Multi-element time series data collaborative compression method and device and computer equipment

Abstract

The invention relates to a multivariate time series data collaborative compression method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a multivariate time sequence data set, wherein the multivariate time sequence data set comprises a plurality of associated original time sequences; key feature detection processing is carried out on the original time sequences, a feature mark set is obtained, and the feature mark set comprises feature mark sequences corresponding to the original time sequences; performing merging processing on each feature tag sequence in the feature tag set to obtain a collaborative feature tag sequence; and performing cooperative compression processing on each original time sequence based on the cooperative feature marking sequence to obtain a compression result of each original time sequence. By adopting the method, the data analysis accuracy and the mining utilization accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of big data processing technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium and computer program product for collaborative compression of multivariate time series data. Background Art

[0002] In large-scale distributed systems, monitoring programs for objects such as servers, cloud hosts, network devices, and applications will continuously generate a large amount of monitoring data. These data are usually indexed by timestamps and record the status and behavior of the monitored objects at different time points, forming a time series data set. Since these data are generated quickly and accumulate over time, if they are not effectively managed and compressed, they will lead to a large consumption of storage space. In addition, if there is a large amount of redundant data in the time series data, it will also lead to an increase in the cost of subsequent data analysis and processing.

[0003] Usually, when compressing multivariate time series data, it is based on compressing single variables one by one, which may destroy the change correlation between variables and is not conducive to the subsequent time series data analysis and mining utilization. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for collaborative compression of multivariate time series data that can improve the accuracy of data analysis and mining and utilization in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for collaborative compression of multivariate time series data. The method comprises: obtaining a multivariate time series data set, the multivariate time series data set comprising a plurality of associated original time series; performing key feature detection processing on each original time series to obtain a feature marker set, the feature marker set comprising a feature marker sequence corresponding to each original time series; merging each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; based on the collaborative feature marker sequence, performing collaborative compression processing on each original time series to obtain a compression result of each original time series.

[0006] In one embodiment, key feature detection processing is performed on each original time series to obtain a feature marker set, including: for each original time series, first-order difference processing is performed on the original time series to obtain a first-order difference sequence of the original time series; the first quartile and the third quartile of the first-order difference sequence are obtained; based on the first quartile, the third quartile and a preset coefficient, a key feature detection threshold is determined; based on the comparison results of each differential value in the first-order difference sequence with the key feature detection threshold, a feature marker sequence corresponding to the original time series is obtained.

[0007] In one embodiment, each signature sequence in a signature set is merged to obtain a collaborative signature sequence, including: for each signature sequence in the signature set, summing each signature value in the signature sequence to obtain a total signature value of the signature sequence; sorting the total signature values ​​of each signature sequence in the signature set in order from large to small to obtain an updated signature set; and obtaining a collaborative signature sequence based on each sampling value in each sequence in the updated signature set.

[0008] In one embodiment, based on each sampling value in each sequence in the update mark set, a collaborative feature marker sequence is obtained, including: for each sequence in the update mark set, determining the first sequence as the target sequence, and determining the first sampling value in the target sequence as the target sampling value; determining the numerical relationship between the target sampling value and the first preset value; when the target sampling value is the same as the first preset value, determining the first preset value as the collaborative feature value, and determining the next sampling value of the target sampling value in the target sequence as the new target sampling value, and returning to the step of determining the numerical relationship between the target sampling value and the first preset value until the new target sampling value is the last sampling value in the target sequence; when the target sampling value is different from the first preset value, determining the next sequence of the target sequence in the update mark set as the new target sequence, returning to the step of determining the first sampling value in the target sequence as the target sampling value, until the new target sequence is the last sequence in the update mark set, and when the first sampling value in the last sequence is different from the first preset value, determining the second preset value as the collaborative feature value; based on the determined collaborative feature value, the collaborative feature marker sequence is obtained.

[0009] In one embodiment, based on the collaborative feature marker sequence, collaborative compression processing is performed on each original time series to obtain the compression results of each original time series, including: expanding the collaborative feature marker sequence to obtain an updated feature sequence; correcting and optimizing the updated feature sequence to obtain a corrected feature sequence; for each feature value in the corrected feature sequence, based on the numerical relationship between the feature value and the first preset value and the second preset value, dividing the sampling time interval in each original time series to determine the key feature interval and the non-key feature interval in each original time series; for each original time series, while keeping the sampling values ​​in the key feature interval in the original time series unchanged, data reduction processing is performed on the sampling values ​​in the non-key feature interval in the original time series to obtain the compression results of the original time series.

[0010] In one embodiment, the updated feature sequence is corrected and optimized to obtain the corrected feature sequence, including: traversing each feature value in the updated feature sequence in turn, determining the first feature value currently traversed, and the second feature value adjacent to the first feature value; when the first feature value is the first preset value and the second feature value is the second preset value, updating the current count value and determining the third feature value adjacent to the second feature value; when the third feature value is the same as the first preset value and the updated current count value is less than or equal to the optimization threshold, updating the N feature values ​​adjacent to the third feature value before the third feature value to the first preset value, clearing the current count value, determining the feature value adjacent to the third feature value as the new first feature value, and returning to the step of determining the second feature value adjacent to the first feature value; N is the same as the optimization threshold; when the third feature value is the same as the first preset value and the updated current count value is less than or equal to the optimization threshold When the third eigenvalue is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; when the third eigenvalue is the same as the second preset value, the updated current count value is updated again; if the current count value after the update is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; if the current count value after the update is less than or equal to the optimization threshold, the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; after traversing each eigenvalue in the update feature sequence, a correction feature sequence is obtained based on the updated update feature sequence.

