Construction method and device of time sequence analysis system, equipment and medium

Abstract

The embodiment of the invention provides a construction method and device of a time sequence analysis system, equipment and a medium, and is applied to the field of artificial intelligence. The to-be-constructed time sequence analysis system comprises a plurality of analysis subsystems, and a to-be-trained time sequence prediction model is stored in each analysis subsystem and used for predicting a time sequence corresponding to the analysis subsystem. The method comprises the following steps: acquiring an original time sequence corresponding to each analysis subsystem; inputting the original time sequence into a time sequence prediction model of the corresponding analysis subsystem, and decomposing the original time sequence by the time sequence prediction model to obtain global variable information; global variable information of each analysis subsystem is obtained, and federated variable information is generated according to the multiple pieces of global variable information; and optimizing each time sequence prediction model according to the federal variable information and the global variable information corresponding to each analysis subsystem, completing training of the plurality of time sequence prediction models, and obtaining a constructed time sequence analysis system.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method, device, equipment and medium for constructing a time series analysis system. Background Art

[0002] Time series are data points arranged in chronological order, usually at consecutive time intervals, such as daily, monthly, and annually. This type of data is common in many fields, such as finance, meteorology, stock markets, sales data, etc. In the financial system, forecasting, classification, and anomaly detection of time series can play a predictive and auxiliary role in risk control, sales, and management.

[0003] However, existing time series-based processing uses high-quality data collected by relevant institutions without missing values. In addition, different clients in real scenarios have their own data. For the sake of data security and privacy, data will not be collected in a centralized organization or retrograde training. At the same time, when processing time series, only the time series itself is processed, resulting in inaccurate analysis results of the time series, and the role it plays in the operation of the financial system is relatively limited. Summary of the invention

[0004] The present application discloses a construction method, device, equipment and medium for a time series analysis system, which aims to solve the problem that existing time series-based processing uses high-quality data collected by relevant institutions without missing values. In addition, different clients in real scenarios have their own data. For the sake of data security and privacy, the data will not be collected in a centralized organization and retrograde training. At the same time, when processing the time series, only the time series itself is processed, resulting in the final analysis results of the time series being inaccurate and the role played in the operation of the financial system being relatively limited.

[0005] In a first aspect, the present application provides a method for constructing a time series analysis system, wherein the time series analysis system to be constructed includes multiple analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem; the method includes:

[0006] Get the original time series corresponding to each analysis subsystem;

[0007] The original time series is input into the time series prediction model of the corresponding analysis subsystem. The time series prediction model decomposes the original time series to obtain global variable information.

[0008] Obtain the global variable information of each analysis subsystem, and generate federated variable information based on multiple global variable information;

[0009] Each time series prediction model is optimized according to the federated variable information and the global variable information corresponding to each analysis subsystem, the training of multiple time series prediction models is completed, and the constructed time series analysis system is obtained.

[0010] In some embodiments, the time series prediction model includes a decomposition layer, and the original time series is input into the time series prediction model of the corresponding analysis subsystem, including:

[0011] The original time series is input into the decomposition layer of each analysis subsystem, and the decomposition layer decomposes the original time series to obtain global variable information.

[0012] Exemplarily, the decomposition layer decomposes the original time series and obtains the first period information and the first trend information; the time series prediction model also includes a coding layer and a decoding layer; the method also includes: inputting the original time series into the coding layer, the coding layer encodes the original time series, and obtains the second period information and the second trend information; inputting the first period information into the decoding layer, the decoding layer decodes the first period information, and obtains the third period information; obtains the period difference information between the third period information and the first period information, and the trend difference information between the first trend information and the second trend information; according to the period difference information and the trend difference information, the decomposition layer is optimized and the global variable information is updated, so that the first period information approaches the third period information, and the first trend information approaches the second trend information.

[0013] It should be noted that, in some embodiments, the coding layer includes at least one periodic trend decomposition module; the original time series is input into the coding layer, and the coding layer encodes the original time series to obtain second period information and second trend information, including: inputting the original time series into the first periodic trend decomposition module to obtain sub-period information and second trend information; transmitting the sub-period information to the remaining periodic trend decomposition modules layer by layer, and outputting the second period information.

