A time series data processing method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a time series data processing method and device, electronic equipment and storage medium, relates to the technical field of data processing, and particularly relates to the field of artificial intelligence. The specific implementation scheme is: determining input data of a feature processing unit in a processing process of a time series data processing model on target data, the target data being: incremental data of time series data corresponding to a current time unit; obtaining, from a cache, previous data of the feature processing unit, the previous data being: data that the feature processing unit has obtained when performing feature processing in a previous time unit; obtaining a first result of the feature processing unit performing feature processing on the input data and the previous data; updating the first result to the cache; and after obtaining a second result of the time series data processing model processing the target data, updating the second result and the target data to the cache. The scheme provided in the embodiment of the present disclosure can improve the efficiency when processing time series data.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and more particularly to the field of artificial intelligence technology. Background Technology

[0002] Time series data is generated over time, and data from adjacent or closely spaced time periods often exhibit correlation. In some applications, time series data can be processed using time series data processing models. To improve the accuracy of time series data processing, these models can include multiple feature processing units. Due to the aforementioned characteristics of time series data, each feature unit processes the entire dataset at each stage of feature processing. Summary of the Invention

[0003] This disclosure provides a time-series data processing method, apparatus, electronic device, and storage medium.

[0004] According to one aspect of this disclosure, a time-series data processing method is provided, comprising:

[0005] The input data of the feature processing unit is determined during the processing of target data by the time series data processing model. The target data is the incremental data of the time series data corresponding to the current time unit.

[0006] The preceding data of the feature processing unit is obtained from the cache, wherein the preceding data is the data that the feature processing unit obtained when performing feature processing in the previous time unit;

[0007] Obtain the first result of feature processing performed by the feature processing unit on the input data and the preceding data;

[0008] Update the cache with the first result;

[0009] After obtaining the second result of the time-series data processing model processing the target data, the second result and the target data are updated to the cache.

[0010] According to another aspect of this disclosure, a time-series data processing apparatus is provided, comprising:

[0011] The input data determination module is used to determine the input data of the feature processing unit during the processing of target data by the time series data processing model. The target data is the incremental data of the time series data corresponding to the current time unit.

[0012] The preceding data acquisition module is used to obtain the preceding data of the feature processing unit from the cache. The preceding data is the data that the feature processing unit has obtained when performing feature processing in the previous time unit.

[0013] The first result acquisition module is used to obtain the first result of the feature processing unit performing feature processing on the input data and the preceding data;

[0014] The first result update module is used to update the first result to the cache;

[0015] The second result and target data update module is used to update the second result and the target data to the cache after obtaining the second result of the time series data processing model processing the target data.

[0016] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0017] At least one processor; and

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the aforementioned timing data processing method.

[0020] According to another aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the above-described timing data processing method.

[0021] According to another aspect of this disclosure, a computer program product is provided, including a computer program that is executed by a processor using the above-described timing data processing method.

[0022] As can be seen from the above, in the solution provided by the embodiments of this disclosure, during the processing of time-series data in each time unit, the first result of feature processing by the feature processing unit and the second result obtained by the time-series data processing model are cached. This allows the feature processing unit to select the required preceding data from the cached data when it needs preceding data for feature processing in subsequent time units. In this way, each feature processing unit can obtain the preceding data of the target data corresponding to the current time unit from the cache, considering the correlation of the time-series data, without needing to recalculate it. Furthermore, since the target data is incremental data, its volume is smaller than the full data volume, resulting in shorter time required for feature processing. This improves the efficiency of feature processing by the feature processing unit and the processing efficiency of the time-series data processing model.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0024] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0025] Figure 1 This is a schematic diagram of the structure of a timing pipeline model provided in an embodiment of this disclosure;

[0026] Figure 2 This is a schematic diagram of another timing pipeline model provided in this embodiment of the disclosure;

[0027] Figure 3 This is a flowchart illustrating a time-series data processing method provided in an embodiment of this disclosure;

[0028] Figure 4 This is a schematic diagram of input data provided in an embodiment of this disclosure;

[0029] Figure 5 This is a schematic diagram of the input data and output results of a feature processing unit provided in an embodiment of this disclosure;

