A method, device and medium for screening data playback time zone
By preprocessing and mining of time series data, combining user attention and sliding time windows, the key time periods in massive time series data are dynamically screened out, solving the problem of incomplete information display in the existing technology and achieving more accurate and efficient data screening and presentation.
Patent Information
- Application Number
- CN202411237648.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-09-05
AI Technical Summary
When faced with massive time-series big data, it is difficult for the existing technology to accurately capture important characteristics of data, especially in industrial data with high volatility and high complexity. A single standard screening method cannot adapt to the variability and multidimensionality of data. This leads to incomplete information display during the identification and screening of key time points, making it difficult for users to quickly capture key information.
A method for filtering data playback time area is proposed. By pre-processing the initial timing data, the number of data points and key points of the target timing data are determined, and the timing data is divided based on the attention mechanism selected by the user and the sliding time window length. For a preset number of time windows, calculate the accumulated value of the attention score of each time window, determine the output sequence, combine multiple time series data key point selection criteria, and introduce a dynamic key point scoring system.
It has realized the most critical time periods accurately screened out in a huge and complex data center, provided more comprehensive and in-depth data insights, improved the ability to identify and present important data intervals, solved the complexity problems in the comprehensive application of multiple methods, and optimized The user's visual experience makes it more intuitive and efficient.
Smart Images

Figure CN118760789B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data analysis, and in particular to a method, device and medium for screening a data playback time zone. Background Art
[0002] When faced with massive amounts of time series big data, it is necessary to form an information summary of the time series data in order to enable users to grasp the core content of the time series data more quickly.
[0003] The current state of technology in the field of IoT time series data analysis presents a variety of screening and visualization methods. When faced with industrial data with high volatility and complexity, screening methods that rely on a single standard may fail because they cannot adapt to the variability and multidimensionality of the data. This results in the existing methods failing to accurately capture all the important features of the data during the identification and screening of key time points.
[0004] In practical applications, combining multiple methods can improve the accuracy and efficiency of key time area screening. However, when multiple methods are used in combination, it brings up the problem of how to display the most valuable information to users within a limited time. Otherwise, when it comes to the display of large-scale time series data, it will face the challenge of data overload, making it difficult for users to quickly capture key information.
[0005] Although some existing products have attempted to highlight key points through dynamic PPT or video effects, these attempts have not fundamentally solved the problem of large data volume and incomplete information display. The display of a combination of multiple methods is often limited to static pictures and tends to be comprehensive, without fully considering the specific needs of users for screening key data points. Summary of the invention
[0006] In order to solve the above problems, the present application proposes a method, device and medium for screening a data playback time zone, wherein the method comprises:
[0007] Preprocess the initial time series data to obtain target time series data, and determine the number of data points corresponding to the target time series data; perform key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; determine the importance scores corresponding to different data playback key points based on the attention mechanism selected by the user; divide the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user; determine the accumulated importance scores corresponding to the preset number of time windows; and determine the output order of the time windows corresponding to the target time series data based on the accumulated importance scores.
[0008] In one example, the preprocessing of the initial time series data to obtain the target time series data specifically includes: verifying whether the timestamp corresponding to the initial time series data conforms to the time series pattern to verify the integrity of the initial time series data; if the initial time series data is incomplete, performing data cleaning on the initial time series data; if the initial time series data is complete, performing volatility detection on the initial time series data to determine the volatility of the initial time series data; and using the initial time series data whose volatility is lower than a preset threshold as the target time series data.
[0009] In one example, the data playback key points include: trend turning points and trend mutation segments; the key point mining of the preprocessed target time series data specifically includes: determining a preset maximum number of line segments, and connecting adjacent time series data points to obtain multiple connected line segments; determining a triangular area composed of two connected line segments before merging and the merged line segments of a group of adjacent line segments as a merging cost value of the group of adjacent line segments; based on the merging cost value, merging the adjacent line segments until the number of line segments is less than the maximum number of line segments to obtain a linear approximate broken line; based on the linear approximate broken line, determining the slope and direction trend of each line segment; taking the line segment with an absolute value of the slope higher than a preset threshold as a trend mutation segment; taking the connection point between two adjacent line segments with different trend directions as a trend turning point.