[0011] In a second aspect, the present application provides a device for collaborative compression of multivariate time series data, the device comprising: an acquisition module for acquiring a multivariate time series data set, the multivariate time series data set comprising a plurality of associated original time series; a first processing module for performing key feature detection processing on each original time series to obtain a feature marker set, the feature marker set comprising a feature marker sequence corresponding to each original time series; a second processing module for merging each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; and a third processing module for performing collaborative compression processing on each original time series based on the collaborative feature marker sequence to obtain a compression result for each original time series.

[0012] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program: obtaining a multivariate time series data set, the multivariate time series data set includes multiple associated original time series; performing key feature detection processing on each original time series to obtain a feature marker set, the feature marker set includes a feature marker sequence corresponding to each original time series; merging each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; based on the collaborative feature marker sequence, performing collaborative compression processing on each original time series to obtain a compression result of each original time series.

[0013] In a fourth aspect, the present application also provides a computer-readable storage medium. A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented: obtaining a multivariate time series data set, the multivariate time series data set includes multiple associated original time series; performing key feature detection processing on each original time series to obtain a feature marker set, the feature marker set includes a feature marker sequence corresponding to each original time series; merging each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; based on the collaborative feature marker sequence, performing collaborative compression processing on each original time series to obtain a compression result of each original time series.

[0014] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program, which implements the following steps when executed by a processor: obtaining a multivariate time series data set, the multivariate time series data set includes multiple associated original time series; performing key feature detection processing on each original time series to obtain a feature marker set, the feature marker set includes a feature marker sequence corresponding to each original time series; merging each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; based on the collaborative feature marker sequence, performing collaborative compression processing on each original time series to obtain a compression result of each original time series.

[0015] The above-mentioned multivariate time series data collaborative compression method, device, computer equipment, storage medium and computer program product obtain a multivariate time series data set, which includes multiple related original time series, and perform key feature detection processing on each original time series to obtain a feature marker set, which includes a feature marker sequence corresponding to each original time series, and merge each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence, and then perform collaborative compression processing on each original time series based on the collaborative feature marker sequence to obtain the compression result of each original time series. Therefore, by performing collaborative compression processing on each original time series, the change correlation between multiple original time series is taken into account, and then, when performing data analysis and mining and utilization on each compressed original time series, the data analysis accuracy and mining and utilization accuracy can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A diagram of an application environment of a multivariate time series data collaborative compression method in an embodiment;

[0017] Figure 2 It is a flowchart of a method for collaborative compression of multivariate time series data in one embodiment;

[0018] Figure 3 is a schematic diagram of an original time series in one embodiment;

[0019] Figure 4 A schematic diagram of a process of performing key feature detection processing on each original time series to obtain a feature tag set in one embodiment;

[0020] Figure 5 A schematic diagram of a feature marker sequence corresponding to an original time series in one embodiment;

[0021] Figure 6 A schematic diagram of a flow chart of merging the feature marker sequences in a feature marker set to obtain a collaborative feature marker sequence in one embodiment;

[0022] Figure 7 A schematic diagram of a process of obtaining a collaborative feature marker sequence based on each sampling value in each sequence in an update marker set in one embodiment;

[0023] Figure 8 is a schematic diagram of a collaborative feature marker sequence in one embodiment;

[0024] Fig. 9 A schematic diagram of a collaborative feature marker sequence in another embodiment;

[0025] Fig.10It is a flowchart of performing collaborative compression processing on each original time series based on a collaborative feature tag sequence to obtain compression results of each original time series in one embodiment;

[0026] Fig.11 A schematic diagram of comparing an updated feature sequence and a corrected feature sequence in one embodiment;

[0027] Fig.12 A schematic diagram for comparing an updated feature sequence and a corrected feature sequence in another embodiment;

[0028] Fig.13 is a schematic diagram of an original time series and a corresponding key feature interval in one embodiment;

[0029] Fig.14 A schematic diagram of a flow chart of a method for collaborative compression of multivariate time series data in another embodiment;

[0030] Fig.15 It is a structural block diagram of a multivariate time series data collaborative compression device in one embodiment;

[0031] Fig.16 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0033] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations; and the acquisition, storage, processing, transmission, etc. of data must comply with relevant laws and regulations.

[0034] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0035] The multivariate time series data collaborative compression method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers.

[0036] Specifically, the server 104 obtains a multivariate time series data set from the terminal 102, and the multivariate time series data set includes multiple related original time series. The server 104 performs key feature detection processing on each original time series to obtain a feature marker set. The feature marker set includes a feature marker sequence corresponding to each original time series, and each feature marker sequence in the feature marker set is merged to obtain a collaborative feature marker sequence. Therefore, based on the collaborative feature marker sequence, each original time series is collaboratively compressed to obtain a compression result of each original time series.

[0037] The terminal 102 may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0038] In one embodiment, Figure 2 As shown, a multivariate time series data collaborative compression method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0039] S202, obtaining a multivariate time series data set, where the multivariate time series data set includes a plurality of associated original time series.

[0040] Wherein, each original time series includes multiple sampling values ​​and a sampling time point corresponding to each sampling value, and each sampling time point in the sampling time interval corresponding to the original time series is a sampling time point corresponding to each sampling value. Each original time series corresponds to a variable, and the sampling value represents the sampling data of the variable corresponding to the original time series. There is a correlation between the multiple variables corresponding to each of the multiple original time series. The sampling time points corresponding to the sampling values ​​in different original time series are the same, and the number of sampling values ​​in different original time series is the same.

[0041] like Figure 3 As shown, a schematic diagram of an original time series is provided, wherein X1~X3 represent different original time series. For example, the original data series X1 can be represented as multiple sampled data of the central processing unit (CPU) utilization of the server, the original time series X2 can be represented as multiple sampled data of the memory utilization of the server, and the original time series X3 can be represented as multiple sampled data of the network utilization of the server.

