Data processing method and electronic device

Abstract

SOLUTION: A method comprises: acquiring a time-series dataset comprising multiple time-series data items each comprising a time and multiple attribute parameters corresponding thereto; acquiring a target attribute parameter being at least one of the multiple attribute parameters; determining, based on a time-series model, an influence factor of the target attribute parameter, the influence factor indicating an attribute parameter influencing the target attribute parameter and a time corresponding thereto; and outputting the influence factor.EFFECT: For a target attribute parameter, at least one influence factor of a target attribute parameter can be determined based on a time-series dataset and a time-series model, thereby facilitating provision of accurate reference information for a decision on the target attribute parameter and enabling the decision more accurate.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] Embodiments of the present disclosure generally relate to the computer field, and more specifically relate to a data processing method, an electronic device, a computer-readable storage medium, and a computer program product. [Background Art]

[0002] With the development of technology, causal analysis is widely applied in various fields such as retail, energy control, and intervention in abnormal events. Causal analysis enables relevant decision-making. However, existing solutions usually require manually setting elements such as the causes of event occurrence. In this case, on one hand, human cost increases excessively; on the other hand, there is uncertainty caused by manual operation, which is prone to errors. For this reason, more appropriate decision-making cannot be obtained. [Summary of the Invention] [Problem to be Solved by the Invention]

[0003] According to an exemplary embodiment of the present disclosure, there is provided a data processing solution capable of determining at least one influencing factor that affects a target attribute parameter. [Means for Solving the Problem]

[0004] A first aspect of this disclosure provides a data processing method. The method includes obtaining a time series dataset comprising a plurality of time series data items, each of which comprises a time and a plurality of corresponding attribute parameters; obtaining a target attribute parameter which is at least one of the plurality of attribute parameters; determining at least one influencing element of the target attribute parameter based on a time series model, wherein the at least one influencing element indicates at least one attribute parameter that affects the target attribute parameter and at least one time corresponding to the at least one attribute parameter; and outputting the at least one influencing element.

[0005] A second aspect of this disclosure provides a data processing method, the method comprising: obtaining a time-series dataset relating to sales records, wherein the time-series dataset comprises a plurality of time-series data items, each of which comprises a time and a plurality of corresponding attribute parameters, the plurality of attribute parameters comprising at least one of purchase price, selling price, sales volume, inventory quantity, or number of customer views; determining at least one influencing factor on an inventory quantity based on a time-series model, wherein the at least one influencing factor indicates at least one attribute parameter that affects the inventory quantity and at least one time corresponding to the at least one attribute parameter; and determining the inventory quantity for the next time based on the at least one influencing factor.

[0006] A third aspect of this disclosure provides a data processing method, the method comprising: obtaining a time-series dataset relating to the power of an IoT device, wherein the time-series dataset comprises a plurality of time-series data items, each of which comprises time and a plurality of corresponding attribute parameters, the plurality of attribute parameters comprising at least one of electricity rates, power demand, voltage, current, temperature, humidity, atmospheric pressure, or the power consumption of the IoT device; determining at least one influencing factor of the power consumption based on a time-series model, wherein the at least one influencing factor indicates at least one attribute parameter that affects the power consumption and at least one time corresponding to the at least one attribute parameter; and determining the uptime of the IoT device based on the at least one influencing factor.

[0007] A fourth aspect of this disclosure provides a data processing method. The method includes obtaining a time-series dataset collected from a social network, wherein the time-series dataset comprises a plurality of time-series data items, each of which comprises a time and a corresponding plurality of attribute parameters, the plurality of attribute parameters comprising at least one of user identification information, user posts, number of views of user posts, number of comments on user posts, or anomaly indicators; determining at least one influencing element of an anomaly indicator based on a time-series model, wherein the at least one influencing element indicates at least one attribute parameter that affects a target attribute parameter and at least one time corresponding to the at least one attribute parameter; and outputting presentation information indicating users with anomaly risk based on the at least one influencing element.

[0008] A fifth aspect of this disclosure provides an electronic device. The electronic device includes at least one processor unit and at least one memory coupled to the at least one processor unit and storing instructions that are executed by the at least one processor unit. When such instructions are executed by the at least one processor unit, the electronic device causes the electronic device to perform the method described in any of the first to fourth aspects of this disclosure.

[0009] A sixth aspect of this disclosure provides an electronic device, which includes memory and a processor. The memory is for storing one or more computer instructions, which are executed by the processor to carry out a method described in any of the first to fourth aspects of this disclosure.

[0010] A seventh aspect of this disclosure provides a computer-readable storage medium which stores machine-readable instructions which, when executed by a device, cause the device to perform a method described in any of the first to fourth aspects of this disclosure.

[0011] In the eighth aspect of this disclosure, a computer program product is provided. The computer program product includes computer-readable instructions, which, when executed by a processor, perform a method described in any of the first to fourth aspects of this disclosure.

[0012] A ninth aspect of this disclosure provides an electronic device, which includes a processing circuit device configured to perform a method described in any of the first to fourth aspects of this disclosure.

[0013] The summary section of the invention is intended to provide a simplified introduction to a set of concepts. These will be further explained in the embodiments described below. The description in the summary section of the invention is not intended to highlight any important or necessary features of the disclosure, nor is it intended to limit the scope of the disclosure. Other features of the disclosure should be readily apparent from the description below. [Brief explanation of the drawing]

[0014] The above and other features, advantages and aspects of each embodiment of this disclosure should become clearer by referring to the following detailed description in conjunction with the drawings. In the drawings, identical or similar reference numerals indicate identical or similar elements. [Figure 1] A block diagram of an exemplary environment according to the embodiments of this disclosure is shown. [Figure 2] A schematic flowchart of the data processing process according to some embodiments of this disclosure is shown. [Figure 3] A schematic block diagram of an exemplary system relating to several embodiments of this disclosure is shown. [Figure 4] A schematic block diagram illustrating the interaction between a user interface module and a time series analysis engine according to some embodiments of this disclosure is shown. [Figure 5A] A schematic diagram of the first time relationship diagram according to some embodiments of the present disclosure is shown. [Figure 5B] A schematic diagram of a second time relationship diagram according to some embodiments of the present disclosure is shown. [Figure 6] An exemplary flowchart of the process for obtaining a second time relationship diagram according to some embodiments of this disclosure is shown. [Figure 7] A schematic flowchart of the process for determining key elements according to some embodiments of this disclosure is shown. [Figure 8A] A schematic diagram illustrating the determination of the relationship between any two numerical intervals, a and b, according to some embodiments of this disclosure is shown. [Figure 8B]FIG. 1 is a schematic diagram illustrating determination of two numerical intervals based on the ordering of three intervals a, b, and c according to some embodiments of the present disclosure. [Figure 9] FIG. 2 shows displays of a plurality of influence values according to some embodiments of the present disclosure. [Figure 10A] FIG. 3 is a schematic diagram showing a relationship between a temporal change in influence of sales volume on a target attribute parameter according to some embodiments of the present disclosure. [Figure 10B] FIG. 4 is a schematic diagram showing a relationship between a temporal change in influence of user views on a target attribute parameter according to some embodiments of the present disclosure. [Figure 11] FIG. 5 is a schematic flowchart of a process for determining a response relationship according to some embodiments of the present disclosure. [Figure 12A] FIG. 6 is a schematic diagram of a temporal change in influence of sales volume adjustment on a target attribute parameter according to some embodiments of the present disclosure. [Figure 12B] FIG. 7 is a schematic diagram of a cumulative temporal change in cumulative influence of sales volume adjustment on a target attribute parameter according to some embodiments of the present disclosure. [Figure 13A] FIG. 8 is a schematic diagram of a temporal change in influence of user view adjustment on a target attribute parameter according to some embodiments of the present disclosure. [Figure 13B] FIG. 9 is a schematic diagram of a cumulative temporal change in cumulative influence of user view adjustment on a target attribute parameter according to some embodiments of the present disclosure. [Figure 14] FIG. 10 is a schematic flowchart of a process for determining an inventory amount according to some embodiments of the present disclosure. [Figure 15] FIG. 11 is a schematic flowchart of a process for determining operating time of a device according to some embodiments of the present disclosure. [Figure 16] FIG. 12 is a schematic diagram of temporal changes of a plurality of collected attribute parameters according to some embodiments of the present disclosure. [Figure 17] FIG. 13 is a schematic flowchart of a process for determining an abnormal event according to some embodiments of the present disclosure. [Figure 18A] Schematic diagrams for determining abnormal risk according to some embodiments of this disclosure are shown. [Figure 18B] Schematic diagrams for determining abnormal risk according to some embodiments of this disclosure are shown. [Figure 18C] Schematic diagrams for determining abnormal risk according to some embodiments of this disclosure are shown. [Figure 19] A block diagram of an exemplary device on which embodiments of this disclosure can be implemented is shown. [Modes for carrying out the invention]

[0015] The embodiments of this disclosure will be described in more detail below with reference to the drawings. While the drawings show several embodiments of this disclosure, it should be understood that this disclosure can be implemented in various forms and should not be interpreted as being limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. Furthermore, the drawings and embodiments of this disclosure are illustrative only and are not intended to limit the scope of protection of this disclosure.