[0010] In one example, the data playback key points include change points, outlier points and alarm record starting points; the key point mining of the preprocessed target time series data specifically includes: detecting the target time series data through a PELT-based change point detection method with a linear kernel to determine the change point position that minimizes a preset cost function; detecting outlier points in the target time series data through a preset outlier detection threshold; and extracting the alarm record starting point based on a user-customized standard configuration alarm mechanism.
[0011] In one example, the method of dividing the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user specifically includes: determining a candidate starting point in the target time series data based on the length of the sliding time window and the number of data points; selecting any data point from the candidate starting points as the target starting point; determining a target end point of the time window corresponding to the target starting point based on the length of the sliding time window and the time value corresponding to the target starting point; determining a corresponding time window based on the target end point and the target starting point; and traversing all data points in the candidate starting point to divide the time series data into a preset number of time windows.
[0012] In one example, the data playback key points include: a global maximum point, a global minimum point, a time window with maximum variance, and a time window with maximum range; the key point mining of the preprocessed target time series data specifically includes: performing statistical analysis on the target time series data, finding the global maximum point, global minimum point, time window with maximum variance, and time window with maximum range corresponding to the target time series data, and determining the corresponding time points.
[0013] In one example, determining the accumulated importance scores corresponding to the preset number of time windows specifically includes: selecting any one time window among the preset number of time windows as the target time window; determining the type of data playback key points contained in the target time window, and the number of key points corresponding to different types of data playback key points; determining the importance scores corresponding to different types of data playback key points; adding the importance scores corresponding to all data playback key points in the target time window to obtain the accumulated importance scores corresponding to the target time window; and traversing all time windows among the preset number of time windows to determine the accumulated importance scores corresponding to the preset number of time windows.
[0014] In one example, the output order of the time windows corresponding to the target time series data is determined based on the accumulated value of the importance score, specifically including: determining a preset output number of time windows; sorting the preset number of time windows according to the size of the accumulated value of the importance scores; and outputting the time windows corresponding to the target time series data in the sorted order until the output number is equal to the preset output number.
[0015] The present application also provides a device for screening data playback time areas, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute: preprocessing initial time series data to obtain target time series data, and determining the number of data points corresponding to the target time series data; performing key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; determining the importance scores corresponding to different data playback key points based on the attention mechanism selected by the user; dividing the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user; determining the accumulated importance scores corresponding to the preset number of time windows; and determining the output order of the time windows corresponding to the target time series data based on the accumulated importance scores.
[0016] The present application also provides a non-volatile computer storage medium storing computer executable instructions, characterized in that the computer executable instructions are configured to: preprocess initial time series data to obtain target time series data, and determine the number of data points corresponding to the target time series data; perform key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; determine the importance scores corresponding to different data playback key points based on the attention mechanism selected by the user; divide the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user; determine the accumulated importance scores corresponding to the preset number of time windows; and determine the output order of the time windows corresponding to the target time series data based on the accumulated importance scores.
[0017] The method proposed in this application can bring the following beneficial effects: integrating a variety of time series data key point selection criteria and methods, pioneeringly introducing a dynamic key point scoring system based on user observation preferences, and finally accurately screening out the most critical time periods, thereby achieving the goal of information summary. This method deeply mines the multivariate attributes of data to provide users with more comprehensive and in-depth data insights. It not only greatly improves the ability to identify and present important data intervals in large and complex data sets, but also successfully resolves the complexity and time constraints faced when multiple methods are used in combination. The system accumulates the key point scores of different features and accurately screens out the key time intervals with the most display value. With the limitation of the number and length of display areas, the present invention optimizes the user's visualization experience, making it more intuitive and efficient.
[0018] In addition, the algorithm of the present invention has excellent flexibility and applicability, and can be adapted to different application scenarios. The automated and intelligent processing flow greatly reduces manual intervention and improves the intelligent level of operation. This enables users to understand data more intuitively and conveniently, and quickly capture key trends and patterns in massive amounts of information. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0020] Figure 1 A schematic diagram of a flow chart of a method for screening a data playback time zone in an embodiment of the present application;
[0021] Figure 2 This is a schematic diagram of the overall process of a method for screening a data playback time zone in an embodiment of the present application;
[0022] Figure 3 This is a schematic diagram of the structure of a device for screening data playback time zones in an embodiment of the present application. DETAILED DESCRIPTION
[0023] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0024] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.