[0042] In actual scenarios, when the variables corresponding to the original time series include CPU utilization and memory utilization, multiple sampled data corresponding to CPU utilization and multiple sampled data corresponding to memory utilization may increase or decrease at the same time, and anomalies or failures may also be transmitted between related services. Therefore, there is a certain correlation between CPU utilization and memory utilization.

[0043] In some cases, the Pearson correlation coefficient, covariance matrix analysis, principal component analysis, time series clustering and other methods can be used to perform correlation analysis on multiple sampling data between different original time series, so that the correlation between different original time series can be viewed more intuitively.

[0044] S204, performing key feature detection processing on each original time series to obtain a feature marker set, where the feature marker set includes a feature marker sequence corresponding to each original time series.

[0045] The feature tag sequence refers to a sequence composed of key feature values ​​in the original time series. There is no limitation on the implementation method of performing key feature detection on each original time series to obtain a feature tag set. The following is an example of possible implementation methods.

[0046] In one embodiment, for each original time series, the average of multiple sampling values ​​in the original time series is determined as the key feature detection threshold; for each sampling value in the multiple sampling values, the sampling value greater than the key feature detection threshold is updated to a first preset value, and the sampling value less than or equal to the key feature detection threshold is updated to a second preset value; according to the order of the sampling values ​​in the original time series, the updated sampling values ​​are combined to obtain a feature marker sequence corresponding to the original time series.

[0047] The sampling value in the original time series that is greater than the key feature detection threshold is the key feature value in the original time series. The first preset value can be 1, the second preset value can be 0, and the first preset value and the second preset value can also have other settings, which are not limited in this embodiment.

[0048] S206: Merge the signature sequences in the signature set to obtain a collaborative signature sequence.

[0049] There is no limitation on the implementation method of merging the signature sequences in the signature set to obtain the collaborative signature sequence. For example, the signature sequences may be directly spliced ​​to obtain the collaborative signature sequence, or other methods may be used, which are not limited in this embodiment.

[0050] S208, based on the collaborative feature tag sequence, perform collaborative compression processing on each original time series to obtain compression results of each original time series.

[0051] The number of collaborative feature values ​​in the collaborative feature tag sequence is the same as the number of sample values ​​in each original time series. The sampling time points when different original time series are compressed are the same, so collaborative compression can be achieved by compressing the sample values ​​at the same sampling time points of each original time series.

[0052] There is no limitation on the implementation method of performing collaborative compression processing on each original time series based on the collaborative feature tag sequence to obtain the compression results of each original time series. The following is an example of possible implementation methods.

[0053] In one embodiment, for each collaborative feature value in the collaborative feature marker sequence, the arrangement position of the collaborative feature value in the collaborative feature marker sequence is determined; for the collaborative feature value in the collaborative feature marker sequence that is different from the first preset value, multiple sampling values ​​matching the arrangement position corresponding to the collaborative feature value are obtained from each original time series, and the sampling mean of the multiple sampling values ​​is determined; and the compression results of each original time series are obtained when the sampling values ​​matching the arrangement position in each original time series are updated to the sampling mean. Among them, for the collaborative feature value in the collaborative feature marker sequence that is the same as the first preset value, the sampling values ​​matching the arrangement position corresponding to the collaborative feature value in each original time series are not adjusted.

[0054] based on Figure 2 The method shown, by obtaining a multivariate time series data set, the multivariate time series data set includes multiple related original time series, and performing key feature detection processing on each original time series, obtaining a feature tag set, the feature tag set includes a feature tag sequence corresponding to each original time series, and merging each feature tag sequence in the feature tag set to obtain a collaborative feature tag sequence, and then based on the collaborative feature tag sequence, performing collaborative compression processing on each original time series to obtain the compression results of each original time series. Therefore, by performing collaborative compression processing on each original time series, the change correlation between multiple original time series is considered, and then, when performing data analysis and mining and utilization on each compressed original time series, the accuracy of data analysis and mining and utilization can be improved.

[0055] In combination with the above content, it can be known that, when the mean of multiple sampling values ​​in the original time series is determined as the key feature detection threshold, the feature tag sequence corresponding to the original time series can be obtained based on the key feature detection threshold. In another embodiment, the key feature detection process is performed on each original time series to obtain the feature tag set (i.e., S204), and the implementation method can also be as follows Figure 4 As shown, the following steps are included:

[0056] S402: For each original time series, perform first-order difference processing on the original time series to obtain a first-order difference sequence of the original time series.

[0057] In this embodiment, the first-order difference sequence refers to a new sequence obtained by performing a difference operation on the original time series. Specifically, the difference between each sample value in the original time series and the sample value before each sample value is calculated to obtain the first-order difference sequence. In order to fill the sequence length, the first eigenvalue in the first-order difference sequence can be set to 0, and the second eigenvalue is the difference between the second sample value and the first sample value in the original time series, and so on.

[0058] For example, a multivariate time series dataset , that is, the multivariate time series data set includes n original time series; for any original time series , that is, the original time series Including m sampling values. Performing first-order difference processing, we can get the first-order difference sequence ,in, , .

[0059] S404, obtaining the first quartile and the third quartile of the first-order difference sequence.

[0060] The first quartile refers to the 25th percentile eigenvalue after all eigenvalues ​​in the first-order difference sequence are arranged from small to large, and the third quartile refers to the 75th percentile eigenvalue after all eigenvalues ​​in the first-order difference sequence are arranged from small to large.

[0061] S406, determining a key feature detection threshold based on the first quartile, the third quartile and a preset coefficient.

[0062] There is no limitation on the implementation method of determining the key feature detection threshold based on the first quartile, the third quartile and the preset coefficient, and several of the methods are described below with examples.

