Data monitoring method, device, equipment and medium

Abstract

The invention provides a data monitoring method, which comprises the following steps of: for monitored data processed in batches, acquiring a true value of a data feature of the monitored data, and predicting a prediction result of the data feature in a current period through a time sequence prediction model according to the true value of the data feature in a historical period; and according to the prediction result of the data feature in the current period and the threshold value of the data feature, determining an early warning threshold value of the data feature, and when the true value of the data feature in the current period is greater than or equal to a first early warning threshold value or the true value of the data feature in the current period is less than or equal to a second early warning threshold value, generating early warning information. According to the method, historical data learning and current data result prediction are realized by using the time sequence prediction model, and the threshold value can be adaptively set for the prediction result according to the tolerance degree of different data processing results, so that the processing result prediction function is realized, and the accuracy of data feature monitoring is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data monitoring method, device, equipment and computer-readable storage medium. Background Art

[0002] Batch data processing refers to the processing of data by the application system according to certain business rules after obtaining data files from the upstream data system. For example, if a bank needs to know the total amount of loans for a year, it can batch summarize the loan amounts of the head office, branches or sub-branches of the bank. For another example, if the regulatory agency needs to know the bank's customers, deposits, etc., summarized by different provinces, it needs to batch calculate the indicator values ​​of the corresponding indicators by province.

[0003] When processing data in batches, it is usually necessary to monitor certain data features of the processed data results to ensure the rationality and correctness of the data processing. In general, the normality of the processing results of the current data is determined by analyzing the changes in the current data compared with the historical data. If the correctness of the current data features is determined manually, it is difficult to manage the features of multiple periods of data and make a judgment on the correctness of the current data features. Summary of the invention

[0004] The present application provides a data monitoring method, which introduces a time series prediction model to predict the value of a data feature in the current period, and determines the warning threshold value of the data feature based on the prediction result and the threshold value of the data feature. The correctness of the data feature can be automatically judged based on the real value of the data feature in the current period and the warning threshold value, and an automatic alarm can be issued based on the judgment result. The present application also provides a device, a computing device, a computer-readable storage medium, and a computer program product for the data monitoring method.

[0005] In a first aspect, the present application provides a data monitoring method. The method comprises:

[0006] For the monitored data processed in batches, obtaining the real value of the data feature of the monitored data, wherein the real value of the data feature includes the real value of the data feature in the historical period and the real value of the data feature in the current period;

[0007] According to the real value of the data feature in the historical period, the prediction result of the data feature in the current period is predicted by the time series prediction model;

[0008] Determine a warning threshold value of the data feature according to a prediction result of the data feature in a current cycle and a threshold value of the data feature, wherein the warning threshold value includes at least one of a first warning threshold value or a second warning threshold value;

[0009] When the true value of the data feature in the current period is greater than or equal to the first warning threshold value, or the true value of the data feature in the current period is less than or equal to the second warning threshold value, warning information is generated.

[0010] In some possible implementations, the method further includes:

[0011] Receiving user feedback on the warning information;

[0012] The threshold of the data feature is adjusted according to the user's feedback on the alarm information.

[0013] In some possible implementations, adjusting the threshold of the data feature according to the user's feedback on the warning information includes:

[0014] When the number of false alarms indicated by the feedback reaches a first number, or the proportion of false alarms indicated by the feedback reaches a first proportion, the first warning threshold value is increased, and / or the second warning threshold value is decreased.

[0015] In some possible implementations, the method further includes:

[0016] The data features indicating a false alarm in the feedback are labeled as normal to obtain labeled data, and the labeled data is used to train or update the time series prediction model.

[0017] In some possible implementations, the prediction result of the data feature in the current period is determined by the time series prediction model according to the predicted trend quantity, period component and holiday component.

[0018] In some possible implementations, the method further includes:

[0019] When the number of missed alarms reaches a second number, or the proportion of missed alarms reaches a second proportion, the first warning threshold value is reduced, and / or the second warning threshold value is increased.