[0014] It should be noted that, in some embodiments, the time series prediction model also includes an attention layer; before obtaining the period difference information between the third period information and the first period information, the first trend information and the trend difference information of the second trend information, it also includes: inputting the second period information and the third period information into the attention layer, the attention layer performs frequency domain correlation learning on the second period information and the third period information to obtain frequency domain correlation information; updating the parameters of the decoding layer according to the frequency domain correlation information to update the third period information.

[0015] In some embodiments, each time series prediction model is optimized based on the federated variable information and the global variable information corresponding to each analysis subsystem, and also includes: obtaining the correlation information between the federated variable information and the global variable information corresponding to each analysis subsystem based on the self-attention mechanism; updating the global variable information corresponding to the analysis subsystem according to the correlation information, so as to optimize the parameters of the time series prediction model corresponding to the analysis subsystem according to the global variable information.

[0016] In some embodiments, obtaining the original time series corresponding to each analysis subsystem includes: obtaining multiple original time series collected by multiple analysis subsystems; constructing a local data set corresponding to each analysis subsystem based on the original time series corresponding to each analysis subsystem; and obtaining the original time series with the same time axis from the local data set corresponding to each analysis subsystem.

[0017] In a second aspect, the present application provides a device for constructing a time series analysis system, wherein the time series analysis system to be constructed includes a plurality of analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem; including:

[0018] A sequence acquisition unit, used to acquire the original time series corresponding to each analysis subsystem;

[0019] The variable acquisition unit is used to input the original time series into the time series prediction model of the corresponding analysis subsystem, and the time series prediction model decomposes the original time series to obtain global variable information;

[0020] A federation generation unit, used for obtaining global variable information of each analysis subsystem, and generating federation variable information according to a plurality of global variable information;

[0021] The construction completion unit is used to optimize each time series prediction model according to the federated variable information and the global variable information corresponding to each analysis subsystem, complete the training of multiple time series prediction models, and obtain the constructed time series analysis system.

[0022] In a third aspect, the present application provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein the memory stores a strategy model, wherein when the computer program is executed by the processor, a method for constructing a time series analysis system as provided in any embodiment of the present application is implemented.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the processor implements a method for constructing a time series analysis system as provided in any embodiment of the present application.

[0024] The embodiments of the present application provide a method, device, equipment and medium for constructing a time series analysis system. The time series analysis system to be constructed includes multiple analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem. The provided method obtains the original time series corresponding to each analysis subsystem and inputs it into the time series prediction model of each analysis subsystem. The time series prediction model decomposes the original time series, obtains global variable information, and then generates federal variable information based on multiple global variable information, so as to optimize each time series prediction model based on the federal variable information and the global variable information corresponding to each analysis subsystem, complete the training of multiple time series prediction models, and obtain the constructed time series analysis system.

[0025] Furthermore, the provided method can be extended to the existing encoding-decoding architecture, has good scalability, and only uploads global variable information, does not involve periodic and trend information, and has better data privacy and security. And based on the federated method, multiple participants can jointly participate in the training of the model while considering the global trend, which can better utilize the data resources and knowledge experience of all parties. In summary, the system constructed by the learning method proposed in this application is scalable, has better privacy and security, and can learn more accurate hidden variables of time series data.

[0026] When applied in financial systems, the provided method can combine multiple sub-suppliers or data managers in multiple regions to conduct federated learning, so that the trained time series analysis system can better predict, classify and detect anomalies of time series in the financial system, and play a predictive and auxiliary role in risk control, sales, management and other aspects.

[0027] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0029] Figure 1 is a schematic flow chart of the steps of a method for constructing a time series analysis system provided in one embodiment of the present application;

[0030] Figure 2 It is a schematic flow chart of the steps of a method for obtaining an original time series provided in an embodiment of the present application;

[0031] Figure 3 It is a schematic flow chart of the steps of a parameter optimization method of a time series prediction model provided by an embodiment of the present application;

[0032] Figure 4 It is a schematic diagram of the structure of a device for constructing a time series analysis system provided in one embodiment of the present application;

[0033] Figure 5 It is a schematic block diagram of the structure of a computer device provided in one embodiment of the present application.

[0034] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are 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 creative work are within the scope of protection of this application.