[0063] In one embodiment, the difference between the third quartile and the first quartile is obtained; and the product of the difference and a preset coefficient and the sum of the third quartile are determined as the key feature detection threshold.

[0064] Among them, the preset coefficient is greater than or equal to the first setting coefficient, and less than or equal to the second setting coefficient. For example, the first setting coefficient can be 1, and the second setting coefficient can be 3. The first setting coefficient and the second setting coefficient can also be set to other values.

[0065] For example, taking i=1 and the first preset coefficient as 1.5, we can calculate The first quartile Q1 and the third quartile Q3, then the key feature detection threshold C = Q3 + 1.5 × (Q3-Q1).

[0066] In one embodiment, based on the correlation between the mean and the preset range, the preset range matching the mean between the third quartile and the first quartile is determined as the target range; based on the correspondence between the preset range and the reference coefficient, the reference coefficient corresponding to the target range is determined as the target coefficient; the coefficient mean of the target coefficient and the preset coefficient is determined; the product of the mean between the third quartile and the first quartile and the coefficient mean is determined as the key feature detection threshold.

[0067] S408, based on the comparison results of each difference value in the first-order difference sequence with the key feature detection threshold, obtain a feature label sequence corresponding to the original time series.

[0068] Specifically, for the first eigenvalue in the first-order difference sequence, the first eigenvalue is determined as the feature mark value; for the other eigenvalues ​​in the first-order difference sequence except the first eigenvalue, when the eigenvalue is greater than or equal to the key feature detection threshold, the first preset value is determined as the feature mark value; when the eigenvalue is less than the key feature detection threshold, the second preset value is determined as the feature mark value; according to the order of each difference value in the first-order difference sequence, the corresponding feature mark values ​​are combined to obtain the feature mark sequence corresponding to the original time series. For example, combined Figure 3 The original time series shown is Figure 5 As shown, the feature marker sequences corresponding to the original time series X1~X3 are provided.

[0069] For example, taking the first preset value as 1 and the second preset value as 0 as an example, the feature marker sequence set , that is, the feature tag sequence set S includes the feature tag sequences corresponding to n original time series; based on any original time series The first-order difference sequence of , when j=1, ,exist hour, , C represents the key feature detection threshold, then the original time series can be obtained The corresponding signature sequence .

[0070] based on Figure 4The content described above performs first-order difference processing on the original time series, so that the first-order difference can amplify the influence of outliers in the original time series and identify key fluctuation points in the original time series, thereby improving the accuracy of the obtained feature marker sequence.

[0071] In one embodiment, Figure 6 As shown, a flow chart of merging the feature marker sequences in the feature marker set to obtain a collaborative feature marker sequence is provided, including the following steps:

[0072] S602: For each signature sequence in the signature set, sum up each signature value in the signature sequence to obtain a total signature value of the signature sequence.

[0073] S604, sorting the total feature values ​​of each feature marker sequence in the feature marker set in descending order to obtain an updated marker set.

[0074] The total characteristic values ​​of each sequence in the update mark set decrease in sequence, and each sequence in the update mark set is a characteristic mark sequence corresponding to the original time series. Can be expressed , The eigenvalues ​​and The corresponding feature values ​​in . For example, the feature marker set is expressed as , then update the feature tag set, and you can get the updated tag set ,in, , , .

[0075] S606: Obtain a collaborative feature marker sequence based on each sampling value in each sequence in the update marker set.

[0076] There is no limitation on the implementation method of obtaining the collaborative feature marker sequence based on each sampling value in each sequence in the update marker set, and the following is an example of possible implementation methods.

[0077] In one embodiment, each sample value in each sequence in the update tag set is input into a pre-trained collaborative tag model to obtain a collaborative feature tag sequence. The collaborative tag model can be trained based on multiple training time series and the collaborative feature tag sequences corresponding to the multiple training time series.

[0078] based on Figure 6The content shown is that by sorting the total characteristic values ​​of each feature sequence in the feature set in order from large to small, the accuracy of the obtained collaborative feature sequence can be improved while taking into account all the sampling values ​​in the feature sequence corresponding to each original time series.

[0079] In combination with the above content, it can be known that the collaborative feature signature sequence can be obtained based on the collaborative signature model. In another embodiment, based on each sampling value in each sequence in the update signature set, the implementation method of obtaining the collaborative feature signature sequence (ie, S606) can also be as follows: Figure 7 As shown, the following steps are included:

[0080] S702: For each sequence in the update mark set, determine the first sequence as a target sequence, and determine the first sampling value in the target sequence as a target sampling value.

[0081] S704, determining a numerical relationship between the target sampling value and the first preset value.

[0082] The numerical relationship represents the result of whether the target sampling value is the same as the first preset value, and the numerical relationship includes that the target sampling value is the same as the first preset value, or that the target sampling value is different from the first preset value.

[0083] S706, when the target sampling value is the same as the first preset value, the first preset value is determined as the collaborative feature value, and the next sampling value of the target sampling value in the target sequence is determined as the new target sampling value, and the step of determining the numerical relationship between the target sampling value and the first preset value is returned until the new target sampling value is the last sampling value in the target sequence.

[0084] S708, when the target sampling value is different from the first preset value, the next sequence of the target sequence in the update mark set is determined as the new target sequence, and the step of determining the first sampling value in the target sequence as the target sampling value is returned until the new target sequence is the last sequence in the update mark set, and when the first sampling value in the last sequence is different from the first preset value, the second preset value is determined as the collaborative feature value.

[0085] S710: Obtain a collaborative feature tag sequence based on the determined collaborative feature value.