[0020] In some possible implementations, the method further includes:

[0021] Constructing a data feature monitoring asset certificate, wherein the data feature monitoring asset certificate integrates at least one of a data feature identifier, a category, a monitoring frequency, an offset, a data acquisition method, a responsible person or a contact person;

[0022] The obtaining of the real value of the data feature of the monitored data includes:

[0023] According to the method for obtaining the data feature, the real value of the data feature of the monitored data is obtained.

[0024] In some possible implementations, the time series prediction model includes a prophet model, a long short-term memory model, or an autoregressive integrated moving average model.

[0025] In some possible implementations, the data characteristics include: at least one of the current data volume of the report, the null value rate of key fields, the summary value of the amount field, and the data quality monitoring value.

[0026] In a second aspect, the present application provides a data monitoring device. The device comprises:

[0027] An acquisition module, used for acquiring, for batch-processed monitored data, real values ​​of data features of the monitored data, wherein the real values ​​of the data features include real values ​​of the data features in historical periods and real values ​​of the data features in a current period;

[0028] A prediction module, used to predict the prediction result of the data feature in the current period through a time series prediction model according to the real value of the data feature in the historical period;

[0029] A determination module, configured to determine a warning threshold value of the data feature according to a prediction result of the data feature in a current cycle and a threshold value of the data feature, wherein the warning threshold value includes at least one of a first warning threshold value or a second warning threshold value;

[0030] The alarm module is used to generate warning information when the true value of the data feature in the current cycle is greater than or equal to the first warning threshold value, or the true value of the data feature in the current cycle is less than or equal to the second warning threshold value.

[0031] In some possible implementations, the device further includes:

[0032] A receiving module, used to receive user feedback on the alarm information;

[0033] An adjustment module is used to adjust the threshold of the data feature according to the user's feedback on the alarm information.

[0034] In some possible implementations, the adjustment module is specifically configured to:

[0035] When the number of false alarms indicated by the feedback reaches a first number, or the proportion of false alarms indicated by the feedback reaches a first proportion, the first warning threshold value is increased, and / or the second warning threshold value is decreased.

[0036] In some possible implementations, the device further includes:

[0037] The labeling module is specifically used to label the data features indicating false alarms in the feedback as normal, and obtain labeling data, and the labeling data is used to train or update the time series prediction model.

[0038] In some possible implementations, the prediction result of the data feature in the current period is determined by the time series prediction model according to the predicted trend quantity, period component and holiday component.

[0039] In some possible implementations, the device further includes:

[0040] The adjustment module is specifically used to reduce the first warning threshold value and / or increase the second warning threshold value when the number of missed alarms reaches a second number or the proportion of missed alarms reaches a second proportion.

[0041] In some possible implementations, the device further includes:

[0042] A construction module is specifically used to construct a data feature monitoring asset certificate, wherein the data feature monitoring asset certificate integrates at least one of a data feature identifier, a category, a monitoring frequency, an offset, a data acquisition method, a responsible person or a contact person;

[0043] The acquisition module is specifically used for:

[0044] According to the method for obtaining the data feature, the real value of the data feature of the monitored data is obtained.

[0045] In some possible implementations, the time series prediction model includes a prophet model, a long short-term memory model, or an autoregressive integrated moving average model.

[0046] In some possible implementations, the data characteristics include: at least one of the current data volume of the report, the null value rate of key fields, the summary value of the amount field, and the data quality monitoring value.

[0047] In a third aspect, the present application provides a computing device, the computing device comprising a processor and a memory. The processor and the memory communicate with each other. The processor is used to execute instructions stored in the memory, so that the computing device performs the data monitoring method in the first aspect or any implementation of the first aspect.

[0048] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, wherein the instructions instruct a computing device to execute the data monitoring method described in the first aspect or any one of the implementations of the first aspect.

[0049] In a fifth aspect, the present application provides a computer program product comprising instructions, which, when executed on a computing device, enables the computing device to execute the data monitoring method described in the first aspect or any one of the implementations of the first aspect.