[0036] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0037] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0038] It should be understood that, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first", "second", etc. are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not necessarily limit the difference.

[0039] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0040] To facilitate understanding of the embodiments of the present application, some words involved in the embodiments of the present application are briefly explained below.

[0041] 1. Federated learning: Federated learning is also known as federated machine learning (Federated machine learning / Federated Learning), joint learning, and alliance learning. Federated learning is a machine learning framework that can effectively help multiple institutions use data and conduct machine learning modeling while meeting the requirements of user privacy protection, data security, and government regulations. As a distributed machine learning paradigm, federated learning can effectively solve the problem of data silos, allowing participants to jointly model without sharing data, and can technically break data silos and achieve AI collaboration.

[0042] Federated learning defines a machine learning framework, under which virtual models are designed to solve the problem of different data owners collaborating without exchanging data. The virtual model is the optimal model for all parties to aggregate data together, and each region serves local goals based on the model. Federated learning requires that the modeling result should be infinitely close to the traditional model, that is, the result of aggregating data from multiple data owners in one place for modeling. Under the federation mechanism, the identity and status of each participant are the same, and a shared data strategy can be established. Since the data is not transferred, it will not leak user privacy or affect data specifications. In order to protect data privacy and meet legal and compliance requirements.

[0043] Federated learning has three major components: data source, federated learning system, and user. The relationship between the three is shown in the figure. Under the federated learning system, each data source performs data preprocessing, jointly builds its learning model, and feeds back the output results to the user.

[0044] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0045] A time series task is a task that involves modeling, forecasting, and analyzing time series data (i.e., the original time series in this application). A time series is a set of data points that are arranged in chronological order, usually in continuous time intervals, such as daily, monthly, and yearly. This type of data is common in many fields, such as finance, meteorology, stock markets, sales data, and so on.

[0046] Here are some key aspects of time series tasks:

[0047] 1. Time Series Forecasting: This is the most common time series task. The goal is to predict future values ​​based on past data points. This forecast can be a single-step forecast (forecasting the next data point) or a multi-step forecast (forecasting multiple data points in the future).

[0048] 2. Anomaly detection: In a time series, there may be abnormal or abrupt data points, which may indicate failures, unusual events, or other important information. The task of anomaly detection is to detect and identify these anomalies.

[0049] 3. Time series classification: In some cases, time series data needs to be classified into different categories. For example, stock price increases and decreases can be classified as increases or decreases based on the patterns and characteristics of the time series.

[0050] 4. Seasonal analysis: Time series data may have seasonality, i.e., recurring patterns at specific time intervals. Seasonal analysis aims to identify these seasonal patterns.

[0051] 5. Rolling Forecast: As time goes by, the model can be continuously updated and rolling forecasts can be performed. This is a dynamic time series forecasting task.

[0052] 6. Time series generation: Based on historical data, new time series samples can be generated, which is useful in some synthetic data applications.

[0053] 7. Time series correlation analysis: Time series data may be associated with other data series or events. Time series correlation analysis helps discover these associations.

[0054] In the problem of time series prediction, the current time series model is basically based on the Transformer method Informer model (a new transformer model for long-series time series prediction). The Transformer-based method significantly improves the results of time series prediction.

[0055] At the same time, a novel decomposition architecture Autoformer is designed in the prior art, which has an automatic correlation mechanism, breaks the tradition of preprocessing of series decomposition, and transforms it into the basic internal block of the deep model. This design gives Autoformer the ability to progressively decompose complex time series. And by combining Transformer with seasonal trend decomposition method, the decomposition method captures the global characteristics of the time series, while Transformer captures more detailed structures.

[0056] The above methods all adopt the idea of ​​decomposition, but there are the following problems:

[0057] 1. Failure to consider real data: The data sets used in the article are all high-quality data collected by relevant institutions, with no missing values.

[0058] 2. No consideration of real-world scenarios: In real-world scenarios, different clients have their own data. For data security and privacy reasons, data will not be collected in a centralized organization or retrograde training.

[0059] 3. The decomposition only results in seasonal trends, but lacks other hidden variables: Taking the pork price curve in reality as an example, in addition to the seasonal trend, it is also affected by the consumer price index (CPI) and the prices of other meats and vegetables.