[0086] The contents described in S702-S710 may also be described in the following manner. For example, assuming that the number of original time series is n and the number of feature mark values ​​in the feature mark sequence corresponding to any original time series is m, the specific description is:

[0087] Step 1.1, initialize the all-zero sequence R of length m, and initialize i=1, j=1.

[0088] Step 1.2: Determine the i-th sequence in the update tag set The jth element in Is the value of 1? ,but , and execute step 1.3; if , , and repeat this step until the jth element in each sequence in the update mark set is traversed. .

[0089] Step 1.3, , , and return to step 1.2.

[0090] Based on steps 1.1 to 1.3, the all-zero sequence R of length m can be updated, and the updated That is the collaborative feature marker sequence. For example, based on steps 1.1 to 1.3, Figure 8 As shown in FIG. 1 , a schematic diagram of traversing each sequence in the update tag set to obtain a collaborative feature tag sequence is provided. Figure 3 The original time series X1~X3 in Fig. 9 As shown, a schematic diagram of a collaborative feature marker sequence is provided.

[0091] Based on the summation of the marker feature values ​​in each feature marker sequence, Figure 7 The method shown, when traversing the eigenvalues ​​in each sequence in descending order of the sorted sum values, in the sequence with the largest sum value, the elements of the first preset value are the largest, thereby increasing the probability of timely termination of traversal of other sequences, especially when the number of variables is large, can significantly reduce the time complexity of the calculation, and more quickly perform collaborative interval division on multiple time series.

[0092] In one embodiment, based on the collaborative feature tag sequence, the original time series are collaboratively compressed to obtain the compression results of the original time series (ie, S208), which can be implemented as follows: Fig.10 As shown, the following steps are included:

[0093] S1002, performing expansion processing on the collaborative feature tag sequence to obtain an updated feature sequence.

[0094] Specifically, for feature values ​​other than the first feature value in the collaborative feature marking sequence, the second feature value is determined as the first feature value, and the feature value before the first feature value is determined as the second feature value; when the first feature value is the same as the first preset value, and the second feature value is the same as the second preset value, the feature value before the first feature value is updated to the first preset value, and the third feature value is determined as the new first feature value, and the step of determining the feature value before the first feature value as the second feature value is returned until the new first feature value is the last feature value in the collaborative feature marking sequence.

[0095] For example, collaborative feature marker sequences ,if ,and , then Set to 1.

[0096] S1004, performing correction and optimization processing on the updated feature sequence to obtain a corrected feature sequence.

[0097] There is no limitation on the implementation method of performing correction processing on the updated feature sequence to obtain the corrected feature sequence, and the following is an example of possible implementation methods.

[0098] In one embodiment, for each feature value in the updated feature sequence, the first preset value that appears continuously is counted, and when the feature value after the first preset value that appears continuously is the second preset value, and the feature value after the second preset value is the first preset value that appears continuously, the second preset value is updated to the first preset value to obtain the correction feature sequence. The number of feature values ​​after the first preset value that appears continuously is at least one.

[0099] In one embodiment, the calibration feature sequence may be obtained by the following steps, specifically:

[0100] S1, traverse each feature value in the updated feature sequence in turn, and determine the first feature value currently traversed and the second feature value adjacent to the first feature value.

[0101] For example, when starting to traverse each eigenvalue in the updated feature sequence, the first eigenvalue refers to the first eigenvalue in the updated feature sequence, and the second eigenvalue refers to the second eigenvalue.

[0102] S2, when the first characteristic value is the first preset value and the second characteristic value is the second preset value, updating the current count value and determining a third characteristic value adjacent to the second characteristic value.

[0103] Among them, when the first eigenvalue is the first eigenvalue in the updated eigenvalue sequence, the second eigenvalue is the second eigenvalue, and the current count value Count=0, the updated current count value Count=1, and the third eigenvalue adjacent to the second eigenvalue refers to the third eigenvalue in the updated eigenvalue sequence.

[0104] S3, when the third eigenvalue is the same as the first preset value and the updated current count value is less than or equal to the optimization threshold, the N eigenvalues ​​adjacent to the third eigenvalue are updated to the first preset value, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned.

[0105] Wherein, N is the same as the optimization threshold; when the third eigenvalue refers to the third eigenvalue in the updated feature sequence, the eigenvalue adjacent to the third eigenvalue refers to the fourth eigenvalue in the updated feature sequence.

[0106] S4, when the third eigenvalue is the same as the first preset value and the updated current count value is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned.

[0107] S5, when the third characteristic value is the same as the second preset value, updating the updated current count value again.

[0108] S6, if the current count value after the update is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the process returns to the step of determining the second eigenvalue adjacent to the first eigenvalue.

[0109] S7, if the current count value after the update is less than or equal to the optimization threshold, the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the process returns to the step of determining the second eigenvalue adjacent to the first eigenvalue.

[0110] S8, after traversing all feature values ​​in the updated feature sequence, obtaining a corrected feature sequence based on the updated feature sequence.

[0111] The contents described in S1-S8 may also be described in the following manner, specifically:

[0112] Step 2.1, Initialization , , , optimization threshold = D.

[0113] Step 2.2, if ,and ,and , ,but , , , and execute step 2.3; otherwise, , and return to execute step 2.2 until the updated feature sequence R is traversed.

[0114] Step 2.3, if ,but , and execute step 2.4; if ,and , then execute step 2.5 to optimize the data in this section.

[0115] Step 2.4, if , the counting is stopped and the , , , and return to step 2.2; if ,but , and return to step 2.3.

[0116] Step 2.5, Set both to 1, and then set , , , and return to step 2.2.