[0030] Figure 6 This is a schematic diagram of the time-series data processing flow of a time-series pipeline model provided in an embodiment of this disclosure;

[0031] Figure 7 This is a schematic diagram illustrating a process for updating cached data according to an embodiment of this disclosure;

[0032] Figure 8 This is a schematic diagram of the structure of a time-series data processing device provided in an embodiment of this disclosure;

[0033] Figure 9 This is a block diagram of an electronic device used to implement the timing data processing method of the embodiments of this disclosure. Detailed Implementation

[0034] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0035] The application scenarios of the solutions provided in the embodiments of this disclosure are described below.

[0036] The time-series data processing method provided in this disclosure can be applied to scenarios where a time-series data processing model containing multiple feature processing units processes time-series data. For example, the time-series processing model can be a time-series pipeline model. Specifically, the time-series data processing process can be a rolling forecasting process based on the time-series pipeline model, where rolling forecasting can be forecasting power load, network traffic, or other data with time-series characteristics, etc.

[0037] The following will describe the solution provided in the embodiments of this disclosure in conjunction with the rolling prediction using a time-series pipeline model.

[0038] First, the differences between the embodiments of this disclosure and the prior art will be explained by way of example.

[0039] In the prior art, it is assumed that there is a time series data processing model, see [link to previous text]. Figure 1 , Figure 1 A schematic diagram of a time-series pipeline model is provided. This model includes: feature processing unit 1, feature processing unit 2, feature processing unit 3, and a prediction unit. This time-series pipeline model can be used for rolling prediction. During the prediction process, the time-series data of the current time unit first undergoes feature processing by feature processing unit 1. Then, feature processing unit 2 processes the output result 1 of feature processing unit 1. Feature processing unit 3 processes the output result 2 of feature processing unit 2 to obtain output result 3. The prediction unit then uses output result 3 to predict the time-series data, obtaining the prediction result for the current time unit.

[0040] In the process of predicting time series data for the next time unit, since time series data often correlates with time series data corresponding to adjacent or similar time periods, it is necessary to consider the time series data and prediction results of the previous time unit during the prediction process. Therefore, the new time series data, the time series data of the previous time unit, and the prediction results all need to be input into the model. This allows the feature processing unit to perform feature processing more accurately. However, as the time series data prediction process progresses, each data input into the model requires feature processing of the full dataset, including the time series data used before the current time unit, the prediction results, and the new time series data.

[0041] In one embodiment of this disclosure, with Figure 1 The corresponding time-series pipeline model is provided; see [link / reference]. Figure 2 , Figure 2A schematic diagram of another time-series pipeline model is provided. Figure 1 Unlike the provided time-series pipeline model, this embodiment allocates a cache area for feature processing unit 1, feature processing unit 2, feature processing unit 3, and prediction unit. This cache area is used to store the processing results of each feature processing unit during feature processing, as well as the processing results of the time-series pipeline model. Thus, when a feature processing unit needs preceding data for feature processing, it can obtain the corresponding preceding data from the cache. Therefore, it can directly perform feature processing on the newly added incremental data of the time-series data corresponding to the current time unit, i.e., the target data.

[0042] The specific implementation of the solution provided in the embodiments of this disclosure will be described below.

[0043] In one embodiment of this disclosure, see [link to embodiment]. Figure 3 , Figure 3 A flowchart illustrating a time-series data processing method is provided. The method includes the following steps S301-S305.

[0044] Step S301: Determine the input data of the feature processing unit during the processing of the target data by the time series data processing model.

[0045] The target data is the incremental data of the time series data corresponding to the current time unit.

[0046] Time-series data can be segmented into different time units, and the time-series data processing model processes the data sequentially within each time unit. A time unit can be an hour, a day, a week, or a month, etc. Incremental data can be data added based on the time-series data from the previous time unit.

[0047] The various feature processing units included in the time-series data processing model can perform different feature processing operations; that is, each feature processing unit can have different functions. The functions of each feature processing unit can be defined by personnel, and the input data of each feature processing unit can also be defined by personnel. Therefore, the input data of each feature processing unit can be different. The feature processing units that need to determine the input data in the above steps can be all feature processing units in the time-series data processing model, or they can be some of the feature processing units in the time-series data processing model; this disclosure does not limit this.