[0060] In response to the above problems and considerations for data security and privacy, this application proposes to use federated learning to build a time series analysis system to achieve better and more accurate learning of hidden variables in time series data. This is a system construction for time series learning that can solve the problems of multi-party participation, data privacy and security, which can be called Fed-former. This paper hopes to propose a unified time series learning system to realize the federation of transformer-based methods.

[0061] See also Figure 1 , Figure 1 The following is a flowchart showing the steps of a method for constructing a time series analysis system provided in an embodiment of the present application. The time series analysis system to be constructed includes multiple analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem.

[0062] It should be noted that the construction method of the time series analysis system provided in this application can be used in the financial system, which has a large amount of data, strong timeliness, and is distributed with multiple different data suppliers (such as financial systems and multiple banks) and data managers in multiple regions (such as corresponding subsidiaries in different regions, etc.). Through the construction of the time series analysis system, it is possible to combine multiple different subsystems to predict, classify, and detect anomalies in time series, which can play a predictive and auxiliary role in risk control, sales, management, etc. It can also be applied to any other system with a large amount of timeliness and multiple subsystems, such as medical, education and other fields. This application is explained using financial scenarios as an example.

[0063] like Figure 1 As shown, the proposed method for constructing a time series analysis system includes steps S101 to S104.

[0064] S101. Obtain the original time series corresponding to each analysis subsystem.

[0065] Specifically, when the method provided in the present application is applied in a financial system, the multiple analysis subsystems can be multiple different types of data suppliers in the financial system, such as multiple banks as analysis subsystems. It can also be data suppliers corresponding to multiple regions in the financial system, such as data suppliers in different provinces and cities. The present application does not limit the type and number of analysis subsystems. The present application aims to use the method of federated learning to jointly complete the construction of a time series analysis system without leaking data privacy by bringing together multiple data suppliers that were originally not interoperable. By obtaining the time series data generated during the operation of each analysis subsystem, that is, the original time series, and processing it, it is possible to complete the training of each analysis subsystem without leaking information.

[0066] In some embodiments, please refer to Figure 2 , Figure 2 It is a schematic flowchart of the steps of a method for obtaining an original time series provided in an embodiment of the present application.

[0067] like Figure 2 As shown, the provided method includes steps S101a to S101c.

[0068] Step S101 a. Acquire multiple original time series collected by multiple analysis subsystems.

[0069] Step S101 b. Construct a local data set corresponding to each analysis subsystem according to the original time series corresponding to each analysis subsystem.

[0070] Step S101 c. Acquire the original time series with the same time axis from the local data set corresponding to each analysis subsystem.

[0071] Because if it is applied in a financial system, each analysis subsystem will generate data at every moment during its operation. In order to ensure the standardized management of the data of each subsystem, the present application constructs a local data set corresponding to each analysis subsystem according to the original time series corresponding to each analysis subsystem, and then obtains the original time series with the same time axis through each local data set for the construction method provided by the present application. This can ensure the time consistency of the training data of multiple systems and improve the training effect. At the same time, due to the differences in the data acquisition devices of each subsystem, there is no restriction on the sampling frequency of the data, so the present application only requires the same time axis, that is, the consistency of the start time and the end time.

[0072] S102. Input the original time series into the time series prediction model of the corresponding analysis subsystem. The time series prediction model decomposes the original time series to obtain global variable information.

[0073] Specifically, by inputting the original time series of each analysis subsystem into the time series prediction model corresponding to each subsystem, the time series prediction model can decompose the input original time series, and then obtain the global variable information corresponding to the original time series. The provided method can extract the training samples of federated learning without leaking data privacy by only extracting the corresponding global variable information in the original time series.

[0074] In some embodiments, the time series prediction model includes a decomposition layer, and the original time series is input into the time series prediction model of the corresponding analysis subsystem, including: inputting the original time series into the decomposition layer of each analysis subsystem, the decomposition layer decomposing the original time series, and obtaining global variable information.

[0075] By setting a decomposition layer in the time series prediction model, such as setting a Feder layer, the Feder layer includes a MoE (Mixture of Experts) decomposition module. The original time series can be decomposed to obtain the corresponding global variable G, and then the global variable G of each analysis subsystem can be obtained to perform federated learning on multiple analysis subsystems.

