WO2022252573A1 - 一种业务数据的监测方法及装置 - Google Patents

一种业务数据的监测方法及装置 Download PDF

Info

Publication number
WO2022252573A1
WO2022252573A1 PCT/CN2021/139697 CN2021139697W WO2022252573A1 WO 2022252573 A1 WO2022252573 A1 WO 2022252573A1 CN 2021139697 W CN2021139697 W CN 2021139697W WO 2022252573 A1 WO2022252573 A1 WO 2022252573A1
Authority
WO
WIPO (PCT)
Prior art keywords
business
business data
image
historical
pixel point
Prior art date
Application number
PCT/CN2021/139697
Other languages
English (en)
French (fr)
Inventor
伍日杰
王志远
Original Assignee
深圳前海微众银行股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳前海微众银行股份有限公司 filed Critical 深圳前海微众银行股份有限公司
Publication of WO2022252573A1 publication Critical patent/WO2022252573A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting

Definitions

  • the invention relates to the field of financial technology (Fintech), in particular to a method and device for monitoring business data.
  • the method of monitoring business data is generally to compare the business data of the current cycle with the business data of multiple historical cycles. For example, if the business data of the current cycle (May 8) is abnormal , it is necessary to compare the business data of the current cycle with the business data of multiple historical cycles (such as May 5, May 6, and May 7) before the current cycle. If the business data of the current cycle fluctuates (increase or drop) is within the threshold range, then it is determined that the business data of the current period is normal data.
  • the above-mentioned threshold range needs to be manually set and adjusted, and the threshold range for a cycle cannot be adapted to the whole period of the cycle, and business data cannot be dynamically monitored, and the accuracy is not high. For example, taking the day as the cycle unit, in a certain scenario, there are more business data from 8:00-17:00, less business data from 17:00-8:00, and only at 17:00 -During the time period of 18:00, the business data dropped sharply and exceeded the threshold range. At this time, the exception generated was wrong, but it was actually normal business data.
  • Embodiments of the present invention provide a service data monitoring method and device, which are used to implement dynamic monitoring of service data and improve the accuracy of service data monitoring.
  • an embodiment of the present invention provides a method for monitoring business data, including:
  • the historical identification image is trained; the historical monitoring result is determined according to the relationship between the first service data and the second service data in the historical identification image.
  • the monitoring threshold in the prior art is artificially set according to experience; while the convolutional neural network model in this application is trained according to historical recognition images; each historical recognition image includes the first business data (current business data) and the second business data. Two business data (historical business data), and have the label that is used to characterize historical monitoring result; Through the training of historical identification image to the first convolutional neural network model, make the first convolutional neural network model can identify the first business data and the relationship between the first business data and the second business data and determine whether the first business data is abnormal according to the relationship between the first business data and the second business data.
  • Each historical identification image in this application represents various historical periods, so the convolutional neural network model actually integrates business data of each historical period, and can be applied to full-time monitoring of business data. Therefore, dynamic monitoring of service data can be realized, and the accuracy of service data monitoring can be improved.
  • the business indicator is at least one of the following: business transaction volume, business average time-consuming and business success rate;
  • the historical period includes a chain period of the preset period and/or a year-on-year period of the preset period.
  • the business indicators are classified into different types, which improves the comprehensiveness of business data monitoring, because the chain period and year-on-year period of the preset time period are more relevant to the preset time period, and business data can be better compared. This increases the accuracy of business data monitoring.
  • generating an image to be recognized representing the first business data and the second business data includes:
  • the existing technical solution it is determined whether the business data is abnormal according to the difference between the current business data and the historical data, and if the difference is not verified whether it is abnormal; in this application, the first business data and the second The first difference between business data is intuitively displayed, and then it is determined whether the first difference is abnormal according to the second difference of the historical recognition image in the convolutional neural network model, and whether the first business data is abnormal according to whether the first difference is abnormal , which is equivalent to realizing two monitoring to increase the accuracy of business data monitoring.
  • preprocessing the graph to determine the image to be recognized includes:
  • Scaling the resolution of the picture to a preset resolution determining the pixel value of any pixel in the image to be identified according to the following formula (1), to obtain the image to be identified;
  • f(P) is the pixel value of any pixel point P in the image to be recognized
  • (x, y) is the coordinate value of the pixel point P
  • (x1, y1) is located at the bottom left of the pixel point P in the picture
  • (x1, y2) is the coordinate value of the adjacent pixel point Q12 located at the upper left corner of the pixel point P in the picture;
  • (x2, y1) is located at the pixel point in the picture
  • (x2, y2) is the coordinate value of the adjacent pixel point Q22 located in the upper right corner of pixel point P in the picture;
  • f(Q11) is the pixel value of pixel point Q11
  • f(Q12) is the pixel value of the pixel point Q12
  • f(Q21) is the pixel value of the pixel point Q21
  • the method further includes:
  • the comprehensive monitoring result of the business data in the preset time period is determined.
  • the detection results obtained by each business indicator are aggregated according to the preset weights to obtain comprehensive monitoring results, so as to prevent the business data of a certain business indicator from being abnormal, and the business data of other business indicators normally lead to wrong judgments.
  • Business data is abnormal, so as to increase the accuracy of business data monitoring.
  • the convolutional neural network model is used for N classification
  • the image to be identified is input to the convolutional neural network model to obtain a classification result
  • the monitoring results corresponding to the classification results are determined, wherein N is greater than M.
  • the comparison relationship between N classification and M type monitoring results is used to determine the classification result of the image to be recognized, so as to determine whether the image to be recognized is abnormal, wherein, the comparison relationship between N classification and M type monitoring results can be It is preset by the user based on experience to increase the flexibility of business data monitoring.
  • the convolutional neural network model is trained according to the image to be recognized and the corrected label corresponding to each false anomaly, and an updated convolutional neural network model is obtained.
  • the convolutional neural network model for the abnormal business data determined by the convolutional neural network model, if the user determines that the abnormal business data is normal, that is, the convolutional neural network model has misjudged, the abnormal business data will be recorded and recorded here.
  • the convolutional neural network model is optimized and trained according to the false and abnormal business data, so as to increase the recognition accuracy of the convolutional neural network model, thereby increasing the monitoring of business data. accuracy.