[0016] In the description of embodiments of this disclosure, “including” and similar terms should be understood as open, i.e., “including, but not limited to…”. The term “based on…” should be understood as “based at least in part”. The term “one embodiment” or “the embodiment” should be understood as “at least one embodiment”. Terms such as “first,” “second,” etc., may refer to different or the same subject matter. Further explicit and implicit definitions may be included in the following text.

[0017] The individual methods and processes described in the embodiments of this disclosure may be applied to various electronic devices such as terminal devices and network devices. Furthermore, the embodiments of this disclosure may be performed in test devices such as signal generators, signal analyzers, spectrum analyzers, network analyzers, test terminal devices, test network devices, and channel emulators.

[0018] In describing embodiments of this disclosure, the term “circuit” may refer to hardware circuits and / or combinations of hardware circuits and software. For example, a circuit may be a combination of analog and / or digital hardware circuits and software / firmware. As another example, a circuit may be any part of a hardware processor that includes software. A hardware processor includes (multiple) digital signal processors, software, and (multiple) memories, and these (multiple) digital signal processors, software, and (multiple) memories work together to enable a device, such as a computing device, to operate and perform various functions. In yet another example, a circuit may be a hardware circuit and / or processor, such as a microprocessor or a part of a microprocessor, which requires software / firmware for operation, but may be absent if software is not required for operation. As used herein, the term “circuit” includes implementations of hardware circuits or (multiple) processors only, or implementations of hardware circuits or (multiple) processors with accompanying software and / or firmware.

[0019] Agents need to make a series of decisions on various problems across many domains. While agents have access to large datasets, making more accurate decisions remains difficult. For example, determining which datasets to collect, identifying sufficiently predictable patterns, and assessing the relative importance of individual variables are typically done manually by people, such as data scientists or departmental managers. This heavy reliance on human intervention in decision-making is both highly inefficient and results in insufficient accuracy.

[0020] In view of this, embodiments of the present disclosure provide a data processing solution to address one or more of the above-mentioned problems and / or other potential problems. This solution utilizes a time-series model to determine at least one influencing factor that affects a target attribute parameter. This eliminates manual specification and improves processing efficiency. Furthermore, the influencing factor determined by the time-series model can be used in subsequent decision-making, leading to more accurate decision-making.

[0021] Figure 1 shows a block diagram of an exemplary environment 100 according to an embodiment of the present disclosure. It should be understood that the environment 100 shown in Figure 1 is merely one example of a possible embodiment of the present disclosure and is not intended to limit the scope of the present disclosure. Embodiments of the present disclosure are equally applicable to other systems or architectures.

[0022] As shown in Figure 1, the environment 100 may include a computing device 110. The computing device 110 may be any device with computing capabilities. The computing device 110 may include, but is not limited to, personal computers, server computers, portable or laptop devices, mobile devices (mobile phones, personal digital assistants (PDAs), media players, etc.), wearable devices, home electrical appliances, minicomputers, mainframes, distributed computing systems, cloud computing resources, etc. It should be understood that the computing device 110 may or may not have additional resources with sufficient computing power for model training, taking into consideration factors such as cost.

[0023] The computing device 110 may be configured to acquire a time-series dataset 120 and output at least one influencing element 140. The decision regarding the at least one influencing element 140 may be made by a time-series model 130.

[0024] For example, the time-series dataset 120 may be entered by a user or retrieved from a storage device, and is not limited thereto. In some embodiments, the time-series dataset 120 may include a plurality of time-series data items, each data item may include time and a plurality of corresponding attribute parameters. Optionally, different data items may have different times. Optionally, the plurality of attribute parameters may be referred to as a plurality of parameters, a plurality of variables, or other names, etc., in some examples, and is not limited thereto.

[0025] In some illustrative examples, embodiments of the present disclosure can be used in the retail sector, for example, for retail demand forecasting. For example, each data item in the time series dataset 120 may include time and a number of corresponding attribute parameters. The attribute parameters may be one or more of the following: purchase price, selling price, sales volume, inventory volume, number of customer views, or others.

[0026] In some illustrative examples, embodiments of the present disclosure can be used in the energy sector, for example, for power control of Internet of Things (IoT) devices. For example, each data item in the time-series dataset 120 may include time and a number of corresponding attribute parameters. The attribute parameters may be one or more of the following: electricity rates, power demand, voltage, current, temperature, humidity, atmospheric pressure, or power consumption of an IoT device.

[0027] In some illustrative examples, embodiments of the present disclosure can be used to intervene in anomaly events, for example, to predict anomaly events such as suicides based on information on social networks. For example, each data item in the time-series dataset 120 may include time and a number of corresponding attribute parameters. The attribute parameters may be one or more of user identification information, user posts, the number of views of user posts, the number of comments on user posts, or anomaly indicators.

[0028] The scenarios listed above are for illustrative purposes only and should not be used to limit the scope of this disclosure. Embodiments of this disclosure may be applied to various fields where similar problems exist, and these are not listed here. Furthermore, the term "action" in embodiments of this disclosure may be referred to as, for example, "decision-making," and this disclosure does not limit this to such terms.

[0029] One point that can be understood is that time series can be used to represent data on spatiotemporal processes in the real world. When making decisions, predictions can be made using correlations over a specific time length, and by determining the "specific time length" as the minimum sufficient statistic, the accuracy of decisions can be improved. The minimum sufficient statistic can be understood as follows: correlations smaller than a specific time length are not sufficient for prediction, and correlations exceeding a specific time length may be contaminated with noise and contain some redundant information.

[0030] In some examples, the correlation between different time points in a time series may be represented by a graph (e.g., a causal graph), where the nodes of the graph represent certain attribute parameters at a given time. Pointing from node to node signifies a causal relationship between two nodes, and a linear relationship can be depicted by reinforcing the quantitative relationships of the edges. This will be explained in detail below in conjunction with Figures 2 to 17.

[0031] Figure 2 shows a schematic flowchart of a process 200 used for data processing according to some embodiments of the present disclosure. For example, process 200 may be executed by the computing device 110 shown in Figure 1. It should be understood that process 200 may further include additional blocks not shown, and / or some of the blocks shown may be omitted. The scope of the present disclosure is not limited in this respect.

[0032] In block 210, a time series dataset is obtained. The time series dataset contains multiple time series data items, each time series data item containing time and corresponding multiple attribute parameters. In block 220, a target attribute parameter is obtained. The target attribute parameter is at least one of the multiple attribute parameters. In block 230, at least one influencing element of the target attribute parameter is determined based on the time series model. The at least one influencing element indicates at least one attribute parameter that affects the target attribute parameter and at least one time corresponding to at least one attribute parameter. In block 240, at least one influencing element is output.

[0033] In some embodiments, the data items of a time-series dataset are:

number

number

[0034] [Table 1]

[0035] To ensure that this is understood, the time series datasets shown in Table 1 above are merely illustrative examples, and in actual applications, time series datasets may be represented in other formats, and this disclosure is not limited to this.

[0036] In some embodiments, a time-series dataset entered by the user and target attribute parameters entered by the user may be obtained. For example, the target attribute parameters may be parameters of interest to the user or parameters that the user intends to use for decision-making.

[0037] In some other embodiments, an initial time-series dataset input by the user may be obtained, and after data processing, the time-series dataset of block 210 may be obtained. Optionally, data processing may include noise reduction and / or generation of augmented data items.

[0038] In some embodiments of this disclosure, the time series model may be obtained by pre-training, pre-built by a data scientist, for example, or by other means. This disclosure is not limited to this. Optionally, the time series model may be determined from a group of candidate models.

[0039] For example, different influence elements among the at least one influence element determined in block 230 may have different influences on the target attribute parameter. Optionally, the influence may be expressed in the form of an interval including, for example, a maximum influence and a minimum influence.

[0040] In some embodiments, if there are multiple influencing elements determined in block 230, the multiple influencing elements may be ordered based on their influence, and some or all of the multiple influencing elements may be output based on the ordering. For example, one or more influencing elements with the greatest influence may be determined.

[0041] To understand it, elemental analysis is a data analysis technique that can be used to determine the impact of influencing factors on target attribute parameters. For example, it can determine the importance of product quality or product price to customer attitudes towards a brand. Therefore, elemental analysis enables prior prediction and leads to more accurate decision-making.