[0025] Figure 1 A flowchart of a method for screening a data playback time zone provided in one or more embodiments of this specification. The method can be applied to the field of IoT time series data analysis. The process can be executed by a computing device in the corresponding field. Certain input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.
[0026] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail by taking a server as an example.
[0027] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific limitations on this.
[0028] like Figure 1 As shown, the embodiment of the present application provides a method for screening a data playback time zone, including:
[0029] S101: Preprocessing initial time series data to obtain target time series data, and determining the number of data points corresponding to the target time series data.
[0030] First, the input time series data is required to be preprocessed before application. During preprocessing, the integrity of the time series data needs to be verified. If the data is complete, it means that the time series data should be continuous, without missing values or gaps, and the timestamps should be evenly distributed or conform to a specific time series pattern. If there are missing or outliers in the data, data cleaning is required, including interpolation, smoothing, or outlier removal to ensure the quality and consistency of the data set.
[0031] Then, the time series data is tested for volatility. Time series data should show a certain volatility or variability, that is, the data points should not be static or constant on the time axis. Volatility can be evaluated by calculating the statistics of the data (such as variance, standard deviation). Through these preprocessing steps, it can be ensured that the processed time series data is of high quality and suitable for subsequent analysis and feature assignment.
[0032] After data preprocessing, the time series data has a total of N data points. The user needs to set the length of the sliding time window L, which defines the number of time points covered by a window. The value of L will directly affect the time scale of the algorithm analysis and the granularity of the results. The user should also specify the number of time windows W to be selected, which is the number of the most important time intervals that need to be identified and returned during the algorithm execution.
[0033] S102: Perform key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data.
[0034] Here, this application selects 10 types of key information for industrial big data analysis, including change points, trend turning points, trend mutation segments, outlier points, alarm record starting points, discrete points, maximum points, minimum points, maximum variance time windows, and maximum range time windows. The following will introduce how to filter out these 10 types of data points.
[0035] The purpose of screening trend turning points and trend mutation segments is to determine the trend of each time period of time series data and find the trend turning points on the time series data. PLA (Piecewise Linear Approximation) is used, which is an effective method for representing and compressing time series data. It divides the time series into multiple segments, each of which is represented by a straight line approximation, so as to divide and approximate under the standard of optimizing storage and compression data quality. By adding nodes or segments, the PLA method can approximate any nonlinear or linear function with arbitrary precision. When using it, first set the maximum number of segments, then start initialization, connect each pair of adjacent points into a straight line, calculate the cost of merging adjacent segments (connecting the first and last points into a line) (the area enclosed by the fitting segment and the original two lines), and choose the method with the lowest cost to merge. Then iterate until the number of segments is reduced to less than the set maximum number of segments and stop the iteration. By detecting the slope of each segment after the linear approximation of the original data and the time points of the first and last points, it can be determined whether the trend in each time period is rising, falling or stable. It significantly reduces the complexity of trend analysis of time series data by converting nonlinear problems into linear forms.
[0036] The application of trend detection method is explained by taking the output power data of a power plant as an example. The fluctuation of the output power of the power plant closely corresponds to the fluctuation of the power demand of users in its service area. Especially in the peak hours of electricity consumption in the morning and evening, the increase of output power is particularly significant, showing obvious periodic characteristics. Therefore, when conducting trend analysis, we selected the output power data within 24 hours as samples and set the maximum number of line segments to 24 to match the time cycle of one day. The first step of the analysis is to perform time series analysis on the output power data, connecting continuous time points to form multiple initial line segments. Next, we merge these line segments and gradually optimize them through an iterative process until the total number of line segments drops to no more than the preset maximum value, that is, 24 segments, for the first time. When this condition is met, the algorithm stops iterating, and the retained line segments and their corresponding time spans constitute the fitting curve and trend time period of the output power trend. In each trend time period, by analyzing the slope of the line segment, we can accurately determine the specific trend of the output power change, whether it is rising, falling or stable.
[0037] A change point is a point in a time series where the statistical characteristics of the data change significantly. Change point analysis is the process of detecting and identifying these change points. The change point analysis described above uses a change point detection method based on PELT (Pruned Exact Linear Time) using a linear kernel. A linear kernel function is usually defined as the sum of the absolute values of the difference between two data points. K represents the value of the kernel function. In the case of a linear kernel, the kernel function is defined as: .