  • an embodiment of the present invention provides a device for monitoring service data, including:
  • An acquisition module configured to acquire first business data within a preset time period and second business data in a historical period associated with the preset time period; the first business data and the second business data are for the same business indicator;
  • a processing module configured to generate an image to be recognized representing the first business data and the second business data
  • the historical identification image is trained; the historical monitoring result is determined according to the relationship between the first service data and the second service data in the historical identification image.
  • the business indicator is at least one of the following: business transaction volume, business average time-consuming and business success rate;
  • the historical period includes a chain period of the preset period and/or a year-on-year period of the preset period.
  • processing module is specifically used for:
  • processing module is specifically used for:
  • Scaling the resolution of the picture to a preset resolution determining the pixel value of any pixel in the image to be identified according to the following formula (1), to obtain the image to be identified;
  • f(P) is the pixel value of any pixel point P in the image to be recognized
  • (x, y) is the coordinate value of the pixel point P
  • (x1, y1) is located at the bottom left of the pixel point P in the picture
  • (x1, y2) is the coordinate value of the adjacent pixel point Q12 located at the upper left corner of the pixel point P in the picture;
  • (x2, y1) is located at the pixel point in the picture
  • (x2, y2) is the coordinate value of the adjacent pixel point Q22 located in the upper right corner of pixel point P in the picture;
  • f(Q11) is the pixel value of pixel point Q11
  • f(Q12) is the pixel value of the pixel point Q12
  • f(Q21) is the pixel value of the pixel point Q21
  • processing module is also used for:
  • the comprehensive monitoring result of the business data in the preset time period is determined.
  • the convolutional neural network model is used for N classification
  • the processing module is specifically used for:
  • the image to be identified is input to the convolutional neural network model to obtain a classification result
  • processing module is also used for:
  • the convolutional neural network model is trained according to the image to be recognized and the corrected label corresponding to each false anomaly, and an updated convolutional neural network model is obtained.
  • an embodiment of the present invention also provides a computer device, including:
  • the processor is configured to call the program instructions stored in the memory, and execute the above-mentioned service data monitoring method according to the obtained program.
  • an embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to cause a computer to execute the above-mentioned service data monitoring method.
  • FIG. 1 is a schematic diagram of a system architecture provided by an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of a method for monitoring business data provided by an embodiment of the present invention
  • Fig. 3 is a schematic diagram of a graph provided by an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of a resolution scaling calculation provided by an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of an image to be recognized provided by an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of a convolutional neural network model provided by an embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of an apparatus for monitoring service data provided by an embodiment of the present invention.
  • the first is to compare the business data of the current cycle with the business data of multiple historical cycles.
  • Take business indicators as an example of business transaction volume for example, take days as the cycle unit, monitor the business data of the current cycle (May 8), and compare the historical cycle of the current cycle (including chain history, such as May 7, month-to-month history , such as April 8) for comparison with the business data
  • the chain-on-quarter threshold is now set to be 30% of the upper and lower limits (that is, the business transaction volume in the current cycle is compared with the business transaction volume in the chain-on-month history, and the increase or decrease in the business transaction volume The decline cannot exceed 30%)
  • the upper limit of the year-on-year threshold is 20% (that is, the business transaction volume in the current cycle is compared with the month-on-month historical business transaction volume, and the increase in business transaction volume cannot exceed 20%)
  • the lower limit of the year-on-year threshold is 40% % (that is, the business transaction volume in the current cycle is compared with the historical business transaction volume on a year-
  • the second is to determine the business data range by fitting the curve. Still taking business indicators as an example of business transaction volume, for example, generate a fitting curve based on business data in the historical cycle, and determine the threshold range of business transaction volume in the current cycle according to the fitting curve is 80-100, if the business transaction in the current cycle If the amount is not within the threshold range, it is determined that the business data of the current period is abnormal data.
  • the threshold needs to be manually set and adjusted, and the threshold range for a period cannot be adapted to the whole period of the period, and the business data cannot be dynamically monitored, and the accuracy is not high. And when the business volume is extremely small, a large number of false abnormalities will appear. For example, the historical business transaction volume is 1, the current business transaction volume is 0, and the drop is 100%. If it exceeds the threshold, it is determined that the current business data is abnormal.
  • the fitting curve cannot be dynamically monitored according to the actual situation of the business data.
  • the historical business data fluctuates greatly, and the accuracy of the threshold range determined by the fitting curve is extremely small, so it cannot
  • the business data optimizes the threshold range, so it is impossible to dynamically determine the threshold range.
  • FIG. 1 exemplarily shows a system architecture to which this embodiment of the present invention is applicable.
  • the system architecture includes a server 100 , and the server 100 may include a processor 110 , a communication interface 120 and a memory 130 .
  • the communication interface 120 is used to obtain the first service data within a preset period and the second service data of a historical period associated with the preset period.
  • the processor 110 is the control center of the server 100, and uses various interfaces and routes to connect various parts of the entire server 100, by running or executing software programs/or modules stored in the memory 130, and calling data stored in the memory 130, Various functions of the server 100 are executed and data is processed.
  • the processor 110 may include one or more processing units.
  • the memory 130 can be used to store software programs and modules, and the processor 110 executes various functional applications and data processing by running the software programs and modules stored in the memory 130 .
  • the memory 130 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application required by a function, etc.; the data storage area may store data created according to business processing, etc.
  • the memory 130 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.
  • FIG. 1 is only an example, which is not limited in this embodiment of the present invention.
  • FIG. 2 exemplarily shows a schematic flowchart of a service data monitoring method provided by an embodiment of the present invention, and the process can be executed by a service data monitoring device.
  • the process specifically includes:
  • Step 210 acquiring the first business data within the preset period and the second business data of the historical period associated with the preset period.
  • the first business data and the second business data are for the same business index, for example, both the first business data and the second business data are business transaction volumes.
  • Step 220 generating images to be recognized representing the first service data and the second service data.
  • the basis for determining whether the first business data is abnormal data includes the first business data and the second business data. The difference is shown intuitively.
