to identify at least one abnormal time series segment in the time series segment to be predicted.

[0145] In this embodiment, by selecting at least one time series segment to be predicted from multiple first time series segments, and inputting each segment of the at least one time series segment to be predicted into the neural network prediction model, the computational consumption of electronic devices can be reduced while ensuring accuracy, thereby further improving the efficiency of abnormal data identification.

[0146] The above-mentioned selection of at least one time series segment to be predicted from multiple first time series segments can be achieved by obtaining the absolute difference between the largest and smallest data in each first time series segment and using the absolute difference as the change magnitude of the first time series segment.

[0147] In some implementations, at least one time series segment to be predicted is selected from a plurality of first time series segments, including:

[0148] The start and end data in each first time segment are compared to obtain the comparison results;

[0149] If the comparison result indicates that the change between the starting data and the ending data is greater than or equal to a preset change, the first time segment corresponding to the comparison result is determined as the time segment to be predicted.

[0150] In this embodiment, by determining the change in the starting and ending data in the first time series segment as the change magnitude, the method of filtering the time series segment to be predicted becomes more flexible.

[0151] In some implementations, the monitoring time series is a time series within the geological disaster monitoring environment.

[0152] In this embodiment, the efficiency of identifying abnormal data in time series within the geological disaster monitoring environment can be improved.

[0153] It should be noted that the execution entity of the method for determining abnormal timing segments provided in this application embodiment can be an abnormal timing segment determination device, or a control module in the abnormal timing segment determination device for executing the method for determining abnormal timing segments. This application embodiment uses the execution of the abnormal timing segment determination method by an abnormal timing segment determination device as an example to illustrate the abnormal timing segment determination device provided in this application embodiment.

[0154] Please see Figure 4 This is a schematic diagram of the structure of the device for determining abnormal timing segments provided in an embodiment of this application. Figure 4 As shown, the device 400 for determining the abnormal timing segment includes:

[0155] The time series acquisition module 401 is used to acquire the monitoring time series under the monitoring environment. The monitoring time series includes multiple monitoring data that are continuous in time.

[0156] The mutation data determination module 402 is used to perform data mutation detection on the monitoring time series and determine at least one mutation data in the monitoring time series.

[0157] The first time series segment division module 403 is used to divide the monitoring time series into multiple first time series segments based on at least one mutation data.

[0158] The abnormal time series segment determination module 404 is used to input each first time series segment into a preset neural network prediction model to determine the abnormal time series segments among the multiple first time series segments. The abnormal time series segment represents a time series segment that includes abnormal data.

[0159] In some implementations, the abnormal timing segment determination module 404 includes:

[0160] The segment splicing unit is used to splice each first time sequence segment with a preset time sequence segment in a preset time sequence sample to obtain a spliced ​​time sequence segment associated with the first time sequence segment. The preset time sequence sample also includes a time sequence segment identifier, which is used to indicate whether the preset time sequence segment is an abnormal time sequence segment.

[0161] The first segment determination unit is used to input the spliced ​​time segments associated with each first time segment into a preset neural network prediction model, and to determine the abnormal time segments among multiple first time segments based on the time segment identifier.

[0162] In some implementations, the fragment splicing unit is specifically used for:

[0163] The target time sequence segment is concatenated with a preset time sequence segment from multiple preset time sequence samples to obtain multiple first concatenated time sequence segments associated with the target time sequence segment. The target time sequence segment is any one of the multiple first time sequence segments.

[0164] The first segment determination unit may include:

[0165] The first prediction result determination subunit is used to input multiple first spliced ​​time segments into a preset neural network prediction model to obtain multiple first prediction results for multiple first spliced ​​time segments. Each first prediction result is used to indicate whether the first spliced ​​time segment corresponding to it is an abnormal time segment.

[0166] The comparison result determination subunit is used to compare multiple first prediction results with at least one second prediction result to obtain a comparison result. The at least one second prediction result is obtained by inputting at least one second spliced ​​time sequence segment into the neural network prediction model. The at least one second spliced ​​time sequence segment is obtained by splicing two preset time sequence segments from multiple preset time sequence samples.

[0167] An abnormal time series segment determination subunit is used to determine a target time series segment as an abnormal time series segment when the comparison result indicates that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, so as to obtain the abnormal time series segment among multiple first time series segments.

[0168] In some implementations, the abnormal timing segment determination module 404 includes:

[0169] A time-series sub-segment partitioning unit is used to divide each first time-series segment into multiple time-series sub-segments;

[0170] The prediction sub-result determination unit is used to input multiple time-series sub-segments of each first time-series segment into the neural network prediction model to obtain multiple prediction sub-results of multiple time-series sub-segments. Each prediction sub-result is used to indicate whether its corresponding time-series sub-segment is an abnormal time-series segment.

[0171] The second segment determination unit is used to determine whether the first time segment corresponding to the multiple time segments is an abnormal time segment based on the multiple prediction sub-results of multiple time segments, so as to obtain the abnormal time segment among the multiple first time segments.