[0038] First, define the change point analysis data. Read the background data set to be analyzed .
[0039] The second step is to construct a penalty function. The penalty function is used to measure the cost of introducing a new change point. It usually consists of the data fitting cost and the change point introduction cost. Indicates at a point in time The total cost of Indicates the time interval The data fitting cost within represents the cost of introducing the change point, represents the number of change points introduced before the time point. .
[0040] The third step is to calculate the data fitting cost (linear kernel). For the linear kernel, the data fitting cost can be calculated by the least squares method. Assume that in the time interval The data can be modeled using a linear Fitting, where is the observed value. By using the least squares method, we can get the parameter and The estimated value of is the residual sum of squares: .
[0041] The fourth step is to solve the problem by dynamic programming. The idea of dynamic programming is used to solve the optimal combination of change points.
[0042] Defining an array , represents the minimum total cost at a point in time. Initially, For each time point ,calculate Values: , that is, traverse all possible change point positions and select the position that minimizes the total cost.
[0043] Step 5: Pruning strategy. Based on the current optimal cost and penalty function, exclude some parts that cannot be the optimal solution. For example, if at a certain time point t, the current minimum total cost is greater than a preset threshold, then the search after that time point can be stopped.
[0044] Finally, determine the change point. After traversing the entire time series, This is the final minimum total cost. Starting from time point T, we step back through the array , determine the locations of all change points. The specific method is to find, for each time point t, The s that holds true is a change point. Repeat this process until you go back to time point 1.
[0045] Taking the change point analysis of the temperature sensor of a production line as an example, the cost function C based on the linear kernel is set to a large value initially, usually positive infinity. The first data point is used as the initial segmentation point. For each time point, the cost of taking it as a change point is calculated. The cost can be calculated by the linear kernel function to calculate the degree of data change between the current segmentation point and the previous segmentation point. Update the minimum cost and the corresponding segmentation point. After traversing all data points, the final change point segmentation scheme is determined according to the minimum cost. The change point divides the time series data of the temperature sensor on the production line into different intervals, and the data in each interval has relatively stable characteristics. After applying the PELT algorithm, change points may be detected at the 11th data point (the temperature suddenly rises from 23.5°C to 25.0°C) and the 20th data point (the temperature drops from 25.5°C to 23.0°C). These change points may indicate equipment overheating or cooling system problems, which require further inspection and maintenance.
[0046] For abnormal value points and alarm record starting points, the abnormal value points are detected using the abnormal value detection threshold, and the alarm record starting point is extracted according to the user-customized standard configuration alarm mechanism.
[0047] Statistical analysis is used to find the global maximum point, global minimum point, time window with maximum variance, and time window with maximum range, and determine the corresponding time points. For the identification of the time windows with maximum variance and maximum range, this technology follows the user-preset sliding time window parameter L, and gradually slides the time window in chronological order. In each independent time window, the system will calculate the mean, variance, and range, and use this as a basis for comprehensive evaluation and selection.
[0048] S103: Based on the attention mechanism selected by the user, determine the importance scores corresponding to different data playback key points.
[0049] This application provides users with a preference selection and default method based on the key point attention mechanism in the scoring rules. The scoring range and details can be set by the user. For ease of explanation, the scoring value of 0 to 10 is used as an example.
[0050] First, determine the importance that users attach to key points. Based on whether users can clearly express the importance they attach to different key points, three scoring methods are preset for users to choose from.
[0051] The first method is suitable for situations where the user's attention is clear. In this case, the user directly scores different key points, with the value range between 0 and 10. If the user does not pay attention to a certain type of key point, the score is 0. For example, the user attaches great importance to the trend turning point type and assigns a score of 10; the user does not pay much attention to the outlier point type and assigns a score of 3.
[0052] The second method is to provide users with a preference option for situations where the user's importance is unclear. That is, let the user sort the key point types according to their importance, and then assign points from 10 to 1 in the order of selection. The key points that are not selected are assigned a score of 0. For example, the user puts the maximum value type first and assigns a score of 10; the minimum value type is ranked last and assigned a score of 1.
[0053] The third method is to use the default mode for situations where the user has no knowledge of the situation and assign the same score of 1 to all key point types.