  • Step 230 input the image to be recognized into the convolutional neural network model, and determine the monitoring results of the service data within the preset period under the service index.
  • the convolutional neural network model is trained based on historical identification images with historical monitoring result labels, and the historical monitoring results are determined based on the relationship between the first business data and the second business data in the historical identification images.
  • the business index includes multiple types, specifically, the business index is at least one of the following: business transaction volume, business average time-consuming and business success rate.
  • the service index can be determined according to the service log collected at the service interface. For example, through the collection module (such as agent), the service log is collected at the preset service interface, and then according to the collected The service log of the service log determines the service indicators, as shown in Table 1 below.
  • the status code is used to indicate whether the service data is successful (that is, normal), for example, the status code 200 indicates that the service data is normal, and the status code 500 indicates that the service data is abnormal.
  • the business success rate is determined according to the normal business quantity and the total business quantity within a preset period, for example, the ratio of the normal business quantity to the total business quantity is determined as the business success rate.
  • the historical period includes a chain period of a preset period and/or a year-on-year period of a preset period.
  • the preset time period is 12:00-12:10 on July 7th
  • the chain-on-month period will be 11:50-12:00 on July 7th
  • the year-on-year period will be 12:00-12:10 on June 7th.
  • the year-on-year period can also be divided into weekly year-on-year, month-on-year and year-on-year, etc.
  • the default time period is 12:00-12:10 on July 7, 2021
  • the weekly year-on-year period is June 30, 2021 at 12:00 00-12:10
  • the month-to-month period is 12:00-12:10, June 7, 2021
  • the year-to-year period is 12:00-12:10, July 7, 2020.
  • the specific year-on-year classification is again not limited.
  • step 220 the coordinate system is first determined according to the preset time period, the first business data and the second business data, and then the graph representing the first business data and the second business data is determined in the coordinate system, and the Get the image to be recognized.
  • a coordinate system is generated with time as the abscissa and business indicators as the ordinate, and then the first curve of the first business data in the coordinate system and the second curve of the second business data in the coordinate system are determined to obtain the graph, Then the graph is preprocessed to determine the image to be recognized.
  • the minimum unit and peak value of the abscissa in the coordinate system are determined according to a preset period, for example, if the preset period is 10 hours, the minimum unit of the abscissa is hour, or half an hour, The peak is 10.
  • the minimum unit and peak value of the ordinate in the coordinate system are determined according to business indicators. For example, if the ordinate is the business success rate, the ordinate takes the success rate of 10% as the minimum unit, or the success rate of 20% as the minimum unit, and the peak value is 100 %. It should be noted that the value of the above minimum unit is just an example, and is not specifically limited here.
  • the preset time period is 10 minutes, and the first service data within the preset time period is acquired, as shown in Table 2 below.
  • the second business data of the historical period associated with the preset period is obtained, wherein the second business data includes the ring-to-quarter historical business data and the weekly historical business data of the preset period, as shown in Table 3 and Table 4 below .
  • A represents the first business data
  • B represents the chain historical business data
  • C represents the weekly historical business data
  • A represents the first business data
  • B represents the chain historical business data
  • C represents the weekly historical business data
  • Figure 3 exemplarily shows a schematic diagram of a graph, which includes the first business data, chain historical business data and weekly historical business data, t Represents the time, in minutes, and n represents the transaction volume.
  • the distinction between the first business data, the chain historical business data and the weekly year-on-year historical business data in the graph can be distinguished according to the line format, such as straight line, dotted line, dotted line, etc., and can also be distinguished according to the color of the line, such as The first business data is a red straight line, the quarter-on-quarter historical business data is a blue straight line, and the weekly year-on-year historical business data is a green straight line, no specific limitations are set here.
  • the graph is generated into a picture, and the image to be recognized is obtained according to the picture.
  • the image to be recognized For obtaining the image to be recognized, for any coordinate point of the image to be recognized with a preset resolution, determine the pixel value of the coordinate point according to the pixel value of the adjacent coordinate point of the coordinate point in the picture, and obtain the image to be recognized.
  • the resolution of the picture is scaled to a preset resolution, and the pixel value of any pixel in the image to be identified is determined according to the following formula (1), to obtain the image to be identified;
  • f(P) is the pixel value of any pixel point P in the image to be recognized
  • (x, y) is the coordinate value of the pixel point P
  • (x1, y1) is the adjacent pixel point located in the lower left corner of the pixel point P in the picture
  • the coordinate value of the pixel point Q11; (x1, y2) is the coordinate value of the adjacent pixel point Q12 located at the upper left corner of the pixel point P in the picture
  • (x2, y1) is the adjacent pixel located at the lower right corner of the pixel point P Q21 in the picture
  • the coordinate value of the point; (x2, y2) is the coordinate value of the adjacent pixel point Q22 located in the upper right corner of the pixel point P in the picture
  • f(Q11) is the pixel value of the pixel point Q11
  • f(Q12) is the pixel value of the pixel point Q12.
  • Pixel value, f(Q21) is the pixel value of pixel point Q21, and f(
  • the default resolution is 224 pixels*224 pixels. Due to the different maximum value of business data within the preset time period, the peak value of the coordinate system may be different, and the height or width of the graph may also be different. The image generated according to the graph The resolution will also be different. Therefore, in order to improve the detection rate, the resolution of the picture is preprocessed to a preset resolution.
  • the resolution of the picture is greater than the preset resolution, the resolution of the picture needs to be reduced to the preset resolution; if the resolution of the picture is smaller than the preset resolution, the resolution of the picture needs to be enlarged to the preset resolution. Set the resolution.
  • FIG. 4 exemplarily shows a schematic diagram of a resolution scaling calculation.
  • the image to be recognized will be described in a specific example in conjunction with the following FIG. 4 .
  • f(R1) is the pixel value of pixel P on the abscissa for pixel points Q11 and Q21
  • f(R2) is the pixel value of pixel P on the abscissa for pixel points Q12 and Q22. If there is Decimal pixel values, the calculation result is rounded.
  • the zooming of the picture can be realized, and then the zoomed picture is determined as the image to be recognized.