[0050] Based on the implementations provided in the above aspects, the present application can also be further combined to provide more implementations.

[0051] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0052] An embodiment of the present application provides a data monitoring method, which provides a data feature situation prediction function based on a time series prediction model, utilizes the time series prediction model to learn historical data and predict current data results, and can adaptively set thresholds for prediction results based on the tolerance level for different data processing results, thereby realizing the function of processing result prediction. This method can take into account the characteristics of multiple periods of data, thereby improving the accuracy of data feature monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical methods of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 A flowchart of a data monitoring method provided in an embodiment of the present application;

[0055] Figure 2 A schematic diagram of an intelligent algorithm prediction system and a data monitoring and early warning system provided in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of the structure of a data monitoring device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The scheme in the embodiments provided in this application will be described below in conjunction with the drawings in this application.

[0058] The terms "first" and "second" in the embodiments of the present application are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0059] First, some technical terms involved in the embodiments of the present application are introduced.

[0060] Batch data processing refers to the processing of data by the application system according to certain business rules after obtaining data files from the upstream data system. Typical application scenarios may include the People's Bank of China needs the number of loans from Bank X in a year, so it needs to batch summarize the loans of all customers of Bank X, or the regulatory agency needs to know the customers and deposits of Bank Y, and needs to summarize the customers and deposits of Bank Y by different provinces.

[0061] Time series prediction algorithms are a technology used to predict future values ​​based on given time series data. They are a type of algorithm that predicts future data by fitting historical data. Time series prediction algorithms can include the Prophet algorithm, the Auto Regressive Integrated Moving Average (ARIMA) algorithm, or the Long Short-Term Memory (LSTM) algorithm.

[0062] The Prophet algorithm is a method based on time series analysis, used to predict future trends. It is based on ARIMA and seasonal decomposition models and can handle non-stationary time series data. Unlike the exponential balance model, which targets trends and seasonality in the data, ARIMA aims to depict the autocorrelations of the data. In ARIMA (p, d, q), AR stands for autoregression, p stands for the number of autoregression terms; MA stands for moving average, q stands for the number of sliding average terms, and d stands for the number of differences (order) made to make it a stationary series. Among them, differentiation is the key to making the time series stationary.

[0063] In batch data processing, it is usually necessary to monitor certain characteristics of the processed data results to ensure the rationality and correctness of data processing, such as monitoring the summary value of the amount field, the field null value rate of the data table, the data volume of the data table, etc. In general, the processing results of the current data are judged to be normal by analyzing the changes in the current data compared with the historical data.

[0064] However, the characteristics of the results of data processing are easily affected by the quality of upstream data and business fluctuations. These influences may cause fluctuations in the growth of the characteristics of current data monitoring, affecting the changes of current data compared with historical data, thereby increasing the difficulty of judging the correctness of the current data processing results. If the correctness of the current data characteristics is judged manually, it is difficult to manage the characteristics of multiple periods of data and make it difficult to judge the correctness of the current data characteristics. And because the amount of processed data is large, if manual methods are used, it will also bring about high labor costs.

[0065] In view of this, the present application provides a data monitoring method. The method can be executed by a computing device. The computing device can be a server or a terminal. Among them, the server can be a physical server or a cloud server. The terminal can be a desktop, a laptop or a mobile terminal. The present application does not limit this.

[0066] Specifically, the computing device can obtain the true value of the data feature of the monitored data for batch processing. The true value of the data feature includes the true value of the data feature in the historical period and the true value of the data feature in the current period. The computing device can predict the prediction result of the data feature in the current period through the time series prediction model based on the true value of the data feature in the historical period. Then the computing device determines the warning threshold value of the data feature based on the prediction result of the data feature in the current period and the threshold value of the data feature. Among them, the warning threshold value includes at least one of the first warning threshold value or the second warning threshold value. The first warning threshold value can also be called the warning upper limit, and the second warning threshold value can also be called the warning lower limit. When the true value of the data feature in the current period is greater than or equal to the first warning threshold value, or the true value of the data feature in the current period is less than or equal to the second warning threshold value, a warning message is generated.