[0172] In some implementations, the mutation data determination module 402 includes:

[0173] The second time series segment division unit is used to divide the monitoring time series into multiple second time series segments based on a preset data length, wherein the data length of each second time series segment is less than or equal to the preset data length.

[0174] The mutation data determination unit is used to perform data mutation detection on each second time segment, determine the mutation data in each second time segment, and obtain at least one mutation data in the monitored time series.

[0175] In some implementations, it also includes:

[0176] The variance acquisition module is used to acquire the variance of the monitored time series.

[0177] The mutation data determination module is specifically used for:

[0178] If the variance of the monitored time series is greater than or equal to a preset threshold, data mutation detection is performed on the monitored time series to identify at least one mutation data in the monitored time series.

[0179] In some implementations, adjacent timing first time segments in a plurality of first time segments have overlapping timing windows.

[0180] In some implementations, it also includes:

[0181] The time series segment filtering module is used to filter out at least one time series segment to be predicted from multiple first time series segments, wherein the change amplitude of each time series segment to be predicted is greater than or equal to a preset amplitude.

[0182] The abnormal timing segment determination module is specifically used for:

[0183] Each time series segment to be predicted is input into a preset neural network prediction model to identify at least one abnormal time series segment in the time series segment to be predicted.

[0184] In some implementations, the time-series segment filtering module includes:

[0185] The comparison result determination unit is used to compare the start data and the end data in each first time segment to obtain the comparison result;

[0186] The time series segment determination unit is used to determine the first time series segment corresponding to the comparison result as the time series segment to be predicted when the change between the starting data and the ending data indicated by the comparison result is greater than or equal to a preset change amount.

[0187] The device for determining abnormal timing segments provided in this application embodiment can achieve... Figure 1 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0188] Please see Figure 5 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application.

[0189] The electronic device may include a processor 501 and a memory 502 storing computer program instructions.

[0190] Specifically, the processor 501 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0191] Memory 502 may include mass storage for data or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In some embodiments, memory 502 may include removable or non-removable (or fixed) media, or memory 502 may be a non-volatile solid-state memory. In some embodiments, memory 502 may be internal or external to a battery device.

[0192] In some implementations, memory 502 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0193] Memory 502 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method for determining abnormal timing segments according to this application.

[0194] The processor 501 reads and executes computer program instructions stored in the memory 502 to achieve... Figure 1 The method in the illustrated embodiment achieves... Figure 1 The technical effects achieved by executing the methods / steps shown in the examples are not elaborated here for the sake of brevity.

[0195] In some embodiments, the electronic device may further include a communication interface 503 and a bus 504. For example, Figure 5 As shown, the processor 501, memory 502, and communication interface 503 are connected through bus 504 and complete communication with each other.

[0196] The communication interface 503 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0197] Bus 504 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 504 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0198] The electronic device can execute the method for determining abnormal timing segments in the embodiments of this application, thereby achieving the combination Figure 1 and Figure 4 The method and apparatus for determining abnormal time segments are described.

[0199] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described method embodiment for determining abnormal timing segments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0200] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0201] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described method embodiment for determining abnormal timing segments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0202] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0203] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0204] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0205] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for determining abnormal time segments, characterized in that, include: Acquire a monitoring time series under the monitoring environment, the monitoring time series including multiple monitoring data that are continuous in time; wherein, the monitoring data is at least one of temperature, humidity and ultraviolet intensity; Data mutation detection is performed on the monitoring time series to identify at least one mutation data in the monitoring time series; Based on the at least one mutation data, the monitoring time series is divided into multiple first time series segments; Each of the first time series segments is input into a preset neural network prediction model to determine the abnormal time series segments among the plurality of first time series segments. The abnormal time series segments represent time series segments that include abnormal data. The step of inputting each of the first time-series segments into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments includes: The target time sequence segment is concatenated with a preset time sequence segment from a plurality of preset time sequence samples to obtain a plurality of first concatenated time sequence segments associated with the target time sequence segment. The target time sequence segment is any one of the plurality of first time sequence segments. The preset time sequence sample further includes a time sequence segment identifier, which is used to indicate whether the preset time sequence segment is an abnormal time sequence segment. The plurality of first spliced ​​time series segments are respectively input into a preset neural network prediction model to obtain a plurality of first prediction results for the plurality of first spliced ​​time series segments. Each first prediction result is used to indicate whether the first spliced ​​time series segment corresponding to it is an abnormal time series segment. The plurality of first prediction results are compared with at least one second prediction result to obtain a comparison result. The at least one second prediction result is obtained by inputting at least one second spliced ​​time sequence segment into the neural network prediction model. The at least one second spliced ​​time sequence segment is obtained by splicing two preset time sequence segments from the plurality of preset time sequence samples. If the comparison result indicates that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, the target time series segment is determined to be an abnormal time series segment, so as to obtain the abnormal time series segment among the multiple first time series segments.