  • the pixel values of the pixel point P on the ordinate for the pixel points Q11, Q21, Q12, and Q22 can be determined first, and then the pixel values of the pixel point P can be determined, such as According to the following formula (5) and formula (6), determine the pixel value of the pixel point P on the ordinate for the pixel points Q11, Q21, Q12 and Q22, and then determine the pixel value of the pixel point P according to the following formula (7) .
  • the pixel value of any pixel in the image to be recognized can also be determined according to the ratio of the horizontal and vertical coordinates of the picture resolution to the preset resolution, for example, the picture resolution is mxn, and the preset If the resolution is axb, the side length ratios are m/a and n/b respectively.
  • the pixel point corresponding to the picture is (im/a, jn/b), If im/a, jn/b are non-integer, then according to the rounding method, determine the pixel point of the corresponding picture, and use the pixel value of the pixel point as the pixel value of the pixel point (i, j), so as to obtain the to-be-recognized
  • the pixel values of all pixels in the image are used to determine the image to be recognized. Therefore, in the embodiment of the present invention, there is no specific limitation on the method for image zooming.
  • step 230 the convolutional neural network model is trained according to historical recognition images with historical monitoring result labels.
  • FIG. 5 exemplarily shows a schematic diagram of an image to be recognized .
  • the historical recognition image is similar to that shown in Figure 5.
  • the operation and maintenance personnel assign historical monitoring result labels to the historical recognition image, and then train the convolutional neural network model to realize supervised machine learning.
  • the convolutional neural network model may be a VggNet convolutional neural network model, a GoogLeNet convolutional neural network model, or the like.
  • the convolutional neural network model is an AlexNet convolutional neural network model.
  • the convolutional neural network model is used for N classification, the image to be recognized is input to the convolutional neural network model to obtain the classification result, and the monitoring result corresponding to the classification result is determined according to the comparison relationship between the N classification and the M type monitoring results, wherein , N is greater than M.
  • FIG. 6 exemplarily shows a schematic diagram of a convolutional neural network model.
  • AlexNet includes 5 convolutional layers (conv), respectively c1, c2, c3, c4 and c5; 3 fully connected layers (fully connected), respectively f1, f2 and f3; the model output is 1000 digital values corresponding to 1000 categories, and the output results are converted into decimals between 0-1 through the softmax function-correspondence The probability of multiple classifications of traffic status results.
  • control relationship between the N classification and the M classification monitoring results can be a value preset by the operation and maintenance personnel based on experience.
  • the N classification includes (m1,...,m1000), and the detection results are divided into 5 types, including Business is normal, business is slightly abnormal, business is generally abnormal, business is majorly abnormal, and business is seriously abnormal.
  • (m1, m2, ..., m200) corresponds to normal business
  • (m201, m202, ..., m400) corresponds to slight abnormal business
  • (m401, m402, ..., m600) corresponds to common abnormal business
  • (m601, m602 ,..., m800) corresponds to major business abnormalities
  • (m801, m802,..., m1000) corresponds to serious business abnormalities.
  • the comprehensive monitoring results may be determined according to the weights corresponding to the service indicators, and whether the first service data is abnormal is determined according to the comprehensive detection results.
  • the image to be recognized is input into the convolutional neural network model, and after determining the monitoring results of the business data under the business indicators within the preset period, according to the preset weight of each business indicator and the business data within the preset time period
  • the monitoring results under each business indicator determine the comprehensive monitoring results of business data within a preset period of time.
  • the weight of the detection result type can also be preset, for example, the weight of normal business is 0.9, the weight of minor business abnormality is 0.8, the weight of normal business abnormality is 0.6, the weight of major business abnormality is 0.3 and the weight of serious business abnormality The weight is 0.1.
  • the preset weights of the service indicators may be as follows: the weight of the business success rate is 0.5, the weight of the average time consumption of the business is 0.3, and the weight of the business transaction volume is 0.2.
  • preset weights and abnormal thresholds are set artificially based on experience, and are not specifically limited here.
  • the convolutional neural network model if the operation and maintenance personnel determine that the abnormal business data is a false abnormality, optimize the convolutional neural network model.
  • the convolutional neural network model is trained according to the images to be recognized and corrected labels corresponding to each false anomaly, and an updated convolutional neural network model is obtained.
  • the operation and maintenance personnel will mark the business data a1 when determining that the abnormal business data a1 is a false abnormality, and then determine the number of false abnormal business data (such as 1001 , including a1,..., a1001) is greater than 1000 (quantity threshold), then use the business data a1,..., a1001 as training samples to optimize the training of the convolutional neural network model to increase the convolutional neural network model. Identify the reliability and reduce the probability of the convolutional neural network model identifying incorrect and abnormal business data.
  • an alarm is issued to indicate that the user has abnormal service data.
  • the specific alarm method may be voice broadcast, etc., which is not specifically limited here.
  • FIG. 7 exemplarily shows a schematic structural diagram of a device for monitoring service data provided by an embodiment of the present invention, and the device can execute a flow of a method for monitoring service data.
  • the device specifically includes:
  • An acquisition module 710 configured to acquire first business data within a preset time period and second business data in a historical period associated with the preset time period; the first business data and the second business data are for the same business indicator ;
  • a processing module 720 configured to generate an image to be recognized representing the first business data and the second business data
  • the historical identification image is trained; the historical monitoring result is determined according to the relationship between the first service data and the second service data in the historical identification image.
  • the business indicator is at least one of the following: business transaction volume, business average time-consuming and business success rate;
  • the historical period includes a chain period of the preset period and/or a year-on-year period of the preset period.
  • processing module 720 is specifically configured to:
  • processing module 720 is specifically configured to:
  • Scaling the resolution of the picture to a preset resolution determining the pixel value of any pixel in the image to be identified according to the following formula (1), to obtain the image to be identified;
  • f(P) is the pixel value of any pixel point P in the image to be recognized
  • (x, y) is the coordinate value of the pixel point P
  • (x1, y1) is located at the bottom left of the pixel point P in the picture
  • (x1, y2) is the coordinate value of the adjacent pixel point Q12 located at the upper left corner of the pixel point P in the picture;
  • (x2, y1) is located at the pixel point in the picture
  • (x2, y2) is the coordinate value of the adjacent pixel point Q22 located in the upper right corner of pixel point P in the picture;
  • f(Q11) is the pixel value of pixel point Q11
  • f(Q12) is the pixel value of the pixel point Q12
  • f(Q21) is the pixel value of the pixel point Q21
  • processing module 720 is also used for:
  • the comprehensive monitoring result of the business data in the preset time period is determined.