[0117] Based on steps 2.1 to 2.5, the correction feature sequence can be obtained. For example, taking the optimization threshold D=2 as an example, Fig.11 and Fig.12 As shown in FIG. 1 , a comparison diagram of an updated feature sequence and a corrected feature sequence is provided. Fig.11 and Fig.12 It can be seen from the figure that the correction and optimization processing of the update feature sequence is essentially to modify part of the second preset values ​​in the update feature sequence to the first preset values.

[0118] S1006, for each eigenvalue in the corrected feature sequence, based on the numerical relationship between the eigenvalue and the first preset value and the second preset value, the sampling time interval corresponding to each original time series is divided to determine the key feature interval and the non-key feature interval in each original time series.

[0119] Among them, the sampling time interval corresponding to the original time series includes the sampling time points of each sampling value in the original time series, and the position of each characteristic value in the corrected characteristic sequence in the corrected characteristic sequence corresponds to the position of each sampling value in the original time series in the original time series.

[0120] Specifically, for each eigenvalue in the corrected feature sequence, when the eigenvalue is the same as the first preset value, the sampling time point of the sampling value corresponding to the eigenvalue is determined as the key feature point; when the eigenvalue is the same as the second preset value, the sampling time point of the sampling value corresponding to the eigenvalue is determined as the non-key feature point; according to the order of each sampling value in each original time series, the key feature points and non-key feature points corresponding to each original time series are respectively combined to determine the key feature interval and non-key feature interval in each original time series.

[0121] S1008, for each original time series, while keeping the sampling values ​​in the key feature interval of the original time series unchanged, perform data reduction processing on the sampling values ​​in the non-key feature interval of the original time series to obtain the compression result of the original time series.

[0122] There is no limit to the implementation method of data reduction processing for the sample values ​​in the non-critical feature interval in the original time series. For example, for the non-critical feature interval composed of non-critical feature points that appear continuously in the original time series, the mean of each sample value in the non-critical feature interval is determined as the updated value of each sample value in the non-critical feature interval, or the middle value of each sample value in the non-critical feature interval is determined as the updated value of each sample value in the non-critical feature interval.

[0123] Based on the contents shown in S1006 and S1008, Figure 3 The original time series X1~X3 in Fig.13 As shown in Figure 1, a schematic diagram of the original time series and the corresponding key feature intervals is provided. Fig.13 It can be seen that the key feature intervals corresponding to each original time series are the same, so the coordinated compression of different original time series can be achieved.

[0124] based on Fig.10 As shown in the figure, by dividing the sampling time interval corresponding to each original time series into key feature interval and non-key feature interval, lossless compression is performed on the key feature interval, and lossy compression is performed on the data in the non-key feature interval. On the one hand, the alignment of key features in the time dimension is ensured and the correlation between multiple variables is retained. On the other hand, the storage cost is reduced and the data utilization rate is improved.

[0125] In combination with the above, in one embodiment, Fig.14 As shown, a multivariate time series data collaborative compression method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps may be included:

[0126] S1402, obtaining a multivariate time series data set, where the multivariate time series data set includes a plurality of associated original time series.

[0127] S1404: For each original time series, perform first-order difference processing on the original time series to obtain a first-order difference sequence of the original time series.

[0128] S1406, obtaining the first quartile and the third quartile of the first-order difference sequence, and determining a key feature detection threshold based on the first quartile, the third quartile and a preset coefficient.

[0129] S1408: Based on the comparison results of each difference value in the first-order difference sequence with the key feature detection threshold, a feature label sequence corresponding to the original time series is obtained.

[0130] S1410 , for each signature sequence in the signature set, summing up each signature value in the signature sequence to obtain a total signature value of the signature sequence.

[0131] S1412, sorting the total feature values ​​of each feature marker sequence in the feature marker set in descending order to obtain an updated marker set.

[0132] S1414: Obtain a collaborative feature marker sequence based on each sampling value in each sequence in the update marker set.

[0133] Among them, the sequence in the update mark set is represented as ,and The sample values ​​in are expressed as In the case of , the collaborative feature marker sequence can be obtained based on steps 1.1 to 1.3, specifically:

[0134] Step 1.1, initialize the all-zero sequence R of length m, initialize i=1, j=1;

[0135] Step 1.2: Determine the i-th sequence in the update tag set The jth element in Is the value of 1? ,but , and execute step 1.3; if ,but , and repeat this step until the jth element in each sequence in the update mark set is traversed. ,;

[0136] Step 1.3, , , , and return to step 1.2.

[0137] Based on steps 1.1 to 1.3, the all-zero sequence R of length m can be updated, and the updated That is the collaborative feature marker sequence.

[0138] S1416, performing expansion processing on the collaborative feature tag sequence to obtain an updated feature sequence.

[0139] For example, collaborative feature marker sequences ,if ,and , then Set to 1.

[0140] S1418, performing correction and optimization processing on the updated feature sequence to obtain a corrected feature sequence.

[0141] Among them, based on the contents described in S1-S8 above, the updated feature sequence can be corrected and optimized to obtain a corrected feature sequence.

[0142] S1420, for each characteristic value in the correction characteristic sequence, based on the numerical relationship between the characteristic value and the first preset value and the second preset value, the sampling time interval in each original time series is divided to determine the key characteristic interval and the non-key characteristic interval in each original time series.

[0143] S1422, for each original time series, while keeping the sampling values ​​in the key feature interval of the original time series unchanged, perform data reduction processing on the sampling values ​​in the non-key feature interval of the original time series to obtain the compression result of the original time series.

[0144] Among them, the contents of S1402-S1422 can be adapted to the description by referring to the aforementioned contents and will not be repeated here.