[0042] In some embodiments of this disclosure, the change over time in the influence of a particular influencing element on a target attribute parameter may be determined for a specific influencing element among the at least one influencing element determined in block 230, for example, for a first influencing element representing a first attribute parameter. Optionally, the first influencing element may be the influencing element with the greatest influence determined through ordering.

[0043] In some embodiments, the effect of adjusting a specific attribute parameter on a target attribute parameter may also be determined. For example, an adjustment instruction for a second attribute parameter may be obtained, and based on that adjustment instruction, the relationship between the influence of the second attribute parameter on the target attribute parameter and time may be determined.

[0044] Figure 3 shows a schematic block diagram of an exemplary system 300 according to some embodiments of the present disclosure. As shown in Figure 3, the system 300 includes a user interface module 310 and a time series analysis engine 320. The user interface module 310 may transmit an input 301 to the time series analysis engine 320, which may perform various operations to obtain an output 302. In addition or optionally, the time series analysis engine 320 may further transmit the output 302 to the user interface module 310 for display to the user. Exemplarily, the time series analysis engine 320 may include a relation discovery module 322 and a model-based analysis module 323. In addition or optionally, the time series analysis engine 320 may further include a data processing module 321.

[0045] Figure 4 shows a schematic block diagram of the interaction between a user interface module 310 and a time series analysis engine 320 according to some embodiments of the present disclosure. As shown in Figure 4, the user interface module 310 includes an input interface 311, a data preparation interface 312, a model selection interface 313, a relation discovery interface 314, and a model-based analysis interface 315. As shown in Figure 4, the time series analysis engine 320 includes a data processing module 321, a relation discovery module 322, and a model-based analysis module 323. The data processing module 321 includes a basic data preprocessing submodule 3211 and a data augmentation submodule 3212, the relation discovery module 322 includes a model selection submodule 3221, a time series modeling submodule 3222, and a model validation submodule 3223, and the model-based analysis module 323 includes a key element identification submodule 3231 and an element analysis submodule 3232.

[0046] As an example, the relationship discovery module 322 may automatically search for potential patterns in a given dataset using causal discovery techniques. Causal discovery techniques may include, but are not limited to, the Peter-Clark (PC) Algorithm, GES (Greedy Equivalent Search), Linear non-Gaussian Model (LinGAM), or Causal Additive Model (CAM). In embodiments of this disclosure, a model may be constructed based on a Bayesian network. The Bayesian network can identify individual direct and indirect causes of a target and imply causal relationships to derive causal conclusions.

[0047] In some examples, embodiments of the present disclosure may perform analyses based on time series models, which may optionally be causal models and may be represented, for example, in the form of graphs.

[0048] In some embodiments, the input interface 311 may be used to receive a time-series dataset. For example, the time-series dataset may be provided to the data processing module 321. In some embodiments, the input interface 311 may be used to receive configuration information for model selection. For example, the configuration information for model selection may include user suggestions for model selection, time-series model structure information, etc. For example, the configuration information for model selection may be provided to the model selection submodule 3221. In some embodiments, the input interface 311 may be used to receive configuration information related to analysis. For example, the configuration information related to analysis may include user requests and / or target attribute parameter instructions, etc. Optionally, for example, the configuration information related to analysis may be provided to the key element identification submodule 3231.

[0049] In some embodiments, the data preparation interface 312 may be used to receive configuration information related to data processing. For example, the configuration information related to data processing may indicate one or more parameters such as data transformation, missing data values, outlier handling, or data augmentation. For example, the configuration information related to data processing may be provided to the data processing module 321, for example, to the basic data preprocessing submodule 3211 and the data augmentation submodule 3212.

[0050] In some embodiments, the model selection interface 313 may be used to receive configuration information related to model selection. For example, the configuration information related to model selection may indicate the criteria for model selection, etc. For example, the configuration information related to model selection may be provided to the model selection submodule 3221.

[0051] In some embodiments, the relationship discovery interface 314 may be used to receive configuration information relating to time series modeling and / or configuration information relating to validation. For example, configuration information relating to time series modeling may be provided to the time series modeling submodule 3222. For example, configuration information relating to validation may indicate validation criteria and may be provided to the model validation submodule 3223.

[0052] In some embodiments, the model-based analysis interface 315 may be used to receive configuration information related to element determination. For example, the configuration information related to element determination may indicate target attribute parameters, the number of influencing elements for which the user desires output, one or more thresholds, adjustment instructions for attribute parameters, etc. For example, the configuration information related to element determination may be provided to a model-based analysis module 323, such as a key element identification submodule 3231 or an element analysis submodule 3232.

[0053] In some embodiments, the data processing module 321 may be used to process a time series dataset to obtain a processed time series dataset. For example, the basic data preprocessing submodule 3211 may perform preprocessing on the time series dataset, including, for example, data cleansing, data range analysis, and missing value imputation.

[0054] In some examples, data cleansing may include removing abnormal data items. For example, if a data item indicates an age of over 200 years, that item may be removed. For example, if a data item indicates a price discount of over 100%, that item may be removed.

[0055] In some examples, data range analysis may include determining the range of values ​​that an attribute parameter of multiple data items can take. In some examples, data range analysis may include determining statistics using quantiles such as 25%, 50%, or 75%. In some examples, data range analysis may include determining the percentage of missing values ​​for an attribute parameter of multiple data items. In some examples, data range analysis may include determining the eigenvalues ​​of each attribute parameter, etc.

[0056] In some examples, if a failure occurs during the time series dataset collection process, such as a sensor failure, and the interval between the two closest time points in the time series dataset is too large, data items may be imputed in the time series dataset by missing value imputation. For example, missing value imputation may be performed based on setting information regarding missing values ​​from the data preparation interface 312.

[0057] Thus, the basic data preprocessing submodule 3211 may preprocess the time series dataset and provide the preprocessed time series dataset to the data augmentation submodule 3212.

[0058] As an example, the basic idea of ​​data augmentation is to generate several synthetic data items that can cover data spaces that have not actually been collected and can maintain accurate labeling. In general situations, multiple data items in a time series dataset can represent a uniform time series, that is, the time series is basically collected at regular time intervals. However, in some fields, such as clinical diagnostics or advertising campaigns, the time of the sample is usually not continuous or even regular. In embodiments of this disclosure, augmentation may be used to obtain augmented data items (also referred to as synthetic data items) to enrich a time series dataset.

[0059] In some embodiments, a data generation model may be used to obtain an augmented data item based on a time series dataset (e.g., a preprocessed time series dataset). An augmented time series dataset can be obtained by adding the augmented data item to the time series dataset (e.g., a preprocessed time series dataset). For example, the first and second data items of a time series dataset (e.g., a preprocessed time series dataset) may be input into the data generation model to obtain an augmented data item, where the first data item has a first time, the second data item has a second time, and the augmented data item has a third time. Optionally, the data generation model may be implemented as a directed acyclic graph (DAG). The data generation model may include two adversarial submodels, and further may include a third submodel for determining the difference between the outputs of the two submodels.

[0060] Thus, the data augmentation submodule 3212 may expand the time series dataset and provide the expanded time series dataset to the model selection submodule 3221.

[0061] In some embodiments, the time series model may be user-inputted. For example, a data scientist may provide a time series model based on their experience. In some embodiments, the time series model may be classified based on the type of data, for example, each data item in different time series datasets may represent a different type, and the corresponding time series model may be different. For example, an appropriate time series model may be determined based on the binary representation of a data item or the continuity feature of a data item. In some embodiments, the time series model may be performance-driven. For example, a time series model may be selected from several candidate models based on the relationship between a time series dataset and a given dataset. Optionally, this relationship may include similarities between features. For example, the extracted feature may be "{%binary=10%, |feature|=100, |sample|=10,000, %missingData=10%}". This indicates that 10% is binary data, the feature dimension is 100, the sample size is 10,000, and the percentage of missing data is 10%. For example, a predetermined dataset with the highest similarity may be determined, and then a candidate model corresponding to that predetermined dataset with the highest similarity may be determined. Optionally, a time series model may be obtained by further fine-tuning the determined candidate model. For example, fine-tuning may be performed through actual interaction with the user, but this disclosure is not limited to this. In some embodiments, the time series model may be determined based on the application scenario.

[0062] Thus, the model selection submodule 3221 may select a time series model for subsequent processing, for example, by providing the best modeling settings to the time series modeling submodule 3222.

[0063] In some examples, the time series modeling submodule 3222 may be used to establish dependencies between multiple attribute parameters across time. Optionally, one method of measuring such dependencies may be based on covariances such as vector autoregressive (VAR) models or Granger causality. Optionally, one method of measuring such dependencies may be based on conditional independence tests such as the PCMCI (Peter and Clark - Momentary Conditional Independence) algorithm, which explores dependencies based on lag time.