  • the convolutional neural network model is used for N classification
  • the processing module 720 is specifically used for:
  • the image to be identified is input to the convolutional neural network model to obtain a classification result
  • the monitoring results corresponding to the classification results are determined, wherein N is greater than M.
  • processing module 720 is also used for:
  • the convolutional neural network model is trained according to the image to be recognized and the corrected label corresponding to each false anomaly, and an updated convolutional neural network model is obtained.
  • the embodiment of the present invention also provides a computer device, including:
  • the processor is configured to call the program instructions stored in the memory, and execute the above-mentioned service data monitoring method according to the obtained program.
  • an embodiment of the present invention also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to perform the monitoring of the above business data method.
  • the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions
  • the device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Finance (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Computing Systems (AREA)
  • Strategic Management (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Technology Law (AREA)
  • Health & Medical Sciences (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种业务数据的监测方法及装置，包括：获取预设时段内的第一业务数据以及与预设时段关联的历史时段的第二业务数据，其中，第一业务数据和第二业务数据针对同一业务指标，生成表征第一业务数据和第二业务数据的待识别图像，将待识别图像输入至卷积神经网络模型中，确定预设时段内的业务数据在业务指标下的监测结果，卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的，历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。因卷积神经网络模型融合了各历史时段的业务数据的情况，可以适用于业务数据的全时段的监测，因此实现了动态的对业务数据进行监测，提高业务数据监测的准确性。

Description

一种业务数据的监测方法及装置
相关申请的交叉引用
本申请要求在2021年05月31日提交中国专利局、申请号为202110597265.4、申请名称为“一种业务数据的监测方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及金融科技(Fintech)领域，尤其涉及一种业务数据的监测方法及装置。
背景技术
随着计算机技术的发展，越来越多的技术(例如：区块链、云计算或大数据)应用在金融领域，传统金融业正在逐步向金融科技转变，大数据技术也不例外，但由于金融、支付行业的安全性、实时性要求，也对大数据技术提出的更高的要求。
目前，监测业务数据的方法一般是，将当前周期的业务数据与多个历史周期的业务数据进行对比，例如，以天为周期单位，若监测当前周期(5月8日)的业务数据是否异常，需要将当前周期的业务数据与当前周期之前的多个历史周期(如5月5日、5月6日和5月7日)的业务数据进行对比，若当前周期的业务数据的波动(增加或下降)在阈值范围内，则确定当前周期的业务数据为正常数据。
但是，上述阈值范围需要人为的进行手工设置和调整，针对一个周期的阈值范围并不能适应于该周期的全时段，无法动态的对业务数据进行监测，且准确性不高。例如，以天为周期单位，某种场景下的业务在每周期内，8:00-17:00的业务数据较多，17:00-8:00的业务数据较少，仅在17:00-18:00时间段内，业务数据急剧下降，超过阈值范围，此时，产生的异常是错误的，实际为正常业务数据。
因此，现需要一种业务数据的监测方法，用于实现动态的对业务数据进行监测，提高业务数据监测的准确性。
发明内容
本发明实施例提供一种业务数据的监测方法及装置，用于实现动态的对业务数据进行监测，提高业务数据监测的准确性。
第一方面，本发明实施例提供一种业务数据的监测方法，包括：
获取预设时段内的第一业务数据以及与所述预设时段关联的历史时段的第二业务数据；所述第一业务数据和所述第二业务数据针对同一业务指标；
生成表征所述第一业务数据和所述第二业务数据的待识别图像；
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的 业务数据在所述业务指标下的监测结果；所述卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的；所述历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
现有技术中的监测阈值是根据经验人为设置的；而本申请中卷积神经网络模型是根据历史识别图像训练得到的；每个历史识别图像中包括第一业务数据(当前业务数据)和第二业务数据(历史业务数据)，且具有用于表征历史监测结果的标签；通过历史识别图像对第一卷积神经网络模型的训练，使得第一卷积神经网络模型可以识别出第一业务数据与第二业务数据之间的关系并根据第一业务数据与第二业务数据之间的关系确定第一业务数据是否异常。本申请中的每个历史识别图像分别代表各种的历史时段，因而卷积神经网络模型实际上是融合了各历史时段的业务数据的情况，可以适用于业务数据的全时段的监测。因此，可以实现动态的对业务数据进行监测，提高业务数据监测的准确性。