[0067] This method introduces a time series prediction model to predict the value of the data feature in the current period, and determines the warning threshold value of the data feature based on the prediction result and the threshold value of the data feature. Based on the actual value of the data feature in the current period and the warning threshold value, the correctness of the data feature can be automatically judged, the accuracy of data feature monitoring can be improved, and automatic alarms can be issued based on the judgment results.

[0068] In order to make the technical solution of the present application clearer and easier to understand, the data monitoring method of the present application is introduced below in conjunction with the accompanying drawings.

[0069] See also Figure 1 A flow chart of a data monitoring method is shown, which can be executed by a computing device, and the method includes the following steps:

[0070] S102: For the monitored data processed in batches, the computing device obtains the real value of the data feature of the monitored data.

[0071] The data feature may be a feature of a processing result in a batch processing data scenario, for example, a feature of a processing result of monitored data processed in batches. In some examples, the data feature includes at least one of the current report data volume, key field null value rate, amount field summary value, and data quality monitoring value.

[0072] Among them, data quality can be characterized by at least one evaluation indicator indicating quality. Data quality monitoring values ​​can be based on report feature technology. For example, a report records all customers of the bank and records basic information of customers, such as name, address, etc. However, due to historical reasons or other reasons, the customer address information input does not meet the requirements, then the compliance rate can be regarded as a data quality monitoring value.

[0073] The true value of the data feature includes the true value of the data feature in the historical period and the true value of the data feature in the current period. The historical period can be the n periods before the current period, where n can be set according to an empirical value and is greater than or equal to 1.

[0074] In some possible implementations, the computing device may also construct a data feature monitoring asset certificate. Among them, the data feature monitoring asset certificate integrates at least one of the data feature identification (or data feature identification code), category, monitoring frequency, offset, data acquisition method, responsible person or contact person. In some scenarios, the responsible person may also be called a stakeholder. In specific implementations, the data feature monitoring asset certificate may use a record table, which records the basic information required for data feature detection, such as data feature identification code, category, monitoring frequency, offset, data acquisition method, and stakeholders. Accordingly, the computing device can obtain the true value of the data feature of the monitored data according to the data feature acquisition method.

[0075] S104. The computing device predicts the prediction result of the data feature in the current period through a time series prediction model according to the real value of the data feature in the historical period.

[0076] The time series prediction model takes time series data (referred to as time series data) as input and time series data as output. The input time series data includes the real value of the data feature in the historical period, and the output time series data includes the prediction result of the data feature in the current period.

[0077] The time series prediction model may include a Prophet model, a long short-term memory model, or an autoregressive integrated moving average model. For ease of description, the time series prediction model is taken as an example of a Prophet model. The prediction result of the data feature in the current period can be determined by the time series prediction model, such as the Prophet model, based on the predicted trend quantity, period component, and holiday component.

[0078] The Prophet model uses the superposition of three forecast quantities: trend quantity, periodic component, and holiday component, taking into account the fact that data features present different data trends in different periods, so as to more accurately predict the forecast results of data features in the current period. Among them, the forecast results can include at least one of the forecast upper limit or the forecast lower limit.

[0079] S106. The computing device determines the warning threshold value of the data feature according to the prediction result of the data feature in the current cycle and the threshold value of the data feature. When the true value of the data feature in the current cycle is greater than or equal to the first warning threshold value, or the true value of the data feature in the current cycle is less than or equal to the second warning threshold value, execute S108.

[0080] The warning threshold value includes at least one of the first warning threshold value or the second warning threshold value. The first warning threshold value may be an upper warning limit, and the second warning threshold value may be a lower warning limit. Accordingly, the computing device may determine the first warning threshold value according to the prediction upper limit and the threshold value of the data feature, and determine the second warning threshold value according to the prediction lower limit and the threshold value of the data feature. It should be noted that the threshold value of the data feature may include a threshold value for determining the first warning threshold value and a threshold value for determining the second warning threshold value, which are respectively recorded as threshold 1 and threshold 2.