[0076] Exemplarily, the decomposition layer decomposes the original time series and obtains the first period information and the first trend information (such as the first period information S1 and the first trend information T1); the time series prediction model also includes an encoding layer and a decoding layer (such as an encoder layer and a decoder layer); the method also includes: inputting the original time series into the encoding layer, the encoding layer encodes the original time series, and obtains the second period information and the second trend information (such as the second period information S2 and the second trend information T2); inputting the first period information into the decoding layer, the decoding layer decodes the first period information, and obtains the third period information (such as the third period information S3); obtaining the period difference information between the third period information and the first period information, and the trend difference information between the first trend information and the second trend information; according to the period difference information and the trend difference information, the decomposition layer is optimized and the global variable information is updated, so that the first period information approaches the third period information, and the first trend information approaches the second trend information.

[0077] Furthermore, the present application can extract the third period information S3 and the second trend information T2 through the Encoder-Decoder part by sampling the framework of the time series analysis system composed of Encoder-Decoder (encoding layer and decoding layer) and Feder (decomposition layer). And calculate the difference between the first period information S1 and the first trend information T1 extracted by the decomposition layer, and then optimize the decomposition layer part, so that the period and trend information output by the decomposition layer part can be close to the output of the encoding layer and the decoding layer.

[0078] At the same time, the provided method can also reversely update the model parameters of Encoder, Decoder, and Feder according to the direct difference between the local true value and the predicted value of the trend information and periodic information corresponding to the original time series.

[0079] It should be noted that, in some embodiments, the coding layer includes at least one periodic trend decomposition module; the original time series is input into the coding layer, and the coding layer encodes the original time series to obtain second period information and second trend information, including: inputting the original time series into the first periodic trend decomposition module to obtain sub-period information and second trend information; transmitting the sub-period information to the remaining periodic trend decomposition modules layer by layer, and outputting the second period information.

[0080] The time series prediction model of each analysis subsystem adopts the above-mentioned Encoder-Decoder-Feder framework. The Encoder input passes through at least one cycle-trend decomposition module, and each layer decomposes the signal into sub-cycle information and second trend information T2. ​​Multiple cycle-trend decomposition modules input the corresponding sub-cycle information layer by layer and finally output the second cycle information S2 to the Decoder. T2 is not input to the Decoder. Then, in the Decoder part, the input of S2 is decoded and output by the Decoder to obtain the third cycle information S3.

[0081] It should be noted that, in some embodiments, the time series prediction model also includes an attention layer; before obtaining the period difference information of the third period information and the first period information, the first trend information and the trend difference information of the second trend information, it also includes: inputting the second period information and the third period information into the attention layer, the attention layer performs frequency domain correlation learning on the second period information and the third period information, and obtains frequency domain correlation information; updates the parameters of the decoding layer according to the frequency domain correlation information to update the third period information. Through the setting of the attention layer, the frequency domain correlation learning can be performed on the second period information and the third period information to obtain frequency domain correlation information; the parameters of the decoding layer are updated according to the frequency domain correlation information to update the third period information.

[0082] S103. Obtain global variable information of each analysis subsystem, and generate federated variable information based on multiple global variable information.

[0083] Specifically, the provided method can obtain the global variable information G of each analysis subsystem, and can determine the federated variable information based on each global variable information G, such as inputting multiple global variable information G through a preset federated learning model to output federated variable information G', or averaging multiple global variable information G to obtain federated variable information G'. Then, the optimization of each analysis subsystem can be completed through the federated variable information.

[0084] S104. Optimize each time series prediction model according to the federated variable information and the global variable information corresponding to each analysis subsystem, complete the training of multiple time series prediction models, and obtain the constructed time series analysis system.

[0085] Specifically, according to the federal variable information G′ and the global variable information G corresponding to each analysis subsystem, the corresponding time series prediction model can be optimized, thereby completing the training of multiple time series prediction models and obtaining a trained time series analysis system. The provided method can be extended to the existing encoding-decoding architecture, has good scalability, and only uploads global variable information, does not involve period and trend information, and has better data privacy and security. And based on the federal method, multiple participants can jointly participate in the training of the model while considering the global trend, which can better utilize the data resources and knowledge experience of all parties. In summary, the system constructed by the learning method proposed in this application has scalability, better privacy and security, and can learn more accurate hidden variables of time series data.