[0048] For example, the functions of each feature processing unit can be feature extraction, feature transformation, normalization, and so on.

[0049] The specific implementation of step S301 will be explained below, and will not be detailed here.

[0050] Step S302: Obtain the preceding data of the feature processing unit from the cache.

[0051] The preceding data refers to the data obtained by the feature processing unit in the previous time unit when performing feature processing.

[0052] The cache can store data acquired by each feature processing unit during feature processing at each time unit, as well as the processing results of incremental data of the time series data corresponding to each time unit by the time series data processing model. Furthermore, units requiring preceding data for feature processing can be set by personnel, or the preceding data for specific units used for feature processing can be set according to business needs. For example, during the first processing of the target data by the time series data processing model, the time series data with preset results and known results can be used as preceding data.

[0053] In one implementation, at least one of the following data is obtained from the cache as preceding data for the aforementioned feature processing unit: the output result of the adjacent and prior feature processing unit before the previous time unit; the obtained processing result for the time series data, the time series of the obtained processing result being located before the time series of the first result; and the time series data corresponding to the current time unit.

[0054] Corresponding to one implementation of step S301, the preceding data should include data corresponding to the data input of the aforementioned feature processing unit. For example, if the input data includes the output result of an adjacent and earlier feature processing unit, then the preceding data should include the output result of the adjacent and earlier feature processing unit before the previous time unit. If the data input includes the obtained processing result for the time series data, then the preceding data should include the obtained processing result for the time series data, wherein the time series of the obtained processing result is prior to the time series of the first result. If the data input includes the target data, then the preceding data should include the time series data corresponding to the current time unit. The obtained processing result for the time series data and the time series data corresponding to the current time unit can be from the previous time unit, or from some time units before the current time unit, or from all time units before the current time unit.

[0055] As can be seen from the above, the content obtained from the preceding data can correspond to the input data, so that each input data has its own preceding data, which makes it easier for the feature processing unit to perform feature processing based on the correlation of time series data, thereby improving the accuracy of feature processing.

[0056] Step S303: Obtain the first result of feature processing of the input data and preceding data by the above feature processing unit.

[0057] In other words, it obtains the output result of the feature processing unit.

[0058] The specific method by which the feature processing unit processes the input and preceding data depends on the application scenario. For example, in a network traffic prediction scenario, the feature processing unit may perform traffic feature prediction processing on the input and preceding data; in a power load prediction scenario, the feature processing unit may perform power load feature prediction on the input and preceding data.

[0059] Step S304: Update the first result to the cache.

[0060] The first result is updated to the cache to provide preceding data for feature processing units that require preceding data for feature processing. For the last feature processing unit that needs feature processing in the same time unit, updating the first result to the cache can preserve a copy of the processed data required by the time series data processing model.

[0061] In one implementation, if the output result of the aforementioned feature processing unit exists in the cache, the first result is appended to the end of the data of the aforementioned output result; if the aforementioned output result does not exist in the cache, the first result is directly stored in the cache.

[0062] Specifically, the splicing process requires splicing the newly obtained first result with the corresponding existing data. The "corresponding" mentioned above refers to the first result of the specified feature processing unit being spliced ​​into the output result of its own feature processing unit after performing feature processing on the target data of the previous time unit.

[0063] As can be seen from the above, by concatenating the first result to the data tail of the output result of the corresponding feature processing unit, and by uniformly caching the data based on each feature processing unit, it is easier to perform unified data processing and improve the efficiency of data processing.

[0064] Step S305: After obtaining the second result of the above time series data processing model on the target data, update the second result and the target data to the cache.

[0065] In one implementation, the second result is appended to the end of the processed results for the time-series data in the cache; the target data is appended to the end of the time-series data in the cache.

[0066] Once the second result is obtained, the processing of the time series data prior to the current time unit can be considered complete, and the obtained second result is taken as the final processing result for the time series data. Therefore, the second result is appended to the end of the cached data representing the final processing result for the time series data.