[0054] The user finally determined a scoring method for each type of key point, and recorded the user's score for the fth type of key point.
[0055] S104: Based on the number of data points and the sliding time window length set by the user, the time series data is divided into a preset number of time windows.
[0056] When dividing the time series data into multiple time windows, the candidate starting point in the target time series data can be determined based on the length of the sliding time window and the number of data points, and then any data point in the candidate starting point can be selected as the target starting point. Then, based on the length of the sliding time window and the time value corresponding to the target starting point, the target end point of the time window corresponding to the target starting point is determined. At this time, the corresponding time window can be determined based on the target end point and the target starting point. By traversing all the data points in the candidate starting point, the time series data can be divided into a preset number of time windows.
[0057] Specifically, a sliding time window can be used to traverse the entire time series data to form N-L+1 time windows, where the mth ( ) time windows correspond to the time points from the mth to the m+L-1th in the time series data.
[0058] S105: Determine the accumulated values of importance scores corresponding to the preset number of time windows.
[0059] When calculating the cumulative value of the importance score corresponding to each time window, you can select any time window from the preset number of time windows as the target time window. Determine the type of data playback key points contained in the target time window, as well as the number of key points corresponding to different types of data playback key points, and determine the importance scores corresponding to different types of data playback key points. If multiple types of key points are detected at the same time point, in this case, the score of the time point is the accumulation of the scores of each type of key points. At this time, add up the importance scores corresponding to all data playback key points in the target time window to obtain the cumulative value of the importance score corresponding to the target time window; traverse all time windows in the preset number of time windows to determine the cumulative value of the importance score corresponding to the preset number of time windows. Specifically, in the nth ( ) time points, the number of key points of the fth category is , then the accumulated value at this point . Then the mth ( The cumulative score corresponding to each time window is
[0060] S106: Determine the output order of the time windows corresponding to the target time series data based on the accumulated value of the importance score.
[0061] In one embodiment, when outputting the time window, all time windows can be traversed, and the selection process follows the priority logic, first selecting the time window with the largest score accumulation value, and then selecting the second largest time window in the remaining intervals outside the largest time window, and continuing in this way until W non-overlapping intervals are selected, or until all time intervals are considered, and all selected time windows are output. The selected time window provides the user with a data overview, pointing out the most critical part of the data, thereby helping the user to focus on analysis and decision-making.
[0062] like Figure 3 As shown, an embodiment of the present application further provides a device for screening a data playback time zone, comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:
[0063] Preprocess the initial time series data to obtain target time series data, and determine the number of data points corresponding to the target time series data; perform key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; determine the importance scores corresponding to different data playback key points based on the attention mechanism selected by the user; divide the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user; determine the accumulated importance scores corresponding to the preset number of time windows; and determine the output order of the time windows corresponding to the target time series data based on the accumulated importance scores.
[0064] The embodiment of the present application further provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured as follows:
[0065] Preprocess the initial time series data to obtain target time series data, and determine the number of data points corresponding to the target time series data; perform key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; determine the importance scores corresponding to different data playback key points based on the attention mechanism selected by the user; divide the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user; determine the accumulated importance scores corresponding to the preset number of time windows; and determine the output order of the time windows corresponding to the target time series data based on the accumulated importance scores.
[0066] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0067] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.
[0068] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0069] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0070] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0072] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0073] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
[0074] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0075] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0076] The above is only the embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A method for screening a data playback time zone, characterized in that: include: Preprocessing the initial time series data to obtain target time series data, and determining the number of data points corresponding to the target time series data; Performing key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; Based on the attention mechanism selected by the user, determine the importance scores corresponding to different data playback key points; Based on the number of data points and the length of the sliding time window set by the user, the time series data is divided into a preset number of time windows; Determine the accumulated values of importance scores corresponding to the preset number of time windows respectively; Based on the accumulated value of the importance score, an output order of the time windows corresponding to the target time series data is determined.
2. The method according to claim 1, characterized in that The preprocessing of the initial time series data to obtain the target time series data specifically includes: Verifying the integrity of the initial time series data by verifying whether the timestamp corresponding to the initial time series data conforms to the time series mode; If the initial time series data is incomplete, performing data cleaning on the initial time series data; If the initial time series data is complete, performing a volatility test on the initial time series data to determine the volatility of the initial time series data; The initial time series data whose volatility is lower than a preset threshold is used as the target time series data.