[0086] In some embodiments, each time series prediction model is optimized according to the federated variable information and the global variable information corresponding to each analysis subsystem, further comprising: obtaining the correlation information between the federated variable information and the global variable information corresponding to each analysis subsystem based on the self-attention mechanism; updating the global variable information corresponding to the analysis subsystem according to the correlation information, so as to optimize the parameters of the time series prediction model corresponding to the analysis subsystem according to the global variable information. By obtaining the correlation information of the federated variable information G′ and the global variable information G corresponding to each analysis subsystem through the self-attention mechanism, the parameters of the time series prediction model corresponding to the analysis subsystem can be optimized according to the global variable information.

[0087] When the provided method is applied in a financial system, for example, the provided method can combine multiple sub-suppliers or data managers in multiple different regions to conduct federated learning, so that the trained time series analysis system can better predict, classify and detect anomalies of time series in the financial system, and play a predictive and auxiliary role in risk control, sales, management and other aspects.

[0088] like Figure 4 As shown, Figure 41 is a schematic diagram of the structure of a device for constructing a time series analysis system provided in an embodiment of the present application, and the device for constructing the time series analysis system is used to execute the aforementioned method for constructing the time series analysis system. The time series analysis system to be constructed includes multiple analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem. The device for constructing the time series analysis system can be configured on a terminal or a server.

[0089] like Figure 4 As shown, the construction device 200 of the time series analysis system includes a sequence acquisition unit 201, a variable acquisition unit 202, a federation generation unit 203 and a construction completion unit 204.

[0090] A sequence acquisition unit 201 is used to acquire the original time series corresponding to each analysis subsystem;

[0091] The variable acquisition unit 202 is used to input the original time series into the time series prediction model of the corresponding analysis subsystem, and the time series prediction model decomposes the original time series to obtain global variable information;

[0092] A federation generating unit 203 is used to obtain global variable information of each analysis subsystem and generate federation variable information according to a plurality of global variable information;

[0093] The construction completion unit 204 is used to optimize each time series prediction model according to the federated variable information and the global variable information corresponding to each analysis subsystem, complete the training of multiple time series prediction models, and obtain a constructed time series analysis system.

[0094] In some embodiments, the time series prediction model includes a decomposition layer, and the variable acquisition unit 202 is also used to: input the original time series into the decomposition layer of each analysis subsystem, and the decomposition layer decomposes the original time series to obtain global variable information.

[0095] Exemplarily, the decomposition layer decomposes the original time series and obtains the first period information and the first trend information; the time series prediction model also includes an encoding layer and a decoding layer; the variable acquisition unit 202 is also used to: input the original time series into the encoding layer, the encoding layer encodes the original time series, and obtains the second period information and the second trend information; input the first period information into the decoding layer, the decoding layer decodes the first period information, and obtains the third period information; obtains the period difference information between the third period information and the first period information, and the trend difference information between the first trend information and the second trend information; optimizes the decomposition layer according to the period difference information and the trend difference information and updates the global variable information, so that the first period information approaches the third period information, and the first trend information approaches the second trend information.

[0096] It should be noted that, in some embodiments, the coding layer includes at least one periodic trend decomposition module; the variable acquisition unit 202 is also used to: input the original time series into the first periodic trend decomposition module to obtain sub-period information and second trend information; transmit the sub-period information to the remaining periodic trend decomposition modules layer by layer, and output the second period information.

[0097] It should be noted that, in some embodiments, the time series prediction model also includes an attention layer; before obtaining the period difference information between the third period information and the first period information, the first trend information and the trend difference information of the second trend information, the variable acquisition unit 202 is also used to: input the second period information and the third period information into the attention layer, the attention layer performs frequency domain correlation learning on the second period information and the third period information, and obtains frequency domain correlation information; updates the parameters of the decoding layer according to the frequency domain correlation information to update the third period information.