[0067] If the time series data before the current time unit has been processed, then the target data is also considered to have been processed. Therefore, the target data is appended to the end of the time series data in the cache as preceding data in the subsequent time series data processing.

[0068] As can be seen from the above, appending the second result to the end of the already processed result data and appending the target data to the end of the time series data in the cache can align the processed time series data and the already processed result data, which facilitates subsequent time series data processing and improves data processing efficiency.

[0069] As can be seen from the above, in the solution provided by the embodiments of this disclosure, during the processing of time-series data in each time unit, the first result of feature processing by the feature processing unit and the second result obtained by the time-series data processing model are cached. This allows the feature processing unit to select the required preceding data from the cached data when it needs preceding data for feature processing in subsequent time units. In this way, each feature processing unit can obtain the preceding data of the target data corresponding to the current time unit from the cache, considering the correlation of the time-series data, without needing to recalculate it. Furthermore, since the target data is incremental data, its volume is smaller than the full data volume, resulting in shorter time required for feature processing. This improves the efficiency of feature processing by the feature processing unit and the processing efficiency of the time-series data processing model.

[0070] Furthermore, since the first and second results are cached in the cache, the feature processing unit can directly use the cached data in subsequent time units when performing feature processing, without having to perform repeated calculations, thus saving computational resources. Also, since repeated calculations consume a large amount of memory resources, the solution provided in this embodiment does not require repeated calculations, thus also saving memory resources.

[0071] The implementation method of step S301 will be explained below.

[0072] In one implementation, for a feature processing unit that is not the first to perform feature processing in a time-series data processing model, the input data of the feature processing unit is determined to include at least one of the following: the output results of adjacent and earlier feature processing units, the processing results obtained for the time-series data, and the target data.

[0073] The first feature processing unit to perform feature processing can be the first feature processing unit in time to begin feature processing among all feature processing units in the time-series data processing model, or it can be a unit that can start feature processing directly without needing the processing results of other feature processing units. Otherwise, it is considered a feature processing unit that is not the first to perform feature processing.

[0074] The following examples illustrate the three types of input data mentioned above.

[0075] 1. Output results of adjacent units that have undergone feature processing earlier.

[0076] For a feature processing unit, the output of the preceding feature processing unit—that is, the output of the feature processing unit adjacent to and preceding this feature processing unit in the data processing flow—is crucial. For example, ... Figure 2 In the time-series data processing model shown, the output of the adjacent and earlier feature processing unit of feature processing unit 2 can be: the output of feature processing unit 1.

[0077] 2. Processing results of time series data

[0078] In one scenario, the obtained processing result for time-series data can be a second result obtained by the time-series data processing model in the previous time unit, where the model processes the target data of that time unit; or it can be any of the second results obtained by the time-series data processing model in previous time units. In another scenario, the obtained processing result can also be a pre-defined, known result for the time-series data.

[0079] 3. Target Data

[0080] As mentioned earlier, the target data is the incremental data processed in the current time unit.

[0081] It is evident that incorporating the outputs of adjacent units that have undergone prior feature processing into the data input allows the current feature processing process to consider other features of the target data, thereby improving the accuracy of feature processing. Incorporating the processed results for time-series data into the data input can serve as label data in the feature processing process, improving the accuracy of feature processing. It also facilitates the synchronous feature extraction of second results that have not undergone feature extraction, unifying the data and making data processing easier. Incorporating the target data into the data input allows for the extraction of more features from the time-series data, improving the accuracy of the time-series data processing model.

[0082] In another implementation, for the first feature processing unit in the time series data processing model, the input data of the feature processing unit is determined to include the target data and the processing results obtained for the time series data.

[0083] The processing results obtained for the time-series data may include a second result obtained for the time-series data and / or a preset result for the time-series data. The aforementioned second result obtained for the time-series data may be the second result obtained in the previous time unit, or it may be some or all of the currently obtained second results.