[0064] Thus, the time series modeling submodule 3222 may establish dependencies between multiple variables, for example, by providing a p-lag coefficient matrix to the model validation submodule 3223 and the key element identification submodule 3231.

[0065] In some embodiments, the model validation submodule 3223 may be used to evaluate the quality of the time series model, for example, by performing the evaluation based on a set of evaluation metrics. Optionally, the set of evaluation metrics may be included in the configuration information provided by the relation discovery interface 314 for validation. As an example, the model validation method is R 2 This may include, but is not limited to, the Mean Squared Error (MSE), F-statistics, or other methods.

[0066] In some embodiments, the key element identification submodule 3231 may be used to determine a set of influencing elements that affect a target attribute parameter, the set of influencing elements including at least one influencing element. In some examples, a given time series model may be used to determine all causal elements that may affect the target attribute parameter, and a key element set may be determined from all possible causal elements. For example, the causal elements may include time lag causal elements, that is, elements of other times that lag compared to the current time have an effect on the target attribute parameter at the current time. For example, the causal elements may include instantaneous causal elements, that is, other attribute parameters at the current time have an effect on the target attribute parameter at the current time.

[0067] Thus, the key element identification submodule 3231 may be used to identify elements that have a significant impact on the target attribute parameters, for example, by providing a set of key elements to the element analysis submodule 3232.

[0068] In some embodiments, the element analysis submodule 3232 may perform forward or backward analysis. For example, forward analysis may be used to identify the disturbance response with the greatest future impact. For instance, it may be determined which attribute parameter disturbance will cause the cumulative impact on the target attribute parameter to exceed a predetermined threshold at a given time. For example, backward analysis may be used to understand the one or more elements that have the greatest and most important impact at a given time. For example, the elements determined for the same target attribute parameter (e.g., restaurant choice) generally differ at different given times (e.g., weekdays and weekends). Optionally, if the causal relationship is linear, the ordering of related elements across different time periods will be consistent.

[0069] It should be noted that while an exemplary system in the embodiments of this disclosure has been described in conjunction with Figures 3 and 4, this is merely illustrative and should not be interpreted as an limitation to the embodiments of this disclosure. For example, in actual scenarios, the data processing module 321 may be omitted, or the system may include other modules. More detailed embodiments will be described later in conjunction with Figures 5 to 17.

[0070] In some embodiments, an initial time series model may be obtained, and a time series model (which may be called the first time series model) may be obtained based on the initial time series model. In some examples, a time series model may be obtained based on the initial time series model and a time series dataset. Optionally, the time series model may be represented in the form of a time relationship diagram, which includes nodes and edges. For ease of explanation, the initial time series model may be referred to as the first time relationship diagram, and the time series model as the second time relationship diagram.

[0071] For example, the first time relationship diagram may be one entered by the user, or it may be a diagram provided by an expert based on their experience. In some examples, a test (e.g., a t-test) may be performed on each edge of the first time relationship diagram to obtain a value representing the significance level, called the p-value. Then, based on the p-value, it may be determined whether each edge represents a causal relationship, for example, whether the edge is significantly equal to 0. If they are equal, the corresponding edge is retained; if they are not equal, the corresponding edge is deleted. The deleted edges may be considered redundant edges in the first time relationship diagram. In this way, the second time relationship diagram can be obtained by removing the redundant edges.

[0072] In this example, the input may be understood to include a first time relationship diagram, target attribute parameters, and a time series dataset, and the output may include a second time relationship diagram. Optionally, the first time relationship diagram may be represented as G, as shown in Figure 5A as an example. The first time relationship diagram, as shown in Figure 5A, may be input by the user (e.g., provided by an expert) or determined by an algorithm (e.g., constructed by the time series modeling submodule 3222), and is not limited to this disclosure.

[0073] The first time relationship diagram may also represent causal relationships between multiple attribute parameters at different times. For example, the arrows in Figure 5A indicate that cause points to effect. For instance, arrow 501 indicates a causal relationship between the online advertising volume at time t-1 and the online advertising volume at time t. In addition, Figure 5A further circles the target attribute parameter, "sales volume."

[0074] To make it easier to understand, the first time relationship diagram has a lag effect. That is, past information (e.g., t-1, t-2) influences the current value at time t. There is also an instantaneous effect. That is, the remaining attribute parameters at time t have an instantaneous effect on the target attribute parameter at time t.

[0075] Figure 6 shows an exemplary flowchart of a process 600 for obtaining a second time-relationship diagram according to some embodiments of the present disclosure. Block 610 determines several time-lag elements that may have a causal relationship with the target attribute parameter in the first time-relationship diagram. Block 620 determines key elements of some of the several possible time-lag elements by statistical testing. Block 630 removes redundant edges. Exemplarily, input 601 in Figure 6 includes a time-series dataset (e.g., represented as X), a first time-relationship diagram (represented as G), and a target attribute parameter (e.g., represented as T), and output 602 in Figure 6 includes a second time-relationship diagram (e.g., represented as G'). As one example, the second time-relationship diagram is shown in Figure 5B. Alternatively, the second time-relationship diagram may be referred to as a time-causal graph or by other names, and the present disclosure does not limit it in this regard.

[0076] Optionally, in block 610, a Structure Vector Autoregressive (SVAR) algorithm may be used, for example, by assuming that all other attribute parameters (i.e., nodes or variables) have direct edges to the target attribute parameter.

[0077] Optionally, in block 620, a t-test algorithm may be used, for example, by quantizing the arithmetic mean to obtain the p-value for each edge. Furthermore, in block 630, if the p-value is less than the threshold, it means that it is not statistically equal to zero. If an edge is statistically not equal to zero, it means that the starting node of that edge has a relatively large influence on the ending node.

[0078] In some examples, the resulting second time relationship diagram (temporal causal graph) may include (tp+1)*K parameters and their influence on the target attribute parameter. Here, tp represents the time lag value (for example, tp=2 in Figure 5B), and K represents the number of key elements (for example, K=3 in Figure 5B). In this way, the causal effect may be determined by the second time relationship diagram (temporal causal graph).

[0079] Thus, based on the second time relationship diagram, at least one influencing factor of the target attribute parameter (e.g., sales volume in Figure 5B) may be determined. As shown in Figure 5B, direct influencing factors on sales volume at time t include (t-1, sales volume), (t, online advertising volume), and (t, user view volume). Each influencing factor includes the attribute parameter and the corresponding time. To make it clear, direct influencing factors may show an edge directly linked to (t, sales volume). Further indirect influencing factors are included as examples, but are not listed here.

[0080] In some embodiments, each influencing element may have an influence on a target attribute parameter. In one example, the influence may be expressed in numerical form, such as the numbers indicated near a particular edge in Figure 5B. In another example, the influence may be expressed in the form of a numerical interval. Thus, influences expressed in interval form can reflect more statistical information and can be used to more accurately determine causal relationships.

[0081] Furthermore, for multiple influencing elements, some key elements among them may be further determined. Figure 7 shows a schematic flowchart of a process 700 for determining key elements according to some embodiments of the present disclosure. In block 710, the influence of each of the multiple influencing elements on the target attribute parameter is determined. In block 720, the multiple influencing elements are ordered based on the influence of each influencing element. In block 730, some or all of the multiple influencing elements are output based on the ordering.

[0082] In some examples, the user may specify the quantities of some or all of the influencing elements that are output. For example, the user may input a value for such quantity, represented, for example, as m, via the model-based analysis interface 315.

[0083] In this example, the input may be understood as including a second time relationship diagram and a quantity m, and the output as including m key influencing elements.

[0084] Optionally, when ordering in block 720, a comparison matrix may be constructed, the influencing element with the minimum (or maximum) value in the comparison matrix may be determined, and then the order may be determined from the influencing element with the minimum (or maximum) value to complete the ordering.

[0085] In some embodiments, when influence is represented as a numerical interval, it is necessary to determine the relationship based on the interval when ordering. For example, the relationship may be

number

[0086] Specifically, we assume that the first influence of the first influencing element is represented as the first interval, and the second influence of the second influencing element is represented as the second interval. The relationship between the first and second influences may include the following: (1) If the upper limit of the first interval is less than the lower limit of the second interval, the first influence is less than the second influence (<). (2) If the first and second intervals have an overlapping region, and the first statistic of the first interval is less than the second statistic of the second interval, the first influence is less than or equal to the second influence (

number

Math

[0087] As can be understood, the upper limit value may also be referred to as an upper bound value, and the lower limit value may also be referred to as a lower bound value. A reliable interval can be determined based on the upper limit value and the lower limit value, and a true value falls within the reliable interval. For example, the reliable interval may represent a probability of 1-α, where α represents a significance level. For example, when α=0.05, as α increases, the corresponding reliable interval decreases, and as α decreases, the corresponding reliable interval increases.