[0081] In some possible implementations, the computing device may determine the first warning threshold value and the second warning threshold value by using the following formula:

[0082] The first warning threshold value = the prediction upper limit of the time series prediction model * (1 + threshold 1).

[0083] The second warning threshold value = the prediction lower limit of the time series prediction model * (1-threshold 2).

[0084] S108. The computing device generates warning information.

[0085] When the true value of the data feature in the current period is greater than or equal to the first warning threshold value, or the true value of the data feature in the current period is less than or equal to the second warning threshold value, it means that the true value of the data feature in the current period is not within the prediction range, and the data feature may have problems or abnormalities. For this reason, the computing device can generate warning information.

[0086] If early warning information notification is required, the computing device can notify the early warning information in the form of email or text message. Among them, the computing device can monitor the asset certificate according to the data characteristics, obtain the responsible person or interested person, and send an email or text message to the responsible person or interested person to notify the early warning information.

[0087] In some possible implementations, the computing device may also receive user feedback on the warning information, and then adjust the threshold of the data feature according to the user feedback on the warning information, so that the threshold of the data feature can be adjusted adaptively.

[0088] Specifically, when the number of false alarms indicated by the feedback reaches a first number, or the proportion of false alarms indicated by the feedback reaches a first proportion, the computing device may increase the first warning threshold value, and / or decrease the second warning threshold value.

[0089] In addition, when the true value of the data feature in the current cycle is greater than the second warning threshold value, and the true value of the data feature in the current cycle is less than the first warning threshold value, it means that the true value of the data feature in the current cycle is within the predicted range, and the computing device can also display the monitoring results to the user (such as a stakeholder or a responsible person), and the monitoring results indicate that the data feature is normal. Furthermore, the computing device can also receive feedback from the user on the monitoring results, and the feedback may include missed alarms. For example, if a stakeholder confirms that there is an abnormality in the data feature of the current cycle, it means a missed alarm. When the number of missed alarms reaches the second number, or the proportion of missed alarms reaches the second proportion, the computing device can reduce the first warning threshold value, and / or increase the second warning threshold value.

[0090] In some possible implementations, the computing device may also annotate the monitored data of the current period based on the user's feedback. Specifically, the computing device may annotate the data features of the feedback indicating a false alarm as normal, and obtain annotated data, which is used to train or update the time series prediction model. Similarly, the computing device may annotate the data features of the feedback indicating a missed alarm as abnormal, and obtain annotated data.

[0091] Based on the above description, the present application provides a data monitoring method. The method introduces a time series prediction model to predict the value of the data feature in the current period, and determines the warning threshold value of the data feature based on the prediction result and the threshold value of the data feature. Based on the true value of the data feature in the current period and the warning threshold value, the correctness of the data feature can be automatically judged, the accuracy of data feature monitoring can be improved, and automatic alarm can be issued based on the judgment result.

[0092] Furthermore, the method determines the warning range by customizing the threshold value based on the time series prediction model. Specifically, after each warning message is issued, the business personnel can provide feedback on the warning result based on the difference between the current warning range and the true value. If the user believes that the warning range is too small, the threshold value can be adaptively adjusted according to the difference range to ensure that the threshold value can meet the needs of the business personnel in subsequent forecasts. The adaptive threshold change scheme of the present application does not require the user to set and adjust the threshold value, reducing the burden on the user.

[0093] It should be noted that a key to data monitoring in this application is the time series prediction model. The following takes the Prophet model (or Prophet algorithm model) as an example of the time series prediction model to explain in detail the artificial intelligence (AI) algorithm prediction system (or intelligent algorithm prediction system for short) and data feature monitoring system based on the Prophet algorithm model.

[0094] The present invention constructs a data monitoring system by establishing data feature monitoring credentials, an intelligent algorithm prediction system, and a quasi-real-time data feature monitoring and early warning system to solve the problem that data report processing results are difficult to monitor.