[0084] In one approach, the input data may further include time-series data corresponding to the previously processed results of the time-series data. This facilitates the feature processing unit in more accurately extracting features from the time-series data. For example... Figure 4 As shown, Figure 4 A schematic diagram of input data is provided. Figure 4 The content within the dashed line on the left represents the processing results of the time series data and the input data of the target data. The content within the dashed line on the right represents the processing results of the time series data, the time series data corresponding to the processing results of the time series data, and the input data of the target data.

[0085] As can be seen from the above, the input data of the first feature processing unit includes the target data and the processing results of the time series data. In this way, feature processing can be performed on the processing results of the time series data at the same time as the incremental data of the time series data corresponding to the current time unit, which can improve the efficiency of feature processing.

[0086] In another implementation, the input data may further include the output results of each unit that performed feature processing earlier. In one approach, the output result of a feature processing unit can be set to include both its own feature processing result and its input data. Then, if the input data of the first unit performing feature processing includes the processed result for the time-series data and the target data, units that are not the first to perform feature processing can set their input data to the output results of adjacent units that performed feature processing earlier. Thus, the second unit performing feature processing can receive the processed result for the time-series data, the target data, and the feature processing result of the first unit. This process continues, with each subsequent unit performing feature processing retaining the output results of each unit that performed feature processing earlier. For example, see [link to relevant documentation]. Figure 5 , Figure 5A schematic diagram of the input data and output results of a feature processing unit is provided. First, the output results of feature processing unit 1 and feature processing unit 2 are set to include both the feature processing result of the unit itself and the input data of the unit itself. The input data of feature processing unit 2 and feature processing unit 3 are set to the output results of adjacent units that are processed first. Therefore, the input data of feature processing unit 1 consists of the processed result for time-series data and the target data, and the output result is the processed result, the target data, and output result 1 of feature processing unit 1. The input data of feature processing unit 2 consists of the processed result, the target data, and output result 1, and the output result 2 of feature processing unit 2 consists of the processed result, the target data, output result 1, and output result 2 of feature processing unit 2. The input data of feature processing unit 3 consists of the processed result, the target data, output result 1, and output result 2. Since feature processing unit 3 uses a different feature processing method, its output result can be set independently. Thus, before feature processing unit 3 processes its own input data, output result 1 and output result 2 are retained.

[0087] In one embodiment of this disclosure, the time-series data processing model is taken as the aforementioned time-series pipeline model as an example, combined with... Figure 6 The solutions provided in the embodiments of this disclosure will be described.

[0088] See Figure 6 , Figure 6 A flowchart illustrating the time-series data processing using a time-series pipeline model is provided. This time-series pipeline model is used for prediction of time-series data.

[0089] The aforementioned time-series pipeline model includes: feature processing unit 1, feature processing unit 2, feature processing unit 3, and prediction unit.

[0090] Once new time-series data is obtained, the time-series data processing process can begin. This new time-series data can be a single time unit or a long period of time. The long period of time-series data is then divided into multiple time units, and the data from the first time unit is processed sequentially. For clarity, the following explanation uses a day as an example.

[0091] Suppose we have a set of processed time-series data and its corresponding time-series data, and this data is stored in a cache. Then, we can use the last time of the time-series data as the prediction start time. If we obtain a segment of time-series data after the prediction start time, we can first segment the time-series data to be predicted into one day's worth of time-series data, which will be used as the incremental data for the corresponding time-series data on the first day, i.e., the target data.

[0092] 1. We will now begin time-series data processing on the target data for the first day.

[0093] The target data, the processing results of the time series data, and the time series data corresponding to the processing results of the time series data are determined as the input data of feature processing unit 1. The preceding data of the target data on the first day is the time series data corresponding to the processing results of the time series data. Since the preceding data of the target data on the first day has not undergone feature processing, it can also be used as the input data of feature processing unit 1.

[0094] After feature processing by feature processing unit 1, output result 1 is obtained. Feature processing unit 2 then performs feature processing on output result 1. Since output result 1 contains the feature processing results of feature processing unit 1 for time series data prior to the current time unit, feature processing unit 2 can use the feature processing results in output result 1 with a time series prior to the first day as preceding data. Therefore, feature processing unit 2 performs feature processing on output result 1 to obtain output result 2.