[0088] As an illustration, at block 730, m influencing factors may be determined based on ordering. As an example, when influence is expressed in the form of numerical intervals, FIG. 8B is a schematic diagram illustrating determining two numerical intervals 820 based on ordering of three intervals a, b and c. In the illustration 822 of FIG. 8B,

Math

[0089] In some other examples, the ordering may be represented in a two-dimensional coordinate graph. For example, FIG. 9 shows a plurality of influences plotted on coordinates, from which m influencing factors may be determined.

[0090] In some embodiments of this disclosure, the change over time of the influence of a particular attribute parameter on a target attribute parameter may be determined. Optionally, the particular attribute parameter may be referred to as the first attribute parameter, which may be specified by the user. For example, the user may input an instruction for the first attribute parameter (e.g., represented as q) via the input interface 311 or the model-based analysis interface 315.

[0091] In this example, the input may be understood as including a second time relationship diagram, a first attribute parameter, and a target attribute parameter, and the output as the change over time in the influence of the first attribute parameter on the target attribute parameter. The change over time can be represented, for example, in the form of a vector.

[0092] Specifically, the influence of the first attribute parameter on the target attribute parameter at each time point can be obtained from the second time relationship diagram, and by extension, its change over time can be obtained. In one example, the change over time in the influence of the first attribute parameter on the target attribute parameter can be represented in the form of a curve (or line graph). Assuming the target attribute parameter is (t, sales volume), Figure 10A shows the relationship 1010 of the change over time in the influence of sales volume on the target attribute parameter, and Figure 10B shows the relationship 1020 of the change over time in the influence of user view volume on the target attribute parameter. In Figures 10A and 10B, the horizontal axis represents the amount of time lag, and the value of the amount of time lag represents the value in time units in which there is a lag with respect to the time of the target attribute parameter. For example, a time lag of 2 represents t-2. If the time unit is days, then t-2 represents 2 days ago. As can be seen from Figure 10A, the influence of past sales volume on current sales volume is decreasing. As can be seen from Figure 10B, past user view volume has an impact on current sales volume in both the short term (e.g., 3 days ago) and the long term (e.g., 2 weeks ago).

[0093] In some embodiments of this disclosure, the model's response may be further determined. For example, one or more attribute parameters may be adjusted to determine the response of a target attribute parameter. For instance, the attribute parameter that maximizes the response of the target attribute parameter may be determined. Figure 11 shows a schematic flowchart of a process 1100 for determining response relationships according to some embodiments of this disclosure. In block 1110, a first adjustment instruction is obtained for a second attribute parameter among a plurality of attribute parameters. In block 1120, based on the first adjustment instruction, the relationship between the influence of the second attribute parameter on the target attribute parameter and time is determined.

[0094] In some examples, the effect of adjusting a second attribute parameter on a target attribute parameter over a predetermined period may be determined. For example, the predetermined period may be entered by the user, for example, by the user entering a value for the predetermined period, represented, for example, D, via the model-based analysis interface 315.

[0095] In this example, the input may be understood to include a second time relationship diagram, a target attribute parameter, a predetermined period D, and a first adjustment instruction for the second attribute parameter. The output may include the change in response over the predetermined period D, represented, for example, by a vector containing D elements.

[0096] In some embodiments, the relationship determined in block 1120 may be expressed as a change in influence over time (with a predetermined time range D), or as a cumulative change in influence over time (with a predetermined time range D). For example, assuming the target attribute parameter is (t, sales volume) and the second attribute parameter is sales volume, Figure 12A shows the change in the influence of sales volume adjustment on the target attribute parameter over time, and Figure 12B shows the cumulative change in the cumulative influence of sales volume adjustment on the target attribute parameter over time.

[0097] Optionally, it may also be determined whether the cumulative influence over a given period is greater than a predetermined threshold. In conjunction with Figure 12B, assuming the predetermined threshold is 15, it can be seen that the cumulative influence over the given period is less than the predetermined threshold.

[0098] In some other examples, two or more attribute parameters may be adjusted separately or simultaneously to determine the response to a target attribute parameter within a specific period D. For example, a second adjustment instruction may be obtained for a third attribute parameter, and based on the second adjustment instruction, the relationship between the influence of the third attribute parameter on the target attribute parameter and a given period may be determined. For example, among the second and third attribute parameters, the attribute parameter that has a relatively large influence on the target attribute parameter may be further determined.

[0099] For example, assuming the target attribute parameter is (t, sales volume) and the third attribute parameter is the number of user views, Figure 13A shows the change over time in the influence of adjusting the number of user views on the target attribute parameter, and Figure 13B shows the change over time in the cumulative influence of adjusting the number of user views on the target attribute parameter. Comparing Figure 12A and Figure 13A, it can be seen that the influence of sales volume is greater.

[0100] Additionally, or optionally, the most effective time point may be determined. The most effective time point represents the time when the optimal effect is obtained if the decision is made. Thus, when the decision is an adjustment instruction for attribute parameters, the time point at which the decision is most effective may be determined based on this. For example, this time point may be determined by determining the threshold for the desired effect, simulating the effect of parameter adjustments, or determining the relationship (e.g., the intersection) between the effect of the adjustments and the threshold for the desired effect. In this way, a more appropriate decision can be made based on this.

[0101] Through the embodiments described above linked to Figures 2 to 13B, this disclosure provides a data processing solution that allows for the determination of at least one influencing factor for a target attribute parameter based on a time-series dataset and a time-series model, and further facilitates the provision of accurate reference information for decision-making regarding the target attribute parameter, thereby making decision-making more accurate.

[0102] To ensure understanding, embodiments of this disclosure are applicable to different fields. Taking the sales field as an example, Figure 14 shows a schematic flowchart of a process 1400 for determining inventory levels according to several embodiments of this disclosure. In block 1410, a time-series dataset relating to sales records is obtained. The time-series dataset includes a plurality of time-series data items. Each time-series data item includes time and a corresponding plurality of attribute parameters. The plurality of attribute parameters include at least one of the following: purchase price, selling price, sales volume, inventory level, or number of customer views. In block 1420, at least one influencing factor on the inventory level is determined based on the time-series model. The at least one influencing factor indicates at least one attribute parameter that affects the inventory level and at least one time corresponding to the at least one attribute parameter. In block 1430, the inventory level for the next time is determined based on the at least one influencing factor. For example, in the sales field, if the target attribute parameter is inventory level, embodiments of this disclosure may determine key factors that have a causal relationship to the inventory level. To ensure clarity, the schematic flowchart shown in Figure 14 is illustrative only, and in actual scenarios, the target attribute parameters may be other variables, such as sales profit, and this disclosure is not limited to this.

[0103] In some other examples, taking the field of green energy as an example, Figure 15 shows a schematic flowchart of a process 1500 for determining the uptime of a device according to some embodiments of the present disclosure. In block 1510, a time-series dataset of power for an IoT device is obtained. The time-series dataset includes a plurality of time-series data items, each time-series data item including time and a plurality of corresponding attribute parameters. The plurality of attribute parameters include at least one of electricity rates, power demand, voltage, current, temperature, humidity, atmospheric pressure, or the power consumption of the IoT device. In block 1520, at least one influencing factor of power consumption is determined based on the time-series model. The at least one influencing factor indicates at least one attribute parameter that affects power consumption and at least one time corresponding to at least one attribute parameter. In block 1530, the uptime of one or more IoT devices is determined based on the at least one influencing factor. For example, in the field of green energy, if the target attribute parameter is power consumption, embodiments of the present disclosure may determine key factors that have a causal relationship with power consumption. To ensure clarity, the schematic flowchart shown in Figure 15 is illustrative only, and in actual scenarios, the target attribute parameters may be other variables, such as household electricity bill payments, and this disclosure is not limited to this.

[0104] Optionally, IoT devices may also be Artificial Intelligence - Internet of Things (AIoT) devices, and devices or systems for energy control of IoT devices may be Green AIoT controllers. As can be understood, modern IoT devices (electrical appliances, cooking appliances, etc.) are becoming increasingly high-performance, but energy consumption comes at the cost of air pollution, water pollution, deforestation, etc. Against this backdrop, Green Transformation refers to actions that use digital technology to solve such environmental problems while achieving economic growth. For example, in a home environment, Green Transformation requires that controllers maintain the internal environment of the home at an appropriate level so that all household electrical appliances operate properly, assuming economic growth. For example, refrigerators consume more power in hot and humid environments. Operating conditions are improved when air conditioning and ventilation systems work together. However, if all electrical appliances are operating at peak times, low voltage will cause the refrigerator to take longer to reach a certain temperature and consume more power. By utilizing the Green AIoT controller in the embodiments of this disclosure, it becomes possible to collect electricity bill data in real time, sense internal temperature and humidity, collect user behavior patterns, and analyze the collected data to determine how to manage energy consumption, thereby achieving the objectives of green transformation.