可选的，所述业务指标为以下至少一种：业务交易量、业务平均耗时和业务成功率；
所述历史时段包括所述预设时段的环比时段和/或所述预设时段的同比时段。
上述技术方案中，将业务指标分类不同类型，提升了对业务数据监测的全面性，因为预设时段的环比时段和同比时段与预设时段更具有关联性，可以更好的比较业务数据，以此增加了业务数据监测的准确性。
可选的，生成表征所述第一业务数据和所述第二业务数据的待识别图像，包括：
以时间为横坐标、业务指标为纵坐标生成坐标系；
确定所述第一业务数据在所述坐标系下的第一曲线和所述第二业务数据在所述坐标系下的第二曲线，得到曲线图；
将所述曲线图进行预处理，确定所述待识别图像。
现有技术方案中，是根据当前业务数据和历史数据的差异来确定业务数据是否异常，当并未对差异进行验证是否异常；本申请中，通过图像的形式，将第一业务数据和第二业务数据之间的第一区别直观的表现出来，进而根据卷积神经网络模型中历史识别图像的第二区别来确定第一区别是否异常，根据第一区别是否异常来确定第一业务数据是否异常，相当于实现两次监测，以增加业务数据监测的准确性。
可选的，将所述曲线图进行预处理，确定所述待识别图像，包括：
将所述曲线图生成图片；
将所述图片的分辨率缩放为预设分辨率，根据下述公式(1)确定所述待识别图像中任一像素点的像素值，得到所述待识别图像；
Figure PCTCN2021139697-appb-000001
其中，f(P)为所述待识别图像内任一像素点P的像素值，(x，y)为像素点P的坐标值，(x1，y1)为所述图片中位于像素点P左下角的相邻像素点Q11的坐标值；(x1，y2)为所述图片中位于像素点P左上角的相邻像素点Q12的坐标值；(x2，y1)为所述图片中位于像素点P右下角Q21的相邻像素点的坐标值；(x2，y2)为所述图片中位于像素点P右上角的相邻像素点Q22的坐标值；f(Q11)为像素点Q11的像素值，f(Q12)为像素点Q12的像素值，f(Q21)为像素点Q21的像素值，f(Q22)为像素点Q22的像素值。
上述技术方案中，因业务数据的值不尽相同，以此存在坐标系中的峰值大小不一，导致生成的图片分辨率不统一，存在卷积神经网络模型监测异常的情况，因此通过预处理，实现待识别图像地分辨率统一，来增加业务数据监测的准确性。
可选的，所述将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果之后，还包括：
根据各业务指标的预设权重和所述预设时段内的业务数据在各业务指标下的监测结果，确定所述预设时段内的业务数据的综合监测结果。
上述技术方案中，将各业务指标得到的检测结果根据预设权重进行聚合，得到综合监测结果，以防止某一业务指标的业务数据异常，其余业务指标的业务数据正常情况下导致错误判断第一业务数据是异常的，以此增加业务数据监测的准确性。
可选的，所述卷积神经网络模型用于进行N分类；
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内业务数据的监测结果，包括：
将所述待识别图像输入至卷积神经网络模型得到分类结果；
根据N分类与M类监测结果的对照关系，确定所述分类结果对应的监测结果，其中，N大于M。
上述技术方案中，N分类与M类监测结果的对照关系，来确定待识别图像的分类结果，以此确定待识别图像是否是异常的，其中，N分类与M类监测结果的对照关系可以是用户根据经验预设的，以增加对业务数据监测的灵活性。
可选的，在确定误异常的监测结果的数量大于数量阈值时，根据各误异常对应的待识别图像及矫正标签对所述卷积神经网络模型进行训练，得到更新后的卷积神经网络模型。
上述技术方案中，针对卷积神经网络模型确定的异常业务数据，若用户确定该异常业务数据是正常的，即卷积神经网络模型发生错误判断，则将该异常业务数据进行记录，并在这种误异常的监测结果的数量大于数量阈值时，根据误异常的业务数据对卷积神经网络模型进行优化训练，以此增加卷积神经网络模型的识别准确性，从而增加了对业务数据监测的准确性。
第二方面，本发明实施例提供一种业务数据的监测装置，包括：
获取模块，用于获取预设时段内的第一业务数据以及与所述预设时段关联的历史时段的第二业务数据；所述第一业务数据和所述第二业务数据针对 同一业务指标；
处理模块，用于生成表征所述第一业务数据和所述第二业务数据的待识别图像；
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果；所述卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的；所述历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
可选的，所述业务指标为以下至少一种：业务交易量、业务平均耗时和业务成功率；
所述历史时段包括所述预设时段的环比时段和/或所述预设时段的同比时段。
可选的，所述处理模块具体用于：
以时间为横坐标、业务指标为纵坐标生成坐标系；
确定所述第一业务数据在所述坐标系下的第一曲线和所述第二业务数据在所述坐标系下的第二曲线，得到曲线图；
将所述曲线图进行预处理，确定所述待识别图像。
可选的，所述处理模块具体用于：
将所述曲线图生成图片；
将所述图片的分辨率缩放为预设分辨率，根据下述公式(1)确定所述待识别图像中任一像素点的像素值，得到所述待识别图像；
Figure PCTCN2021139697-appb-000002
其中，f(P)为所述待识别图像内任一像素点P的像素值，(x，y)为像素点P的坐标值，(x1，y1)为所述图片中位于像素点P左下角的相邻像素点Q11的坐标值；(x1，y2)为所述图片中位于像素点P左上角的相邻像素点Q12的坐标值；(x2，y1)为所述图片中位于像素点P右下角Q21的相邻像素点的坐标值；(x2，y2)为所述图片中位于像素点P右上角的相邻像素点Q22的坐标值；f(Q11)为像素点Q11的像素值，f(Q12)为像素点Q12的像素值，f(Q21)为像素点Q21的像素值，f(Q22)为像素点Q22的像素值。
可选的，所述处理模块还用于：
根据各业务指标的预设权重和所述预设时段内的业务数据在各业务指标下的监测结果，确定所述预设时段内的业务数据的综合监测结果。
可选的，所述卷积神经网络模型用于进行N分类；
所述处理模块具体用于：
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内业务数据的监测结果，包括：
将所述待识别图像输入至卷积神经网络模型得到分类结果；
根据N分类与M类监测结果的对照关系，确定所述分类结果对应的监测 结果，其中，N大于M。
可选的，所述处理模块还用于：
在确定误异常的监测结果的数量大于数量阈值时，根据各误异常对应的待识别图像及矫正标签对所述卷积神经网络模型进行训练，得到更新后的卷积神经网络模型。