[0095] Among them, data feature monitoring vouchers are established, and a data feature monitoring asset information display system is constructed to achieve one-stop configuration of data feature monitoring. Relying on data feature monitoring vouchers, integrating data feature identification codes, categories, monitoring frequencies, offsets, data acquisition methods, stakeholders and other asset information, it can provide base information for intelligent algorithm prediction, data feature monitoring and early warning, and construct data feature monitoring base characteristics. For example, the AI ​​algorithm prediction system focuses on data features, such as the current data volume of the report, the null value rate of key fields, the summary value of the amount field, the data quality monitoring value, and other data features that reflect the processing results of the report data are registered in the data feature monitoring voucher as the basis for subsequent early warnings.

[0096] Establishing an AI intelligent algorithm prediction system to predict data features can provide a basis for data monitoring and early warning. In specific implementation, use automatic labeling tools to label data features, automatically remove abnormal extreme value data, select appropriate data for model training, and improve the accuracy of data prediction. Establish a data feature early warning mechanism based on the Prophet algorithm. The algorithm superimposes three prediction quantities: trend, seasonality, and holidays, and fully considers the characteristics of data features that present different data trends in different periods, so as to more accurately predict the prediction information of current data. In addition, the algorithm can adjust the upper and lower limits of the prediction results according to the credibility, providing a basis for subsequent early warnings. The adaptive threshold setting mechanism can avoid the work of manually determining the threshold.

[0097] Establishing a data monitoring and early warning system can realize the function of quickly notifying stakeholders of data monitoring and early warning information. First, obtain the real data (real value) of the data feature according to the data acquisition method in the data feature monitoring certificate. Then, determine whether an early warning notification is currently required based on the upper and lower limits of the prediction results of the AI ​​intelligent algorithm prediction system. Finally, if an early warning notification is required, obtain the corresponding stakeholders from the data feature monitoring certificate, and notify each stakeholder of the early warning information by email or text message to achieve timely delivery of the early warning information.

[0098] The AI ​​intelligent algorithm prediction system and data monitoring and early warning system constitute the process of data feature early warning information notification, see Figure 2 Among them, the AI ​​intelligent algorithm prediction system includes algorithm training function and algorithm prediction function, and the data monitoring and early warning system includes early warning information notification function. The specific implementation process is as follows:

[0099] The first is the algorithm training function. The main function of this stage is to train the model parameters based on the current data and write the optimal model parameter records into the database. This stage is mainly divided into two steps. The first step is to label the current data. When the algorithm is trained for the first time, the computing device can label all the existing data as normal data. In the subsequent algorithm training, if the prediction result indicates that the real data is problematic data and is confirmed by stakeholders, the real data, i.e., the training data, can be labeled as abnormal; if the prediction result indicates that the data is normal data and is confirmed by stakeholders, the real data will be labeled as normal as subsequent training data. The second step is to select the parameters of the Prophet algorithm. The cost function is obtained by training different parameters. The parameters with the cost function less than a certain threshold and the smallest value are selected as the training parameters of the data features and written into the database.

[0100] The second is the algorithm prediction function. The function of this stage is to predict the data features of the day, and dynamically set the threshold according to the feedback of stakeholders on the historical prediction results, so as to obtain the upper and lower limits of the prediction results of this data feature. This stage can be mainly divided into two steps. The first step is to predict the Prophet algorithm results of the data features of the day. The process is to read the Prophet algorithm parameters during the algorithm training phase as the prediction model for this time, so as to predict the current data features; the second step is to determine the upper and lower limits of the final algorithm based on the threshold of each data feature. This step requires a default threshold for the first prediction. In subsequent predictions, the threshold will be dynamically adjusted according to the feedback of stakeholders on the current prediction. For example, when the actual result is not within the prediction range, and the stakeholders feedback that the actual data is not within the prediction range, but it is normal data, the computing device can adjust the threshold according to the difference between the predicted value and the actual value, so as to achieve the effect of automatically adjusting the threshold.