[0095] Similarly, feature processing unit 3 can use the feature processing results in output result 2 that are prior to the first day as preceding data. Feature processing unit 3 then normalizes output result 2 to obtain output result 3.

[0096] Since the processing results of feature processing unit 1, feature processing unit 2 and feature processing unit 3 are not stored in the cache, output result 1, output result 2 and output result 3 are directly stored in the cache.

[0097] The prediction unit processes output result 3 to predict the target data for day 1, which becomes the second result. Then, the prediction result and the target data are used to update the cached processed results for the time-series data and their corresponding time-series data.

[0098] 2. After completing the time-series data processing of the target data for the first day, begin the time-series data processing of the target data for the second day.

[0099] At this point, the processing results of the time series data and its corresponding time series data already contain the prediction results of the time series data processing of the target data for the first day and the target data for the first day.

[0100] Since the prediction results of the target data on the first day have not undergone feature processing, the target data on the second day and the already processed results of the target data on the first day can be used as the input data of feature processing unit 1.

[0101] Feature processing unit 1 performs feature processing to obtain output result 1, which is then processed by feature processing unit 2. Since output result 1 no longer contains the feature processing results of feature processing unit 1 for time series data prior to the current time unit, feature processing unit 2 determines output result 1 as input data, retrieves output result 1 obtained from the cache for feature processing of the target data of the first day as preceding data, performs feature processing on it, and obtains output result 2.

[0102] Similarly, feature processing unit 3 can determine that the output result 2 is the input data, obtain the output result 2 obtained by feature processing of the target data of the first day from the cache as the preceding data, and obtain the output result 3 after feature processing.

[0103] At this point, the cache already contains output results 1, 2, and 3 obtained from feature processing of the target data from day one. Therefore, output results 1, 2, and 3 obtained from feature processing of the target data from day two can be updated in the cache respectively. For example, see... Figure 7 , Figure 7 A schematic diagram illustrating the process of updating cached data is provided. For example... Figure 7 As shown, the output result 1 obtained by feature processing of the target data for the second day is updated to the end of the data in the cached output result 1, and each data corresponds one-to-one. Similarly, output result 2 and output result 3 are concatenated in the same way.

[0104] The cache unit predicts the concatenated data result 3 in the current cache to obtain a prediction result for the target data of the next day, which serves as the second result for the next day. Then, the prediction result and the target data are used to update the cached processing results for the time-series data and their corresponding time-series data.

[0105] 3. After completing the time-series data processing for the target data of the second day, begin the time-series data processing for the target data of the third day.

[0106] At this point, the processing results of the time-series data, as well as the prediction results of the time-series data that already includes the target data before the current time unit, are considered. Similarly, the target data for the third day is processed in a manner similar to the time-series processing of the target data on the second day.

[0107] Once all unpredicted time series data has been processed by the time series pipeline model, data processing can be stopped, or we can wait for new time series data. If new time series data is obtained, it can be processed in the same way described above.

[0108] In one embodiment of this disclosure, see [link to embodiment]. Figure 8 , Figure 8 A schematic diagram of a time-series data processing device is provided. The device includes:

[0109] The input data determination module 801 is used to determine the input data of the feature processing unit during the processing of target data by the time series data processing model. The target data is the incremental data of the time series data corresponding to the current time unit.

[0110] The preceding data acquisition module 802 is used to obtain the preceding data of the feature processing unit from the cache. The preceding data is the data that the feature processing unit has obtained when performing feature processing in the previous time unit.

[0111] The first result acquisition module 803 is used to obtain the first result of the feature processing unit performing feature processing on the input data and the preceding data;

[0112] The first result update module 804 is used to update the first result to the cache;

[0113] The second result and target data update module 805 is used to update the second result and the target data to the cache after obtaining the second result of the time series data processing model processing the target data.

[0114] As can be seen from the above, in the solution provided by the embodiments of this disclosure, during the processing of time-series data in each time unit, the first result of feature processing by the feature processing unit and the second result obtained by the time-series data processing model are cached. This allows the feature processing unit to select the required preceding data from the cached data when it needs preceding data for feature processing in subsequent time units. In this way, each feature processing unit can obtain the preceding data of the target data corresponding to the current time unit from the cache, considering the correlation of the time-series data, without needing to recalculate it. Furthermore, since the target data is incremental data, its volume is smaller than the full data volume, resulting in shorter time required for feature processing. This improves the efficiency of feature processing by the feature processing unit and the processing efficiency of the time-series data processing model.