第三方面，本发明实施例还提供一种计算机设备，包括：
存储器，用于存储程序指令；
处理器，用于调用所述存储器中存储的程序指令，按照获得的程序执行上述业务数据的监测方法。
第四方面，本发明实施例还提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于使计算机执行上述业务数据的监测方法。
附图说明
为了更清楚地说明本发明实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简要介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域的普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本发明实施例提供的一种系统架构示意图；
图2为本发明实施例提供的一种业务数据的监测方法的流程示意图；
图3为本发明实施例提供的一种曲线图的示意图；
图4为本发明实施例提供的一种分辨率缩放计算的示意图；
图5为本发明实施例提供的一种待识别图像的示意图；
图6为本发明实施例提供的一种卷积神经网络模型的示意图；
图7为本发明实施例提供的一种业务数据的监测装置的结构示意图。
具体实施方式
为了使本发明的目的、技术方案和优点更加清楚，下面将结合附图对本发明作进一步地详细描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
现有技术中，监测业务数据的方法一般分为以下两种：
第一种，将当前周期的业务数据与多个历史周期的业务数据进行对比。以业务指标为业务交易量举例，例如，以天为周期单位，监测当前周期(5月8日)的业务数据，将当前周期的历史周期(包括环比历史，如5月7日，月同比历史，如4月8日)的业务数据与之进行对比，现设环比阈值为上下限均为30％(即当前周期的业务交易量与环比历史的业务交易量相比，业务交易量的上升或下降均不能超过30％)，同比阈值的上限为20％(即当前周期的业务交易量与月同比历史的业务交易量相比，业务交易量的上升不能超过 20％)同比阈值的下限为40％(即当前周期的业务交易量与月同比历史的业务交易量相比，业务交易量的下降不能超过40％)，若超过阈值，则确定当前周期的业务数据为异常数据。
第二种，通过拟合曲线的方式确定业务数据范围。仍以业务指标为业务交易量举例，例如，根据历史周期内的业务数据，生成拟合曲线，根据拟合曲线确定当前周期的业务交易量的阈值范围在80-100，若当前周期的业务交易量不在阈值范围内，则确定当前周期的业务数据为异常数据。
然而，上述第一种方法中，阈值需要人为的进行手工设置和调整，针对一个周期的阈值范围并不能适应于该周期的全时段，无法动态的对业务数据进行监测，且准确性不高，且针对业务量极小的情况下，会出现大量的误异常，例如，历史业务交易量为，1当前业务交易量为0，下降100％，超过阈值，确定当前业务数据异常。
第二种方法中，拟合曲线不能根据业务数据的实际情况进行动态的监测，例如，历史业务数据的波动较大，拟合曲线确定出的阈值范围准确性极小，无法根据确定的误异常业务数据对阈值范围进行优化，因此无法实现动态的确定阈值范围。
因此，现需要一种业务数据的监测方法，用于实现动态的对业务数据进行监测，提高业务数据监测的准确性。
图1示例性的示出了本发明实施例所适用的一种系统架构，该系统架构包括服务器100，该服务器100可以包括处理器110、通信接口120和存储器130。
其中，通信接口120用于获取预设时段内的第一业务数据以及与预设时段关联的历史时段的第二业务数据。
处理器110是服务器100的控制中心，利用各种接口和路线连接整个服务器100的各个部分，通过运行或执行存储在存储器130内的软件程序/或模块，以及调用存储在存储器130内的数据，执行服务器100的各种功能和处理数据。可选地，处理器110可以包括一个或多个处理单元。
存储器130可用于存储软件程序以及模块，处理器110通过运行存储在存储器130的软件程序以及模块，从而执行各种功能应用以及数据处理。存储器130可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序等；存储数据区可存储根据业务处理所创建的数据等。此外，存储器130可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。
需要说明的是，上述图1所示的结构仅是一种示例，本发明实施例对此不做限定。
基于上述描述，图2示例性的示出了本发明实施例提供的一种业务数据的监测方法的流程示意图，该流程可由业务数据的监测装置执行。
如图2所示，该流程具体包括：
步骤210，获取预设时段内的第一业务数据以及与所述预设时段关联的历 史时段的第二业务数据。
本发明实施例中，第一业务数据和第二业务数据针对同一业务指标，例如，第一业务数据和第二业务数据均为业务交易量。
步骤220，生成表征所述第一业务数据和所述第二业务数据的待识别图像。
本发明实施例中，确定第一业务数据是否为异常数据的依据包括第一业务数据和第二业务数据，具体是以图像的形式，将第一业务数据和第二业务数据之间的第一区别直观的表现出来。
步骤230，将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果。
本发明实施例中，卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的，历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
在步骤210中，业务指标包括了多个种类，具体的，业务指标为以下至少一种：业务交易量、业务平均耗时和业务成功率。
在一种可实施的方式中，业务指标可以根据在业务接口采集到的业务日志来确定，举例来说，通过采集模块(如agent)，在预设业务接口处采集业务日志，然后根据采集到的业务日志确定出业务指标，如下述表1所示。
表1
Figure PCTCN2021139697-appb-000003
其中，状态码用于表征业务数据是否成功(即正常)，例如，状态码200表征业务数据正常，状态码500表征业务数据异常。业务成功率是根据预设时段内正常的业务数量与总业务数量确定的，如将正常的业务数量与总业务数量的比值确定为业务成功率。
本发明实施例中，历史时段包括预设时段的环比时段和/或预设时段的同比时段。例如，预设时段为7月7日12:00-12:10，则环比时段为7月7日11:50-12:00，同比时段为6月7日12:00-12:10。其中，同比时段还可以分为周同比，月同比和年同比等，例如，预设时段为2021年7月7日12:00-12:10，周同比时段为2021年6月30日12:00-12:10，月同比时段为2021年6月7日12:00-12:10，年同比时段为2020年7月7日12:00-12:10。具体的同比分类再 次不做限定。
在步骤220中，先根据预设时段、第一业务数据和第二业务数据来确定坐标系，进而在坐标系中确定出表征第一业务数据和第二业务数据的曲线图，根据曲线图来得到待识别图像。