[0101] The third is the function of early warning notification. In this stage, early warning information is generated based on the result of comparing the true value of the data feature with the upper and lower limits predicted by the algorithm, so as to determine whether early warning notification is needed. The first step is to obtain the predicted upper and lower limits of the data feature, and then compare the range with the real data obtained. If the true value is within the predicted upper and lower limits, it is considered that the true value of the current data feature is fine, and no early warning notification is required; if the true value is outside the predicted upper and lower limits, it is considered that there is a problem with the current data feature and early warning information notification is required. In the second step, if early warning information notification is required, the early warning information notification will be sent to stakeholders in the form of emails or text messages. Stakeholders can provide feedback on the results of this early warning based on the early warning information. After the early warning information feedback is completed, the algorithm training module can mark the real data according to the feedback.

[0102] The data feature monitoring voucher solution designed in this application can centralize all configurations in one place, which can effectively reduce the cost of use for users and can flexibly configure various configuration information. The adaptive threshold change solution designed in this application does not require users to set and adjust the threshold, reducing the burden on users. After each warning information is issued, the business personnel can give feedback on the warning results based on the difference between the current warning range and the true value. If the user believes that the warning range is too small, this method will make adaptive adjustments to the threshold according to the difference range to ensure that the threshold can meet the needs of business personnel in subsequent predictions. The warning mechanism solution designed in this application can complete the acquisition of current data in the first time according to the configured data acquisition method after data processing is completed, and can further complete the issuance of data warning information, and can quickly and effectively provide warning information to users.

[0103] Corresponding to the above method embodiment, the present application also provides a data monitoring device, see Figure 3 , the device 300 comprises:

[0104] An acquisition module 302 is used to acquire, for batch-processed monitored data, a true value of a data feature of the monitored data, wherein the true value of the data feature includes a true value of the data feature in a historical period and a true value of the data feature in a current period;

[0105] A prediction module 304 is used to predict the prediction result of the data feature in the current period through a time series prediction model according to the real value of the data feature in the historical period;

[0106] A determination module 306, configured to determine a warning threshold value of the data feature according to a prediction result of the data feature in a current cycle and a threshold value of the data feature, wherein the warning threshold value includes at least one of a first warning threshold value or a second warning threshold value;

[0107] The alarm module 308 is used to generate an alarm message when the real value of the data feature in the current cycle is greater than or equal to the first alarm threshold value, or the real value of the data feature in the current cycle is less than or equal to the second alarm threshold value.

[0108] In some possible implementations, the apparatus 300 further includes:

[0109] The receiving module 310 is used to receive user feedback on the alarm information;

[0110] The adjustment module 312 is used to adjust the threshold of the data feature according to the user's feedback on the alarm information.

[0111] In some possible implementations, the adjustment module 312 is specifically configured to:

[0112] When the number of false alarms indicated by the feedback reaches a first number, or the proportion of false alarms indicated by the feedback reaches a first proportion, the first warning threshold value is increased, and / or the second warning threshold value is decreased.

[0113] In some possible implementations, the apparatus 300 further includes:

[0114] The labeling module 314 is specifically used to label the data features indicating false alarms in the feedback as normal, and obtain labeled data, and the labeled data is used to train or update the time series prediction model.

[0115] In some possible implementations, the prediction result of the data feature in the current period is determined by the time series prediction model according to the predicted trend quantity, period component and holiday component.

[0116] In some possible implementations, the apparatus 300 further includes:

[0117] The adjustment module 312 is specifically configured to reduce the first warning threshold value and / or increase the second warning threshold value when the number of missed alarms reaches a second number or the ratio of missed alarms reaches a second ratio.

[0118] In some possible implementations, the apparatus 300 further includes:

[0119] A construction module 316 is specifically used to construct a data feature monitoring asset certificate, wherein the data feature monitoring asset certificate integrates at least one of a data feature identifier, a category, a monitoring frequency, an offset, a data acquisition method, a responsible person or a contact person;

[0120] The acquisition module 302 is specifically used for:

[0121] According to the method for obtaining the data feature, the real value of the data feature of the monitored data is obtained.