[0115] In one embodiment of this disclosure, the input data determination module is specifically used to: determine, for a feature processing unit that is not the first to perform feature processing in the time-series data processing model, that the input data of the feature processing unit includes at least one of the following data:

[0116] The output of adjacent units that have undergone feature processing first;

[0117] The first result obtained from the time-series data;

[0118] The target data.

[0119] It is evident that incorporating the outputs of adjacent units that have undergone prior feature processing into the data input allows the current feature processing process to consider other features of the target data, thereby improving the accuracy of feature processing. Incorporating the processed results for time-series data into the data input can serve as label data in the feature processing process, improving the accuracy of feature processing. It also facilitates the synchronous feature extraction of second results that have not undergone feature extraction, unifying the data and making data processing easier. Incorporating the target data into the data input allows for the extraction of more features from the time-series data, improving the accuracy of the time-series data processing model.

[0120] In one embodiment of this disclosure, the preceding data acquisition module is specifically configured to: obtain at least one of the following data from the cache as preceding data for the feature processing unit:

[0121] The output of the adjacent and earlier feature-processed unit before the previous time unit;

[0122] Regarding the second result obtained from the time-series data, the time sequence of the second result precedes the time sequence of the first result;

[0123] The time series data preceding the current time unit.

[0124] As can be seen from the above, the content obtained from the preceding data can correspond to the input data, so that each input data has its own preceding data, which makes it easier for the feature processing unit to perform feature processing based on the correlation of time series data, thereby improving the accuracy of feature processing.

[0125] In one embodiment of this disclosure, the input data determination module is specifically used to: determine, for the first feature processing unit performing feature processing in the time series data processing model, that the input data of the feature processing unit includes target data and the obtained processing results for the time series data.

[0126] As can be seen from the above, the input data of the first feature processing unit includes the target data and the processing results of the time series data. In this way, feature processing can be performed on the processing results of the time series data at the same time as the incremental data of the time series data corresponding to the current time unit, which can improve the efficiency of feature processing.

[0127] In one embodiment of this disclosure, the first result update module is specifically used for: if the output result of the feature processing unit exists in the cache, appending the first result to the end of the data of the output result; if the output result does not exist in the cache, directly storing the first result in the cache.

[0128] As can be seen from the above, by concatenating the first result to the data tail of the output result of the corresponding feature processing unit, and by uniformly caching the data based on each feature processing unit, it is easier to perform unified data processing and improve the efficiency of data processing.

[0129] In one embodiment of this disclosure, the second result and target data update module is specifically used to: append the second result to the end of the data of the obtained processing result for the time-series data in the cache; and append the target data to the end of the data of the time-series data in the cache.

[0130] As can be seen from the above, appending the second result to the end of the already processed result data and appending the target data to the end of the time series data in the cache can align the processed time series data and the already processed result data, which facilitates subsequent time series data processing and improves data processing efficiency.

[0131] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0132] In one embodiment of this disclosure, an electronic device is provided, comprising:

[0133] At least one processor; and

[0134] A memory communicatively connected to the at least one processor; wherein,

[0135] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the aforementioned timing data processing method.

[0136] In one embodiment of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the above-described timing data processing method.

[0137] In one embodiment of this disclosure, a computer program product is provided, including a computer program that is executed by a processor using the aforementioned timing data processing method.

[0138] Figure 9 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0139] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 902 or a computer program loaded from storage unit 908 into random access memory (RAM) 903. RAM 903 may also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0140] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0141] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as method timing data processing. For example, in some embodiments, method timing data processing may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the method timing data processing described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to perform method timing data processing by any other suitable means (e.g., by means of firmware).