[0105] For example, a time-series dataset may be obtained by collecting the changes over time of multiple attribute parameters. Referring to Figure 16, the changes over time of electricity rates, changes in urban demand, household voltage, household current, kitchen air conditioner temperature, kitchen humidifier humidity, bedroom air conditioner temperature, and bedroom humidifier humidity are shown, respectively. Optionally, the horizontal axis represents time, and its units may be minutes, hours, or other time units, but this disclosure is not limited to this.

[0106] By utilizing embodiments of this disclosure, the total electricity consumption of a household may be assumed to be a target attribute parameter, and key factors influencing this target attribute parameter may be determined to include the electricity price one hour ago, the current electricity consumption of the kitchen air conditioner, etc. Alternatively, based on a time-series dataset, it may be determined that peak demand in a city may lead to peak electricity prices. Based on this, the controller may determine the operating time of each electrical appliance in the household. For example, most electrical appliances may be operated during off-peak demand in the city.

[0107] In some other examples, taking the field of dealing with anomalous events based on social networks as an example, Figure 17 shows a schematic flowchart of a process 1700 for determining an anomalous event according to some embodiments of the present disclosure. In block 1710, a time-series dataset collected from the social network is obtained. The time-series dataset includes a plurality of time-series data items, each time-series data item including time and a corresponding plurality of attribute parameters. The plurality of attribute parameters include at least one of user identification information, user posts, number of views of user posts, number of comments on user posts, or an anomaly indicator. In block 1720, at least one influencing factor of the anomaly indicator is determined based on the time-series model. The at least one influencing factor indicates at least one attribute parameter that affects the target attribute parameter and at least one time corresponding to the at least one attribute parameter. In block 1730, a suggestion information is output based on the at least one influencing factor. The suggestion information indicates users where an anomaly risk exists.

[0108] For example, abnormal instructions may include indicating whether or not a user committed suicide, or whether or not a user committed illegal acts. To simplify the explanation, a user's suicide will be described below as an abnormal event.

[0109] According to World Health Organization data, nearly 800,000 people commit suicide each year, meaning a tragedy occurs almost every 40 seconds. Due to the rarity of these events, very little data is available. However, the booming internet has led to a surge in data on social networking media. Therefore, identifying the causes and consequences of suicide from social networking media and implementing effective interventions is a highly desirable and feasible solution. Devices or systems used to determine unusual incidents via social networks may be called emotional robots. By collecting users' timely activity, updates on social networks, and influential news, fluctuations in user states can be captured. For example, a time-series dataset can be obtained by collecting attribute parameters at multiple different times.

[0110] In some embodiments, the time-series dataset may be divided into a first subset and a second subset. The first subset may be a suicide dataset, representing users who have committed suicide. The second subset may be a non-suicide dataset, representing users who have not committed suicide. In some examples, similarity matching may further be used to identify one or more non-suicide users from the non-suicide dataset whose similarity to the suicide dataset is higher than a threshold.

[0111] As one example, one could obtain the time series data for 20 suicide users and the time series data for 80 non-suicidal users (normal users). Based on each user's own time series, a separate time series model would be constructed for each user.

[0112] "Mood" can also be used as a target attribute parameter. For example, mood can be quantified and represented through analysis based on the user's text information. The user's text information may include user posts and comments.

[0113] For example, we might assume that the current "mood" of one non-suicidal user (e.g., user 1) out of 80 determined non-suicidal users (normal users) can be expressed as follows: Mood(t) = 5 × Mood(t-1) + 2 × Number of Reads (t-1) + 5 × Number of Reads(t-2).

[0114] For example, we can assume that the current "mood" of one suicide user (e.g., user 2) out of 20 suicide users can be expressed as follows: Mood (t) = -3 × Number of posts at midnight (t-1) - 15 × Mood (t-1) - 5 × Mood (t-3) + 2 × Number of views (t-2).

[0115] In the above formula, mood (t) represents the current mood, t-1 represents the lag of one time unit, t-2 represents the lag of two time units, and t-3 represents the lag of three time units. For example, if the time unit is "days," then t-1, t-2, and t-3 represent 1 day ago, 2 days ago, and 3 days ago, respectively.

[0116] Figures 18A to 18C show schematic diagrams for determining abnormal risk according to some embodiments of the present disclosure. Specifically, for a particular user (e.g., user xyz), a time series of that particular user is collected, and their mood is determined (e.g., predicted) from that time series.

[0117] Referring to Figure 18A, we assume that the actual data collected for a specific user is shown as "x" and the predicted mood is shown as the dashed line 1801. Figure 18A also shows, for example, the mood curves of four non-suicidal users (solid lines) and the mood curves of two suicidal users (dotted lines). Through the analysis, it may be determined that the similarity between a specific user and a suicidal user is 52%, and the similarity with non-suicidal users is 48%. As an example, assuming a threshold of 80%, since 52% < 80%, based on Figure 18A, it may be unnecessary to output the presented information.

[0118] Referring to Figure 18B, we assume that the actual data collected for a specific user is shown as "x" and the predicted mood is shown as the dashed line 1802. Figure 18B also shows, for example, the mood curves of four non-suicidal users (solid lines) and the mood curves of two suicidal users (dotted lines). Through the analysis, it may be determined that the similarity between a specific user and a suicidal user is 73%, and the similarity with a non-suicidal user is 27%. As an example, assuming a threshold of 80%, since 73% < 80%, based on Figure 18B, it may be unnecessary to output the presented information.

[0119] Referring to Figure 18C, we assume that the actual data collected for a specific user is shown as "x" and the predicted mood is shown as the dashed line 1803. Figure 18C also shows, for example, the mood curves (solid lines) for three non-suicidal users and the mood curves (dotted lines) for two suicidal users. Through the analysis, it may be determined that the similarity between a specific user and suicidal users is 89%, and the similarity with non-suicidal users is 11%. As an example, assuming a threshold of 80%, since 89% > 80%, presentation information may be output based on Figure 18C. This presentation information may indicate that the specific user is at risk of suicide.

[0120] To ensure understanding, the examples shown in Figures 18A to 18C are merely illustrative; for example, the figures may show more or fewer mood curves for non-suicidal users and mood curves for suicidal users. This disclosure is not limited in this respect. Also, while the horizontal axis in Figures 18A to 18C represents time (days), in actual scenarios, this time unit may be other time units, and this disclosure is not limited in this respect.

[0121] Thus, embodiments of this disclosure can determine users who may have abnormal risks based on a time series of users on a social network and output relevant information. This facilitates timely intervention and support, reducing the occurrence of tragedies and even making it possible to avoid them altogether.

[0122] To facilitate understanding, the above has shown applicability in some fields in conjunction with Figures 14 to 18C; however, these fields are shown for illustrative purposes only and are not intended to limit the scope of the present invention in any way. Embodiments of this disclosure may be applied to various fields where similar problems exist, and these will not be listed here one by one.

[0123] In some embodiments, the computing device includes circuitry configured to perform the following operations: acquiring a time series dataset containing a plurality of time series data items, each time series data item containing a time and a plurality of corresponding attribute parameters; acquiring a target attribute parameter which is at least one of the plurality of attribute parameters; determining at least one influencing element of the target attribute parameter based on a time series model, wherein the at least one influencing element indicates at least one attribute parameter that affects the target attribute parameter and at least one time corresponding to the at least one attribute parameter; and outputting at least one influencing element.

[0124] In some embodiments, each of at least one influencing element has an influence on a target attribute parameter, and the influence is expressed in the form of an interval.

[0125] In some embodiments, at least one influencing element comprises multiple influencing elements, and the computing device includes circuitry configured to perform the following operations: determining the influence of each of the multiple influencing elements on a target attribute parameter; ordering the multiple influencing elements based on the influence of each influencing element; and outputting some or all of the multiple influencing elements based on the ordering.

[0126] In some embodiments, the multiple influencing elements include a first influencing element and a second influencing element, the first influencing element having a first influence on the target attribute parameter, the second influencing element having a second influence on the target attribute parameter, the first influence being represented as a first interval, the second influence being represented as a second interval, and the computing device includes circuitry configured to perform the following operations: if the upper limit of the first interval is less than the lower limit of the second interval, the first influence is less than the second influence; if the first and second intervals have an overlapping region and the first statistic of the first interval is less than the second statistic of the second interval, the first influence is less than or equal to the second influence; or if the difference between the lower limit of the first interval and the lower limit of the second interval is less than a first threshold, the upper limit of the first interval and the upper limit of the second interval are less than a second threshold, and the difference between the first statistic and the second statistic is less than a third threshold, the first influence is equal to the second influence.