[0122] In some possible implementations, the time series prediction model includes a prophet model, a long short-term memory model, or an autoregressive integrated moving average model.

[0123] In some possible implementations, the data characteristics include: at least one of the current data volume of the report, the null value rate of key fields, the summary value of the amount field, and the data quality monitoring value.

[0124] The present application also provides a computing device for implementing a data monitoring method. The computing device includes a processor and a memory. The processor and the memory communicate with each other. The processor is used to execute instructions stored in the memory so that the computing device executes the data monitoring method.

[0125] The present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computing device, enable the computing device to execute the above-mentioned data monitoring method.

[0126] The present application provides a computer program product containing instructions, which, when executed on a computing device, enables the computing device to execute the above-mentioned data monitoring method.

[0127] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0128] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computing device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0129] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0130] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

Claims

1. A data monitoring method, characterized in that: The method comprises: For the monitored data processed in batches, obtaining the real value of the data feature of the monitored data, wherein the real value of the data feature includes the real value of the data feature in the historical period and the real value of the data feature in the current period; According to the real value of the data feature in the historical period, the prediction result of the data feature in the current period is predicted by the time series prediction model; Determine a warning threshold value of the data feature according to a prediction result of the data feature in a current cycle and a threshold value of the data feature, wherein the warning threshold value includes at least one of a first warning threshold value or a second warning threshold value; When the true value of the data feature in the current period is greater than or equal to the first warning threshold value, or the true value of the data feature in the current period is less than or equal to the second warning threshold value, warning information is generated.

2. The method according to claim 1, characterized in that The method further comprises: Receiving user feedback on the warning information; The threshold of the data feature is adjusted according to the user's feedback on the alarm information.

3. The method according to claim 2, characterized in that The adjusting the threshold of the data feature according to the user's feedback on the alarm information includes: When the number of false alarms indicated by the feedback reaches a first number, or the proportion of false alarms indicated by the feedback reaches a first proportion, the first warning threshold value is increased, and / or the second warning threshold value is decreased.

4. The method according to claim 2, characterized in that: The method further comprises: The data features indicating a false alarm in the feedback are labeled as normal to obtain labeled data, and the labeled data is used to train or update the time series prediction model.

5. The method according to any one of claims 1 to 4, characterized in that: The prediction result of the data feature in the current period is determined by the time series prediction model according to the predicted trend quantity, period component and holiday component.

6. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: When the number of missed alarms reaches a second number, or the proportion of missed alarms reaches a second proportion, the first warning threshold value is reduced, and / or the second warning threshold value is increased.

7. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Constructing a data feature monitoring asset certificate, wherein the data feature monitoring asset certificate integrates at least one of a data feature identifier, a category, a monitoring frequency, an offset, a data acquisition method, a responsible person or a contact person; The obtaining of the real value of the data feature of the monitored data includes: According to the method for obtaining the data feature, the real value of the data feature of the monitored data is obtained.

8. The method according to any one of claims 1 to 4, characterized in that: The time series prediction model includes a prophet model, a long short-term memory model or an autoregressive integrated moving average model.

9. The method according to any one of claims 1 to 4, characterized in that: The data characteristics include: at least one of the current data volume of the report, the null value rate of key fields, the summary value of the amount field, and the data quality monitoring value.

10. A data monitoring device, characterized in that: The device comprises: An acquisition module, used for acquiring, for batch-processed monitored data, real values ​​of data features of the monitored data, wherein the real values ​​of the data features include real values ​​of the data features in historical periods and real values ​​of the data features in a current period; A prediction module, used to predict the prediction result of the data feature in the current period through a time series prediction model according to the real value of the data feature in the historical period; A determination module, configured to determine a warning threshold value of the data feature according to a prediction result of the data feature in a current cycle and a threshold value of the data feature, wherein the warning threshold value includes at least one of a first warning threshold value or a second warning threshold value; The alarm module is used to generate warning information when the true value of the data feature in the current cycle is greater than or equal to the first warning threshold value, or the true value of the data feature in the current cycle is less than or equal to the second warning threshold value.

Images