[0145] based on Fig.14The content shown in the figure, on the one hand, uses the first-order difference to amplify the influence of outliers, identifies the key fluctuation points in the original time series, and thus obtains the feature marker sequence and the collaborative feature marker sequence of each original time series; according to the calculation characteristics of the first-order difference, the collaborative feature marker sequence is expanded, and the key points are automatically extended forward by 1 bit, retaining the key reference data, and improving the accuracy and interpretability of the compressed data after decompression. On the other hand, each feature marker sequence is summed, and then the elements in each sequence are traversed in descending order according to the sum value of each sequence. In the sequence with the largest sum value, the number of elements 1 is the largest, thereby increasing the probability of timely termination of traversal of other sequences, especially when the number of variables is large, which can significantly reduce the time complexity of the calculation, and more quickly divide multiple original time series into collaborative intervals. On the other hand, by setting the two parameters of key feature detection threshold and optimization threshold, the compression rate and the proportion of key data can be flexibly adjusted according to needs. By dividing the data into key feature intervals and non-key feature intervals, lossless compression is performed on the key feature intervals, and lossy compression is performed on the data in the non-key feature intervals. On the one hand, this ensures the alignment of key features in the time dimension and retains the correlation between multiple variables. On the other hand, it also reduces storage costs and improves data utilization.

[0146] Furthermore, when the subsequent time series data processing and analysis requires a full-length sequence, the compressed data can be quickly restored based on the sampling frequency and time interval. The reduced part can be restored by means of mean filling, noise addition, etc., and the important fluctuation feature part can achieve lossless restoration of the original value.

[0147] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0148] Based on the same inventive concept, the embodiment of the present application also provides a multivariate time series data collaborative compression device for implementing the multivariate time series data collaborative compression method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more multivariate time series data collaborative compression device embodiments provided below can refer to the limitations of the multivariate time series data collaborative compression method above, and will not be repeated here.

[0149] In one embodiment, Fig.15 As shown, a device for collaborative compression of multivariate time series data is provided, including: an acquisition module 1502, a first processing module 1504, a second processing module 1506 and a third processing module 1508, wherein: the acquisition module 1502 is used to acquire a multivariate time series data set, the multivariate time series data set includes multiple associated original time series; the first processing module 1504 is used to perform key feature detection processing on each original time series to obtain a feature marker set, the feature marker set includes a feature marker sequence corresponding to each original time series; the second processing module 1506 is used to merge each feature marker sequence in the feature marker set to obtain a collaborative feature marker sequence; the third processing module 1508 is used to perform collaborative compression processing on each original time series based on the collaborative feature marker sequence to obtain a compression result of each original time series.

[0150] In one embodiment, the first processing module is also used to: perform first-order difference processing on each original time series to obtain a first-order difference sequence of the original time series; obtain the first quartile and the third quartile of the first-order difference sequence; determine the key feature detection threshold based on the first quartile, the third quartile and the preset coefficient; and obtain the feature marker sequence corresponding to the original time series based on the comparison results of each differential value in the first-order difference sequence with the key feature detection threshold.

[0151] In one embodiment, the second processing module is also used to: for each feature marker sequence in the feature marker set, sum up each feature marker value in the feature marker sequence to obtain a total feature value of the feature marker sequence; sort the total feature values ​​of each feature marker sequence in the feature marker set in order from large to small to obtain an updated marker set; and obtain a collaborative feature marker sequence based on each sampling value in each sequence in the updated marker set.

[0152] In one embodiment, the second processing module is further used to: for each sequence in the update mark set, determine the first sequence as the target sequence, and determine the first sampling value in the target sequence as the target sampling value; determine the numerical relationship between the target sampling value and the first preset value; when the target sampling value is the same as the first preset value, determine the first preset value as the collaborative feature value, and determine the next sampling value of the target sampling value in the target sequence as the new target sampling value, and return to the step of determining the numerical relationship between the target sampling value and the first preset value until the new target sampling value is the last sampling value in the target sequence; when the target sampling value is different from the first preset value, determine the next sequence of the target sequence in the update mark set as the new target sequence, return to the step of determining the first sampling value in the target sequence as the target sampling value, until the new target sequence is the last sequence in the update mark set, and when the first sampling value in the last sequence is different from the first preset value, determine the second preset value as the collaborative feature value; based on the determined collaborative feature value, obtain the collaborative feature mark sequence.

[0153] In one embodiment, the third processing module is also used to: perform expansion processing on the collaborative feature mark sequence to obtain an updated feature sequence; perform correction and optimization processing on the updated feature sequence to obtain a corrected feature sequence; for each feature value in the corrected feature sequence, based on the numerical relationship between the feature value and the first preset value and the second preset value, divide the sampling time interval in each original time series to determine the key feature interval and the non-key feature interval in each original time series; for each original time series, while keeping the sampling values ​​in the key feature interval in the original time series unchanged, perform data reduction processing on the sampling values ​​in the non-key feature interval in the original time series to obtain the compression result of the original time series.

[0154] In one embodiment, the third processing module is further used to: sequentially traverse each feature value in the updated feature sequence, determine the first feature value currently traversed, and the second feature value adjacent to the first feature value; when the first feature value is the first preset value and the second feature value is the second preset value, update the current count value and determine the third feature value adjacent to the second feature value; when the third feature value is the same as the first preset value and the updated current count value is less than or equal to the optimization threshold, update the N feature values ​​adjacent to the third feature value before to the first preset value, clear the current count value, determine the feature value adjacent to the third feature value as the new first feature value, and return to the step of determining the second feature value adjacent to the first feature value; N is the same as the optimization threshold; when the third feature value is the same as the first preset value and the updated current count value is greater than the optimization threshold In this case, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; when the third eigenvalue is the same as the second preset value, the updated current count value is updated again; if the current count value after the update is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; if the current count value after the update is less than or equal to the optimization threshold, the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; after traversing each eigenvalue in the update feature sequence, a correction feature sequence is obtained based on the updated update feature sequence.