[0098] In some embodiments, the construction completion unit 204 is also used to: obtain the correlation information between the federal variable information and the global variable information corresponding to each analysis subsystem based on the self-attention mechanism; update the global variable information corresponding to the analysis subsystem according to the correlation information, so as to optimize the parameters of the time series prediction model corresponding to the analysis subsystem according to the global variable information.

[0099] In some embodiments, the sequence acquisition unit 201 is also used to: acquire multiple original time series collected by multiple analysis subsystems; construct a local data set corresponding to each analysis subsystem based on the original time series corresponding to each analysis subsystem; and acquire the original time series with the same time axis from the local data set corresponding to each analysis subsystem.

[0100] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0101] The above-mentioned device can be implemented in the form of a computer program. Figure 5 Runs on the computer device shown.

[0102] See also Figure 5 , Figure 5 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device may be a server. Figure 5 The computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0103] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any method for constructing a time series analysis system.

[0104] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0105] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any method for constructing a time series analysis system.

[0106] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0107] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0108] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0109] Get the original time series corresponding to each analysis subsystem;

[0110] The original time series is input into the time series prediction model of the corresponding analysis subsystem. The time series prediction model decomposes the original time series to obtain global variable information.

[0111] Obtain the global variable information of each analysis subsystem, and generate federated variable information based on multiple global variable information;

[0112] Each time series prediction model is optimized according to the federated variable information and the global variable information corresponding to each analysis subsystem, the training of multiple time series prediction models is completed, and the constructed time series analysis system is obtained.

[0113] In some embodiments, the time series prediction model includes a decomposition layer, and the original time series is input into the time series prediction model of the corresponding analysis subsystem, including:

[0114] The original time series is input into the decomposition layer of each analysis subsystem, and the decomposition layer decomposes the original time series to obtain global variable information.

[0115] Exemplarily, the decomposition layer decomposes the original time series and obtains the first period information and the first trend information; the time series prediction model also includes a coding layer and a decoding layer; the method also includes: inputting the original time series into the coding layer, the coding layer encodes the original time series, and obtains the second period information and the second trend information; inputting the first period information into the decoding layer, the decoding layer decodes the first period information, and obtains the third period information; obtains the period difference information between the third period information and the first period information, and the trend difference information between the first trend information and the second trend information; according to the period difference information and the trend difference information, the decomposition layer is optimized and the global variable information is updated, so that the first period information approaches the third period information, and the first trend information approaches the second trend information.

[0116] It should be noted that, in some embodiments, the coding layer includes at least one periodic trend decomposition module; the original time series is input into the coding layer, and the coding layer encodes the original time series to obtain second period information and second trend information, including: inputting the original time series into the first periodic trend decomposition module to obtain sub-period information and second trend information; transmitting the sub-period information to the remaining periodic trend decomposition modules layer by layer, and outputting the second period information.

[0117] It should be noted that, in some embodiments, the time series prediction model also includes an attention layer; before obtaining the period difference information between the third period information and the first period information, the first trend information and the trend difference information of the second trend information, it also includes: inputting the second period information and the third period information into the attention layer, the attention layer performs frequency domain correlation learning on the second period information and the third period information to obtain frequency domain correlation information; updating the parameters of the decoding layer according to the frequency domain correlation information to update the third period information.

[0118] In some embodiments, each time series prediction model is optimized based on the federated variable information and the global variable information corresponding to each analysis subsystem, and also includes: obtaining the correlation information between the federated variable information and the global variable information corresponding to each analysis subsystem based on the self-attention mechanism; updating the global variable information corresponding to the analysis subsystem according to the correlation information, so as to optimize the parameters of the time series prediction model corresponding to the analysis subsystem according to the global variable information.

[0119] In some embodiments, obtaining the original time series corresponding to each analysis subsystem includes: obtaining multiple original time series collected by multiple analysis subsystems; constructing a local data set corresponding to each analysis subsystem based on the original time series corresponding to each analysis subsystem; and obtaining the original time series with the same time axis from the local data set corresponding to each analysis subsystem.

[0120] A computer-readable storage medium is also provided in an embodiment of the present application. The computer-readable storage medium stores a computer program. The computer program includes program instructions. The processor executes the program instructions to implement any method for constructing a time series analysis system provided in an embodiment of the present application.