2. The method according to claim 1, characterized in that, The step of inputting each of the first time-series segments into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments includes: Each of the first time segments is divided into multiple time sub-segments; Multiple time-series sub-segments of each of the first time-series segments are respectively input into the neural network prediction model to obtain multiple prediction sub-results of the multiple time-series sub-segments. Each prediction sub-result is used to indicate whether its corresponding time-series sub-segment is an abnormal time-series segment. Based on multiple prediction sub-results of the multiple time series sub-segments, it is determined whether the first time series segment corresponding to the multiple time series sub-segments is an abnormal time series segment, so as to obtain the abnormal time series segment among the multiple first time series segments.

3. The method according to claim 1, characterized in that, The step of detecting data mutations in the monitored time series to determine at least one mutation data point in the monitored time series includes: Based on a preset data length, the monitoring time series is divided into multiple second time series segments, and the data length of each second time series segment is less than or equal to the preset data length. Data mutation detection is performed on each of the second time series segments to identify the mutation data in each of the second time series segments, thereby obtaining at least one mutation data in the monitored time series.

4. The method according to claim 1, characterized in that, Before performing data mutation detection on the monitored time series and determining at least one mutation data in the monitored time series, the method further includes: Obtain the variance of the monitoring time series; The step of detecting data mutations in the monitored time series to determine at least one mutation data point in the monitored time series includes: If the variance of the monitoring time series is greater than or equal to a preset threshold, data mutation detection is performed on the monitoring time series to identify at least one mutation data in the monitoring time series.

5. The method according to claim 1, characterized in that, The adjacent first timing segments in the plurality of first timing segments have overlapping timing windows.

6. The method according to claim 1, characterized in that, Before inputting each of the first time series segments into a preset neural network prediction model to determine the abnormal time series segments among the plurality of first time series segments, the method further includes: Among the plurality of first time series segments, at least one time series segment to be predicted is selected, and the change magnitude of each time series segment to be predicted is greater than or equal to a preset magnitude. The step of inputting each of the first time-series segments into a preset neural network prediction model to determine the abnormal time-series segments among the plurality of first time-series segments includes: Each of the time series segments to be predicted is input into a preset neural network prediction model to determine the abnormal time series segments in the at least one time series segment to be predicted.

7. The method according to claim 6, characterized in that, The step of selecting at least one time series segment to be predicted from the plurality of first time series segments includes: The start and end data in each of the first time segments are compared to obtain the comparison results; If the comparison result indicates that the change between the starting data and the ending data is greater than or equal to a preset change, the first time segment corresponding to the comparison result is determined as the time segment to be predicted.

8. The method according to claim 1, characterized in that, The monitoring time series is a time series within the geological disaster monitoring environment.

9. A device for determining abnormal time segments, characterized in that, include: The time series acquisition module is used to acquire the monitoring time series under the monitoring environment, wherein the monitoring time series includes multiple monitoring data that are continuous in time; wherein the monitoring data is at least one of temperature, humidity and ultraviolet intensity; A mutation data determination module is used to perform data mutation detection on the monitoring time series and determine at least one mutation data in the monitoring time series. The first time series segment division module is used to divide the monitoring time series into multiple first time series segments based on the at least one mutation data. An abnormal time series segment determination module is used to input each of the first time series segments into a preset neural network prediction model to determine the abnormal time series segments among the plurality of first time series segments, wherein the abnormal time series segment represents a time series segment that includes abnormal data. The abnormal time segment determination module is specifically used for: The target time sequence segment is concatenated with a preset time sequence segment from a plurality of preset time sequence samples to obtain a plurality of first concatenated time sequence segments associated with the target time sequence segment. The target time sequence segment is any one of the plurality of first time sequence segments. The preset time sequence sample further includes a time sequence segment identifier, which is used to indicate whether the preset time sequence segment is an abnormal time sequence segment. The plurality of first spliced ​​time series segments are respectively input into a preset neural network prediction model to obtain a plurality of first prediction results for the plurality of first spliced ​​time series segments. Each first prediction result is used to indicate whether the first spliced ​​time series segment corresponding to it is an abnormal time series segment. The plurality of first prediction results are compared with at least one second prediction result to obtain a comparison result. The at least one second prediction result is obtained by inputting at least one second spliced ​​time sequence segment into the neural network prediction model. The at least one second spliced ​​time sequence segment is obtained by splicing two preset time sequence segments from the plurality of preset time sequence samples. If the comparison result indicates that the difference between multiple first prediction results and at least one second prediction result exceeds a preset difference threshold, the target time series segment is determined to be an abnormal time series segment, so as to obtain the abnormal time series segment among the multiple first time series segments.

10. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method for determining abnormal timing segments as described in any one of claims 1-8.

11. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the method for determining abnormal timing segments as described in any one of claims 1-8.

Images