[0155] Each module in the above-mentioned multivariate time series data collaborative compression device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0156] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.15As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as multivariate time series data sets. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for collaborative compression of multivariate time series data is implemented.

[0157] Those skilled in the art will understand that Fig.15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for collaborative compression of multivariate time series data, characterized in that: The method comprises: Acquire a multivariate time series data set, wherein the multivariate time series data set includes a plurality of associated original time series; Performing key feature detection processing on each of the original time series to obtain a feature marker set, wherein the feature marker set includes a feature marker sequence corresponding to each of the original time series; Merging the feature marker sequences in the feature marker set to obtain a collaborative feature marker sequence; Based on the collaborative feature mark sequence, collaborative compression processing is performed on each of the original time series to obtain compression results of each of the original time series.

2. The method according to claim 1, characterized in that The performing key feature detection processing on each of the original time series to obtain a feature tag set includes: For each of the original time series, performing first-order difference processing on the original time series to obtain a first-order difference sequence of the original time series; Obtaining the first quartile and the third quartile of the first-order difference sequence; Determining a key feature detection threshold based on the first quartile, the third quartile and a preset coefficient; Based on the comparison results of each difference value in the first-order difference sequence with the key feature detection threshold, a feature marker sequence corresponding to the original time series is obtained.

3. The method according to claim 1, characterized in that The merging of the feature marker sequences in the feature marker set to obtain a collaborative feature marker sequence includes: For each of the feature marker sequences in the feature marker set, summing up the feature marker values ​​in the feature marker sequence to obtain a total feature value of the feature marker sequence; Sort the total feature values ​​of the feature marker sequences in the feature marker set in descending order to obtain an update marker set; Based on each sampling value in each sequence in the update mark set, a collaborative feature mark sequence is obtained.

4. The method according to claim 3, characterized in that The obtaining of the collaborative feature marker sequence based on each sampling value in each sequence in the update marker set comprises: For each of the sequences in the update mark set, determining the first sequence as a target sequence, and determining the first sampling value in the target sequence as a target sampling value; Determine a numerical relationship between the target sampling value and a first preset value; In the case where the target sampling value is the same as the first preset value, the first preset value is determined as the collaborative feature value, and the next sampling value of the target sampling value in the target sequence is determined as the new target sampling value, and the step of determining the numerical relationship between the target sampling value and the first preset value is returned until the new target sampling value is the last sampling value in the target sequence; In the case where the target sampling value is different from the first preset value, determining the next sequence of the target sequence in the update mark set as a new target sequence, returning to the step of determining the first sampling value in the target sequence as the target sampling value, until the new target sequence is the last sequence in the update mark set, and when the first sampling value in the last sequence is different from the first preset value, determining the second preset value as the collaborative feature value; Based on the determined collaborative feature value, a collaborative feature signature sequence is obtained.

5. The method according to any one of claims 1 to 4, characterized in that: The step of performing collaborative compression processing on each of the original time series based on the collaborative feature mark sequence to obtain compression results of each of the original time series includes: Performing expansion processing on the collaborative feature marker sequence to obtain an updated feature sequence; Performing correction and optimization processing on the updated feature sequence to obtain a corrected feature sequence; For each characteristic value in the correction characteristic sequence, based on the numerical relationship between the characteristic value and the first preset value and the second preset value, the sampling time interval in each of the original time series is divided to determine the key characteristic interval and the non-key characteristic interval in each of the original time series; For each of the original time series, while keeping the sampling values ​​in the key feature interval of the original time series unchanged, data reduction processing is performed on the sampling values ​​in the non-key feature interval of the original time series to obtain the compression result of the original time series.

6. The method according to claim 5, characterized in that The step of performing correction optimization processing on the updated feature sequence to obtain a corrected feature sequence includes: Traversing each feature value in the updated feature sequence in turn, and determining a first feature value currently traversed and a second feature value adjacent to the first feature value; When the first characteristic value is a first preset value and the second characteristic value is a second preset value, updating the current count value and determining a third characteristic value adjacent to the second characteristic value; In the case where the third eigenvalue is the same as the first preset value and the updated current count value is less than or equal to the optimization threshold, the N eigenvalues ​​adjacent to the third eigenvalue are updated to the first preset value, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; N is the same as the optimization threshold; When the third eigenvalue is the same as the first preset value and the updated current count value is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the step of determining the second eigenvalue adjacent to the first eigenvalue is returned; When the third characteristic value is the same as the second preset value, updating the updated current count value again; If the current count value after the update is greater than the optimization threshold, the current count value is cleared, and the eigenvalue adjacent to the third eigenvalue is determined as the new first eigenvalue, and the process returns to the step of determining the second eigenvalue adjacent to the first eigenvalue; If the current count value after the update is less than or equal to the optimization threshold, the eigenvalue adjacent to the third eigenvalue is determined as a new first eigenvalue, and the process returns to the step of determining the second eigenvalue adjacent to the first eigenvalue; After all the feature values ​​in the updated feature sequence are traversed, a correction feature sequence is obtained based on the updated feature sequence.

7. A multivariate time series data collaborative compression device, characterized in that: The device comprises: An acquisition module, used for acquiring a multivariate time series data set, wherein the multivariate time series data set includes a plurality of associated original time series; A first processing module is used to perform key feature detection processing on each of the original time series to obtain a feature marker set, wherein the feature marker set includes a feature marker sequence corresponding to each of the original time series; A second processing module is used to merge the feature marker sequences in the feature marker set to obtain a collaborative feature marker sequence; The third processing module is used to perform collaborative compression processing on each of the original time series based on the collaborative feature mark sequence to obtain a compression result of each of the original time series.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images