[0127] In some embodiments, the computing device includes circuitry configured to perform the following operation: determining the change over time in the influence of a first attribute parameter among a plurality of attribute parameters on a target attribute parameter.

[0128] In some embodiments, the computing device includes circuitry configured to perform an operation to obtain instructions for a first attribute parameter entered by a user.

[0129] In some embodiments, the computing device includes circuitry configured to perform the following operations: obtaining a first adjustment instruction for a second attribute parameter among a plurality of attribute parameters, and determining, based on the first adjustment instruction, the relationship between the influence of the second attribute parameter on a target attribute parameter and time.

[0130] In some embodiments, the indication of the relationship between the influence of the second attribute parameter on the target attribute parameter and time indicates at least one of the following: the change in the influence of the second attribute parameter on the target attribute parameter over time, or the cumulative change in the cumulative influence of the second attribute parameter on the target attribute parameter over time.

[0131] In some embodiments, the computing device includes circuitry configured to perform the following operation: determining whether the cumulative influence of a second attribute parameter on a target attribute parameter has reached a predetermined threshold over a predetermined period, based on the cumulative change in the cumulative influence of the second attribute parameter on the target attribute parameter over time.

[0132] In some embodiments, the computing device includes circuitry configured to perform the following operations: obtaining a second adjustment instruction for a third attribute parameter among a plurality of attribute parameters, and determining, based on the first and second adjustment instructions, which of the second and third attribute parameters has a relatively large influence on the target attribute parameter.

[0133] In some embodiments, the computing device includes circuitry configured to perform an operation to acquire a first adjustment instruction entered by a user.

[0134] In some embodiments, the computing device includes circuitry configured to perform the following operations: obtaining an initial time series model and obtaining a time series model by performing a test on the initial time series model to remove redundant edges.

[0135] In some embodiments, the computing device includes circuitry configured to perform the following operation: selecting a time series model from among several candidate models based on the relationship between a time series dataset and a given dataset.

[0136] In some embodiments, the computing device includes circuitry configured to perform the following operations: acquiring an initial time-series dataset and performing at least one of the following operations on the initial time-series dataset in order to obtain the time-series dataset: removing noise data items or generating augmented data items.

[0137] In some embodiments, the computing device includes circuitry configured to perform the following operation: inputting a first data item and a second data item of an initial time-series dataset into a data generation model to obtain an extended data item. The first data item has a first time, the second data item has a second time, and the extended data item has a third time.

[0138] In some embodiments, the computing device includes circuitry configured to perform the following operations: acquiring a time-series dataset relating to sales records, wherein the time-series dataset comprises a plurality of time-series data items, each time-series data item comprising a time and a plurality of corresponding attribute parameters, the plurality of attribute parameters comprising at least one of the following: purchase price, selling price, sales volume, inventory quantity, or number of customer views; determining at least one influencing factor on the inventory quantity based on a time-series model, wherein the at least one influencing factor indicates at least one attribute parameter that affects the inventory quantity and at least one time corresponding to the at least one attribute parameter; and determining the inventory quantity for the next time based on the at least one influencing factor.

[0139] In some embodiments, the computing device includes circuitry configured to perform the following operations: acquiring a time-series dataset of power for an IoT device, wherein the time-series dataset comprises a plurality of time-series data items, each time-series data item comprising time and a plurality of corresponding attribute parameters, the plurality of attribute parameters comprising at least one of electricity rates, power demand, voltage, current, temperature, humidity, atmospheric pressure, or the power consumption of the IoT device; determining at least one influencing factor of power consumption based on a time-series model, wherein the at least one influencing factor indicates at least one attribute parameter that affects power consumption and at least one time corresponding to the at least one attribute parameter; and determining the uptime of the IoT device based on the at least one influencing factor.

[0140] In some embodiments, the computing device includes circuitry configured to perform the following operations: acquiring a time-series dataset collected from a social network, wherein the time-series dataset comprises a plurality of time-series data items, each time-series data item comprising a time and a plurality of corresponding attribute parameters, the plurality of attribute parameters comprising at least one of user identification information, user posts, number of views of user posts, number of comments on user posts, or anomaly indicators; determining at least one influencing element of an anomaly indicator based on a time-series model, wherein the at least one influencing element indicates at least one attribute parameter that affects a target attribute parameter and at least one time corresponding to the at least one attribute parameter; and outputting presentation information indicating users with anomaly risk based on the at least one influencing element.

[0141] Figure 19 shows a schematic block diagram of an exemplary device 1900 that can implement embodiments of the present disclosure. For example, the computing device 110 shown in Figure 1 can be implemented by device 1900. As shown in Figure 19, device 1900 includes a central processing unit (CPU) 1901. The CPU 1901 may perform various appropriate operations and processes based on computer program instructions stored in read-only memory (ROM) 1902 or computer program instructions loaded from memory unit 1908 into random access memory (RAM) 1903. The RAM 1903 may further store various programs and data necessary for the operation of device 1900. The CPU 1901, ROM 1902, and RAM 1903 are connected to each other via a bus 1904. An input / output (I / O) interface 1905 is also connected to the bus 1904.

[0142] Multiple components of device 1900 are connected to the I / O interface 1905. These components include input units 1906 such as a keyboard and mouse, output units 1907 such as various types of displays and speakers, storage units 1908 such as magnetic disks and optical disks, and communication units 1909 such as a network interface card, modem, and wireless communication transceiver. The communication unit 1909 allows device 1900 to exchange information / data with other devices via computer networks such as the Internet and / or various telegraph networks. It should be understood that, in this disclosure, the output unit 1907 may be used to display information regarding real-time dynamic changes in user satisfaction, key element identification information for user groups or individual users regarding satisfaction, information regarding optimization policies, and information regarding the evaluation of the effectiveness of policy implementation.

[0143] The processor unit 1901 can be implemented by one or more processing circuits. The processor unit 1901 may be configured to execute each of the processes and operations described above. For example, in some embodiments, the aforementioned processes can be implemented as computer software programs and stored in tangible storage in a machine-readable medium such as the storage unit 1908. In some embodiments, part or all of the computer program may be loaded and / or installed into the device 1900 via the ROM 1902 and / or the communication unit 1909. Once the computer program is loaded into the RAM 1903 and executed by the CPU 1901, one or more steps of the processes described above can be executed.

[0144] This disclosure may be a system, method, and / or a computer program product. A computer program product may include a computer-readable storage medium containing computer-readable program instructions for performing each aspect of this disclosure.

[0145] A computer-readable storage medium can be a tangible device capable of holding and storing instructions used by an instruction execution device. Computer-readable storage media may be, but are not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples of computer-readable storage media (though not all) include portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable and writable read-only memory (EPROM or flash memory), static RAM (SRAM), portable compact disk read-only memory (CD-ROM), digital multipurpose disks (DVDs), memory sticks, floppy disks, mechanical encoder disks, punch cards or grooved projection structures on which instructions are stored, and any suitable combination thereof. The computer-readable storage medium used herein is not to be interpreted as an instantaneous signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating via guides or other transmitting media (e.g., optical pulses via optical cables), or electrical signals transmitted via wires.

[0146] The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to each computing / processing device, or they may be downloaded to an external computer or external storage device via a network such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include copper transmission cables, optical cable transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. A network interface card or network interface in each computing / processing device receives computer-readable program instructions from the network, transfers the computer-readable program instructions, and stores them in the computer-readable storage medium of each computing / processing device.

[0147] Computer program instructions for performing the operations disclosed herein may be assembler directives, instruction set architectures (ISAs), machine language instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or target code written in any combination of one or more programming languages. Programming languages ​​include object-oriented programming languages ​​such as Smalltalk and C++, and general process-oriented programming languages ​​such as "C" or similar programming languages. Computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a single independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it may be connected to an external computer (for example, by using an internet service provider and connecting via the internet). In some embodiments, status information of computer-readable program instructions is used to personalize electronic circuits such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs). These electronic circuits may implement each embodiment of the disclosure by executing computer-readable program instructions.

[0148] Herein, each aspect of the present disclosure has been described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products relating to embodiments of the present disclosure. It should be understood that each block in a flowchart and / or block diagram, and each combination of blocks in a flowchart and / or block diagram, may be implemented by computer-readable program instructions.