[0142] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0143] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0144] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0145] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0146] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0147] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0148] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0149] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A time-series data processing method, comprising: The input data of the feature processing unit during the processing of target data by the time series data processing model is determined. The target data is the incremental data of the time series data corresponding to the current time unit. The time series data processing model includes a feature processing unit and a prediction unit. The time series data processing model is a pipeline model. The input data of the first feature processing unit includes the processing results of the time series data and the target data. The input data of non-first feature processing units includes the output results of adjacent and earlier feature processing units. The preceding data of the feature processing unit is obtained from the cache. The preceding data is the data obtained by the feature processing unit when performing feature processing in the previous time unit. If the input data includes the output result of the adjacent and earlier feature processing unit, then the preceding data includes the output result of the adjacent and earlier feature processing unit before the previous time unit. If the input data includes the obtained processing result for the time series data, then the preceding data includes the obtained processing result for the time series data, and the time series of the obtained processing result is located before the time series of the first result. If the input data includes the target data, then the preceding data includes the time series data corresponding to the current time unit. A first result is obtained from the feature processing unit performing feature processing on the input data and the preceding data, wherein the first result is the output result of the feature processing unit. Update the cache with the first result; After obtaining the second result of the time series data processing model processing the target data, the second result and the target data are updated to the cache. The second result is the prediction result output by the prediction unit based on the output result of the feature processing unit.

2. The method according to claim 1, wherein, The input data of the feature processing unit during the processing of target data by the deterministic time-series data processing model includes: For a feature processing unit that is not the first to perform feature processing in a time-series data processing model, the input data of the feature processing unit further includes at least one of the following: The processing results of the time series data; The target data.

3. The method according to claim 1 or 2, wherein, Updating the first result to the cache includes: If the output result of the feature processing unit exists in the cache, the first result is appended to the end of the data of the output result; If the output result is not present in the cache, the first result is stored directly in the cache.

4. The method according to claim 1 or 2, wherein, The step of updating the cache with the second result and the target data includes: The second result is appended to the end of the already processed results for time-series data in the cache; The target data is appended to the end of the time-series data in the cache.

5. A time-series data processing apparatus, comprising: The input data determination module is used to determine the input data of the feature processing unit during the processing of target data by the time series data processing model. The target data is the incremental data of the time series data corresponding to the current time unit. The time series data processing model includes a feature processing unit and a prediction unit. The time series data processing model is a pipeline model. The input data of the first feature processing unit includes the processing results of the time series data and the target data. The input data of non-first feature processing units includes the output results of adjacent and earlier feature processing units. A preceding data acquisition module is used to obtain preceding data of the feature processing unit from a cache. The preceding data is data obtained by the feature processing unit during feature processing in the previous time unit. If the input data includes the output results of adjacent and earlier feature processing units, then the preceding data includes the output results of adjacent and earlier feature processing units before the previous time unit. If the input data includes the obtained processing results for time-series data, then the preceding data includes the obtained processing results for time-series data, where the time sequence of the obtained processing results precedes the time sequence of the first result. If the input data includes target data, then the preceding data includes the time-series data corresponding to the current time unit. The first result acquisition module is used to obtain a first result of the feature processing unit performing feature processing on the input data and the preceding data, wherein the first result is the output result of the feature processing unit. The first result update module is used to update the first result to the cache; The second result and target data update module is used to update the second result and the target data to the cache after obtaining the second result of the time series data processing model processing the target data. The second result is the prediction result output by the prediction unit based on the output result of the feature processing unit.

6. The apparatus according to claim 5, wherein, The input data determination module is specifically used to: for feature processing units that are not the first to perform feature processing in the time series data processing model, determine that the input data of the feature processing unit further includes at least one of the following data: The processing results of the time series data; The target data.

7. The apparatus according to claim 5 or 6, wherein, The first result update module is specifically used for: if the output result of the feature processing unit exists in the cache, appending the first result to the end of the output result; if the output result does not exist in the cache, storing the first result directly in the cache.

8. The apparatus according to claim 5 or 6, wherein, The second result and target data update module is specifically used for: appending the second result to the end of the data of the obtained processing result for the time series data in the cache; and appending the target data to the end of the data of the time series data in the cache.

9. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-4.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-4.

11. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-4.

Images