[0121] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0122] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for constructing a time series analysis system, characterized in that: The time series analysis system to be constructed includes a plurality of analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem; the method includes: Obtaining the original time series corresponding to each of the analysis subsystems; Inputting the original time series into the corresponding time series prediction model of the analysis subsystem, the time series prediction model decomposes the original time series to obtain global variable information; Acquire the global variable information of each analysis subsystem, and generate federated variable information according to a plurality of the global variable information; Each of the time series prediction models is optimized according to the federated variable information and the global variable information corresponding to each of the analysis subsystems, the training of the multiple time series prediction models is completed, and a constructed time series analysis system is obtained.

2. The method according to claim 1, characterized in that The time series prediction model includes a decomposition layer, and the inputting of the original time series into the corresponding time series prediction model of the analysis subsystem includes: The original time series is input into the decomposition layer of each of the analysis subsystems, and the decomposition layer decomposes the original time series to obtain the global variable information.

3. The method according to claim 2, characterized in that The decomposition layer decomposes the original time series and obtains first period information and first trend information; the time series prediction model also includes an encoding layer and a decoding layer; the method also includes: Inputting the original time series into the encoding layer, the encoding layer encodes the original time series to obtain second period information and second trend information; Inputting the first cycle information into the decoding layer, the decoding layer decoding the first cycle information to obtain third cycle information; Acquire period difference information between the third period information and the first period information, and trend difference information between the first trend information and the second trend information; The decomposition layer is optimized and the global variable information is updated according to the period difference information and the trend difference information, so that the first period information approaches the third period information, and the first trend information approaches the second trend information.

4. The method according to claim 3, characterized in that The encoding layer includes at least one period trend decomposition module; the original time series is input into the encoding layer, and the encoding layer encodes the original time series to obtain the second period information and the second trend information, including: Input the original time series into the first cycle trend decomposition module to obtain sub-cycle information and second trend information; The sub-period information is transmitted layer by layer to the remaining period trend decomposition modules, and the second period information is output.

5. The method according to claim 3, characterized in that: The time series prediction model further includes an attention layer; before obtaining the period difference information of the third period information and the first period information, the trend difference information of the first trend information and the second trend information, further includes: Inputting the second period information and the third period information into the attention layer, and the attention layer performs frequency domain correlation learning on the second period information and the third period information to obtain frequency domain correlation information; The parameters of the decoding layer are updated according to the frequency domain correlation information to update the third period information.

6. The method according to claim 1, characterized in that The optimizing each of the time series prediction models according to the federated variable information and the global variable information corresponding to each of the analysis subsystems further includes: Acquire the correlation information between the federated variable information and the global variable information corresponding to each of the analysis subsystems based on the self-attention mechanism; The global variable information corresponding to the analysis subsystem is updated according to the correlation information, so as to optimize the parameters of the time series prediction model corresponding to the analysis subsystem according to the global variable information.

7. The method according to claim 1, characterized in that The obtaining of the original time series corresponding to each of the analysis subsystems comprises: Acquire multiple original time series collected by multiple analysis subsystems; Constructing a local data set corresponding to each analysis subsystem according to the original time series corresponding to each analysis subsystem; The original time series with the same time axis is obtained from the local data set corresponding to each analysis subsystem.

8. A device for constructing a time series analysis system, characterized in that: The time series analysis system to be constructed includes multiple analysis subsystems, each of which stores a time series prediction model to be trained, and the time series prediction model is used to predict the time series corresponding to the analysis subsystem; including: A sequence acquisition unit, used for acquiring the original time series corresponding to each of the analysis subsystems; A variable acquisition unit, used for inputting the original time series into the corresponding time series prediction model of the analysis subsystem, wherein the time series prediction model decomposes the original time series to obtain global variable information; A federation generating unit, configured to obtain the global variable information of each of the analysis subsystems, and generate federation variable information according to a plurality of the global variable information; The construction completion unit is used to optimize each of the time series prediction models according to the federal variable information and the global variable information corresponding to each of the analysis subsystems, complete the training of multiple time series prediction models, and obtain the constructed time series analysis system.

9. A computer device, characterized in that: The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and implement the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to implement the method according to any one of claims 1 to 7.

Images