3. The method according to claim 1, characterized in that The key points of data playback include: trend turning points and trend mutation segments; The key point mining of the preprocessed target time series data specifically includes: Determine a preset maximum number of line segments, and connect adjacent time series data points to obtain a plurality of connected line segments; Determine the area of a triangle formed by two connected line segments before merging a set of adjacent line segments and the line segments after merging, as the merging cost value of the set of adjacent line segments; Based on the merging cost value, the adjacent line segments are merged until the number of line segments is less than the maximum number of line segments, so as to obtain a linear approximate polyline; Based on the linear approximate broken line, determine the slope and trend of each line segment; The line segment whose absolute value of slope is higher than the preset threshold is regarded as a trend mutation segment; The connecting point between two adjacent line segments with different trends is regarded as the trend turning point.
4. The method according to claim 1, characterized in that: The key points of data playback include change points, abnormal value points and alarm record starting points; The key point mining of the preprocessed target time series data specifically includes: The target time series data is detected by a linear kernel based PELT change point detection method to determine the change point position at which a preset cost function is minimized; Detecting outlier points in the target time series data using a preset outlier detection threshold; Extract the alarm record starting point based on the user-customized standard configuration alarm mechanism.
5. The method according to claim 1, characterized in that The method of dividing the time series data into a preset number of time windows based on the number of data points and the length of the sliding time window set by the user specifically includes: Determine a candidate starting point in the target time series data based on the length of the sliding time window and the number of data points; Select any data point from the candidate starting points as the target starting point; Determine a target end point of the time window corresponding to the target starting point based on the length of the sliding time window and the time value corresponding to the target starting point; Based on the target end point and the target start point, determining a corresponding time window; All data points in the to-be-selected starting point are traversed, and the time series data is divided into a preset number of time windows.
6. The method according to claim 5, characterized in that The key points of data playback include: global maximum point, global minimum point, time window with maximum variance and time window with maximum range; The key point mining of the preprocessed target time series data specifically includes: By performing statistical analysis on the target time series data, the global maximum point, the global minimum point, the time window with the maximum variance and the time window with the maximum range corresponding to the target time series data are found, and the corresponding time points are determined.
7. The method according to claim 1, characterized in that The step of determining the accumulated values of importance scores corresponding to the preset number of time windows respectively includes: Selecting a time window from the preset number of time windows as a target time window; Determine the types of data playback key points contained in the target time window, and the number of key points corresponding to different types of data playback key points; Determine the importance scores corresponding to the key points of different types of data playback; Adding the importance scores corresponding to all the data playback key points in the target time window to obtain the accumulated importance score corresponding to the target time window; All time windows in the preset number of time windows are traversed to determine the accumulated values of importance scores corresponding to the preset number of time windows respectively.
8. The method according to claim 1, characterized in that The step of determining the output order of the time windows corresponding to the target time series data based on the accumulated value of the importance score specifically includes: Determine the preset output quantity for the time window; Sorting the preset number of time windows according to the accumulated values of their importance; The time windows corresponding to the target time series data are output in a sorted order until the output quantity is equal to the preset output quantity.
9. A device for screening a data playback time zone, characterized in that: include: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute: Preprocessing the initial time series data to obtain target time series data, and determining the number of data points corresponding to the target time series data; Performing key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; Based on the attention mechanism selected by the user, determine the importance scores corresponding to different data playback key points; Based on the number of data points and the length of the sliding time window set by the user, the time series data is divided into a preset number of time windows; Determine the accumulated values of importance scores corresponding to the preset number of time windows respectively; Based on the accumulated value of the importance score, an output order of the time windows corresponding to the target time series data is determined.
10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Preprocessing the initial time series data to obtain target time series data, and determining the number of data points corresponding to the target time series data; Performing key point mining on the preprocessed target time series data to obtain data playback key points corresponding to the target time series data; Based on the attention mechanism selected by the user, determine the importance scores corresponding to different data playback key points; Based on the number of data points and the length of the sliding time window set by the user, the time series data is divided into a preset number of time windows; Determine the accumulated values of importance scores corresponding to the preset number of time windows respectively; Based on the accumulated value of the importance score, an output order of the time windows corresponding to the target time series data is determined.