[0149] These computer-readable program instructions may be provided to the processor unit of a general-purpose computer, a dedicated computer, or other programmable data processing device to generate a machine, which, when executed by the processor unit of the computer or other programmable data processing device, generates a device that performs the functions / operations defined in one or more blocks of a flowchart and / or block diagram. These computer-readable program instructions may be stored on a computer-readable storage medium. These instructions enable computers, programmable data processing devices, and / or other devices to operate in a specific manner. Therefore, a computer-readable medium containing instructions may include a product that contains instructions for each mode of performing the functions / operations defined in one or more blocks of a flowchart and / or block diagram.

[0150] Computer-readable program instructions may be loaded into a computer, other programmable data processing device, or other device to execute a series of operational steps on the computer, other programmable data processing device, or other device, thereby generating a process that the computer implements. In this way, the instructions executed by the computer, other programmable data processing device, or other device implement the functions / operations defined in one or more blocks of a flowchart and / or block diagram.

[0151] The flowcharts and block diagrams in the figures represent the implementable architectures, functions, and operations of systems, methods, and computer program products relating to multiple embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of an instruction, which includes one or more executable instructions for implementing a defined logic function. In some alternative implementations, the functions represented within a block may occur in a different order than those shown in the figures. For example, two consecutive blocks may actually be executed essentially in parallel, or in some cases in the opposite order. This depends on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, may be implemented by a dedicated hardware-based system that performs the defined function or operation, or by a combination of dedicated hardware and computer instructions.

[0152] While embodiments of this disclosure have been described above, the above descriptions are illustrative, not exhaustive, and are not limited to the embodiments disclosed. It will be apparent that a person skilled in the art can make several modifications and changes without departing from the scope and spirit of the embodiments described. The terms used herein have been selected to best describe the principles, practical applications, or market improvements of each embodiment, or to enable a person skilled in the art to understand the embodiments disclosed herein.

Claims

1. A data processing method that performs processing on a time series dataset containing multiple time series data items using a computer, Obtaining the aforementioned time series dataset, wherein each of the plurality of time series data items includes a corresponding plurality of attribute parameters at each time point. The aforementioned time-series dataset relates to sales records, and the aforementioned multiple attribute parameters include at least one of the following: purchase price, selling price, sales volume, inventory volume, sales profit, and number of customer views. The aforementioned time-series dataset relates to the power consumption of the IoT device, and the multiple attribute parameters include at least one of the following: electricity cost, power demand, voltage, current, temperature, humidity, atmospheric pressure, or the power consumption of the IoT device, or The aforementioned time-series dataset is collected from social networks, and the multiple attribute parameters include at least one of the following: user identification information, the user's posts, the number of views of the user's posts, the number of comments on the user's posts, or an anomaly indicator. The process involves obtaining a target attribute parameter from the aforementioned plurality of attribute parameters, wherein the target attribute parameter is indicated by the user's input. The method involves determining a time series model that shows the variation of the multiple attribute parameters in a time series, based on the aforementioned time series dataset, wherein the time series model is pre-trained or pre-constructed such that it is a model in which at least one influencing element, which indicates at least one time corresponding to the at least one attribute parameter, and the variation of the target attribute parameter are associated. The method for determining the at least one influencing element is to input the time-series data items of the target attribute parameter into the time-series model, wherein the at least one time is either or both of the following: a time corresponding to the at least one attribute parameter that has a time lag in influencing the variation of the target attribute parameter, which is earlier than the time corresponding to the variation of the target attribute parameter; and a time corresponding to the at least one attribute parameter that has an instantaneous influence on the variation of the target attribute parameter, which is the same time as the time corresponding to the variation of the target attribute parameter. Outputting at least one of the aforementioned influencing elements, including, Data processing method.

2. Instead of determining the aforementioned time series model, Obtain an initial time series model represented in the form of a time relationship diagram including nodes and edges, The process involves performing a t-test on each edge of the initial time series model to obtain a p-value, and then removing redundant edges based on the p-value to obtain the time series model. Further including, The method according to claim 1.

3. Instead of determining the aforementioned time series model, Obtaining multiple predetermined datasets similar to the aforementioned time-series dataset, The further step includes selecting a candidate model from among multiple candidate models that corresponds to the predetermined dataset with the highest similarity, based on the similarity between the time series dataset and each of the multiple predetermined datasets, as the time series model. The method according to claim 1.

4. Obtaining the aforementioned time-series dataset is, Obtaining an initial time series dataset, When multiple data items in the initial time series dataset are not in a uniform time series, in order to obtain a time series dataset consisting of a uniform time series, the initial time series dataset is expanded to obtain additional data items to be added to the initial time series dataset. including, The method according to claim 1.

5. The generation of the aforementioned extended data items is, This includes inputting a first data item having a first time interval and a second data item having a second time interval from the initial time series dataset into a data generation model to obtain an extended data item having a third time interval that is later than the first and second time intervals, The first time, the second time, and the third time are shown in that order in a time series. The method according to claim 4.

6. A data processing method that performs processing related to retail demand forecasting by computer based on a time-series dataset of sales records that includes multiple time-series data items, Obtaining the aforementioned time-series dataset, wherein each of the multiple time-series data items includes a corresponding set of attribute parameters for each time period, and each of the multiple attribute parameters includes at least one of the following: purchase price, selling price, sales volume, inventory volume, or number of customer views. Based on the aforementioned time series dataset, a time series model is determined that shows the variation of the plurality of attribute parameters in the time series, wherein the time series model associates at least one attribute parameter that affects the variation in the inventory quantity, at least one time corresponding to the at least one attribute parameter, at least one influencing element, and the variation in the inventory quantity, and is obtained by pre-training or constructed in advance. The method involves determining the at least one influencing element by inputting the inventory quantity into the time series model, wherein the at least one time is either or both of the following: a time corresponding to the at least one attribute parameter that has a time lag in influencing the inventory quantity fluctuation, which is earlier than the time corresponding to the inventory quantity fluctuation; and a time corresponding to the at least one attribute parameter that has an instantaneous influence on the inventory quantity fluctuation, which is the same as the time corresponding to the inventory quantity fluctuation. Based on at least one of the aforementioned influencing factors, the inventory quantity for the next time period is determined, including, Data processing method.

7. A data processing method comprising: obtaining a time-series dataset of power consumption of an IoT device including multiple time-series data items; and performing a process by computer to determine the operating time of the IoT device, The acquisition of the aforementioned time-series dataset, wherein each of the plurality of time-series data items includes a corresponding plurality of attribute parameters at each time, and the plurality of attribute parameters include at least one of electricity charges, power demand, voltage, current, temperature, humidity, atmospheric pressure, or the power consumption of the IoT device. Based on the aforementioned time series dataset, a time series model is determined that shows the variation of the plurality of attribute parameters in the time series, wherein the time series model associates at least one attribute parameter that affects the variation in power consumption, at least one time corresponding to the at least one attribute parameter, at least one influencing element, and the variation in power consumption, and is obtained by pre-training or constructed in advance. The power consumption is input into the time series model to determine the at least one influencing element, wherein the at least one time is either or both of the following: a time corresponding to the at least one attribute parameter that has a time lag in influencing the fluctuation of power consumption, which is earlier than the time corresponding to the fluctuation of power consumption; and a time corresponding to the at least one attribute parameter that has an instantaneous influence on the fluctuation of power consumption, which is the same as the time corresponding to the fluctuation of power consumption. The operating time of the IoT device is determined based on the aforementioned at least one influencing factor, including, Data processing method.

8. A data processing method that performs processing on a time-series dataset collected from a social network containing multiple time-series data items, using a computer, Obtaining the aforementioned time-series dataset, wherein each of the plurality of time-series data items includes a corresponding plurality of attribute parameters for each time period, and the plurality of attribute parameters include at least one of user identification information, the user's posts, the number of views of the user's posts, the number of comments on the user's posts, or an anomaly indicator. Based on the aforementioned time series dataset, a time series model is determined that shows the variation of the plurality of attribute parameters in the time series, wherein the time series model associates at least one attribute parameter that affects the variation of the abnormal indication, at least one time corresponding to the at least one attribute parameter, at least one influencing element, and the variation of the abnormal indication, and is obtained by pre-training or constructed in advance. The method involves inputting the abnormality indication into the time series model to determine the at least one influencing element, wherein the at least one time is either or both of the following: a time corresponding to the at least one attribute parameter that has a time lag in influencing the variation of the abnormality indication, which is earlier than the time corresponding to the variation of the abnormality indication; and a time corresponding to the at least one attribute parameter that has an instantaneous influence on the variation of the abnormality indication, which is the same as the time corresponding to the variation of the abnormality indication. Based on the aforementioned at least one influencing factor, the system outputs information indicating users who are at risk of abnormality. including, Data processing method.

9. A processing circuit configured to perform the method described in any one of claims 1 to 8, electronic equipment.

Images