具体的，以时间为横坐标、业务指标为纵坐标生成坐标系，再确定第一业务数据在坐标系下的第一曲线和第二业务数据在坐标系下的第二曲线，得到曲线图，然后将曲线图进行预处理，确定待识别图像。
在本发明实施例中，坐标系中横坐标的最小单位和峰值是根据预设时段确定的，例如，预设时段为10小时，则横坐标以小时为最小单位，或半小时为最小单位，峰值为10。坐标系中纵坐标的最小单位和峰值是根据业务指标确定的，例如，纵坐标为业务成功率，则纵坐标以成功率10％为最小单位，或成功率20％为最小单位，峰值为100％。需要说明的是，上述最小单位的取值仅是实例，在此不做具体限定。
为了更好的描述上述技术方案，下面将在具体实例中以业务指标Wie业务交易量进行阐述。
实例1
预设时段为10分钟，获取预设时段内的第一业务数据，如下述表2所示。
表2
Figure PCTCN2021139697-appb-000004
根据同样的方法，获取预设时段关联的历史时段的第二业务数据，其中，第二业务数据包括预设时段的环比历史业务数据和周同比历史业务数据，如下述表3和表4所示。
表3
Figure PCTCN2021139697-appb-000005
表4
Figure PCTCN2021139697-appb-000006
现以分钟为最小时间单位，根据表2至表4的数据可以得到汇总后的业务数据，如下述表5所示。
表5
  1 2 3 4 5 6 7 8 9 10
A 70 75 71 74 69 79 68 61 72 84
B 61 79 59 78 65 66 74 55 73 77
C 74 83 72 82 73 75 77 64 74 72
其中，A表征第一业务数据，B表征环比历史业务数据，C表征周同比历史业务数据，然后根据表5生成坐标系，因表5中最大数据为84，因此坐标系中纵坐标的峰值可以取85、90等，本发明实施例中峰值为90。
然后根据坐标系确定出曲线图，如图3所示，图3示例性的示出了一种曲线图的示意图，其中包括了第一业务数据、环比历史业务数据和周同比历史业务数据，t表征时间，单位为分钟，n表征交易业务量。
需要说明的是，在曲线图中区分第一业务数据、环比历史业务数据和周同比历史业务数据可以根据线条格式区分，如直线、虚线、点画线等，还可以根据线条的颜色进行区分，如第一业务数据为红色直线，环比历史业务数据为蓝色直线，周同比历史业务数据为绿色直线，在此不做具体限定。
在本发明实施例中，在得到曲线图之后，将曲线图生成图片，根据图片得到待识别图像。
对于曲线图生成图片，具体将图3中的坐标系删除，仅根据曲线图来生成图片，然后对图片进行缩放，将图片的分辨率缩放至预设分辨率，进而得到待识别图像。
对于得到待识别图像，针对预设分辨率的待识别图像的任一坐标点，根据坐标点在图片中的相邻坐标点的像素值，确定坐标点的像素值，得到待识别图像。
将曲线图生成图片；
将图片的分辨率缩放为预设分辨率，根据下述公式(1)确定待识别图像中任一像素点的像素值，得到待识别图像；
Figure PCTCN2021139697-appb-000007
其中，f(P)为待识别图像内任一像素点P的像素值，(x，y)为像素点P的坐标值，(x1，y1)为图片中位于像素点P左下角的相邻像素点Q11的坐标值；(x1，y2)为图片中位于像素点P左上角的相邻像素点Q12的坐标值；(x2，y1)为图片中位于像素点P右下角Q21的相邻像素点的坐标值；(x2，y2)为图片中位于像素点P右上角的相邻像素点Q22的坐标值；f(Q11)为像素点Q11的像素值，f(Q12)为像素点Q12的像素值，f(Q21)为像素点Q21的像素值，f(Q22)为像素点Q22的像素值。
举例来说，预设分辨率为224像素*224像素，因预设时段内业务数据最大值不同，导致坐标系的峰值可能不同，曲线图高度或宽度也会不同，则根据曲线图生成的图片分辨率也会不同，因此，为了提升监测确定率，将图片的分辨率预处理为预设分辨率。
若图片的分辨率大于预设分辨率，则需要将图片分辨率进行缩小，缩小至预设分辨率，若图片的分辨率小于预设分辨率，则需要将图片分辨率进行放大，放大至预设分辨率。
其中，在进行分辨率缩放时，会依次对待识别图像中的所有像素点进行处理，在两个方向，即X轴和Y轴方向上分别进行一次计算，通过四个相邻像素点插值得到待求像素点的像素值，计算过程中距离待求像素点坐标位置越近的坐标像素权重越大。
结合上述公式(1)，图4示例性的示出了一种分辨率缩放计算的示意图，下面将结合下图4，在具体实例中阐述得到待识别图像。
实例2
将2*2像素缩小成1*1像素的场景，图4所示，Q11，Q12，Q21，Q22为图4中像素点P的四个相邻像素点，像素点P为待求像素点，公式(1)可拆分为下述公式(2)、公式(3)和公式(4)，具体步骤如下：
1、通过Q11(x1，y1)，Q21(x2，y1)，根据下述公式(2)得到f(R1)，通过Q12(x1，y2)，Q22(x2，y2)，根据下述公式(3)得到f(R2)；
Figure PCTCN2021139697-appb-000008
Figure PCTCN2021139697-appb-000009
其中，f(R1)为像素点P针对像素点Q11和Q21在横坐标上的像素值，f(R2)为像素点P针对像素点Q12和Q22在横坐标上的像素值，若计算后存在小数像素值，则对计算结果进行四舍五入。
2、通过R1(x，y1)，R2(x，y2)，根据下述计算公式(4)得到P。
Figure PCTCN2021139697-appb-000010
由此可以实现图片的缩放，然后将缩放后的图片确定为待识别图像。需 要说明的是，在一种可实施的方式中，可以先确定出像素点P针对像素点Q11、Q21、Q12和Q22在纵坐标上像素值，然后在确定出像素点P的像素值，如根据下述公式(5)和公式(6)确定出像素点P针对像素点Q11、Q21、Q12和Q22在纵坐标上像素值，再根据下述公式(7)确定出像素点P的像素值。
Figure PCTCN2021139697-appb-000011
Figure PCTCN2021139697-appb-000012
Figure PCTCN2021139697-appb-000013
在另一种可实施的方式中，还可以根据图片分辨率与预设分辨率的横纵坐标比值来确定待识别图像中任一像素点的像素值，例如，图片分辨率为mxn，预设分辨率为axb，则边长比分别为m/a和n/b，针对待识别图像中任一像素点(i，j)，对应图片的像素点为(im/a,jn/b)，若im/a,jn/b非整数，则根据四舍五入的方式，确定出对应图片的像素点，并将该像素点的像素值作为像素点(i，j)的像素值，以此得到待识别图像中所有像素点的像素值，进而确定出待识别图像。因此，在本发明实施例中，对于图片缩放的方法不做具体的限定。
在步骤230中，卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的，结合图3和上述公式(1)，图5示例性的示出了一种待识别图像的示意图。其中历史识别图像类似于图5所示，由运维人员对历史识别图像赋予历史监测结果标签，进而对卷积神经网络模型进行训练，实现有监督的机器学习。
在一种可实施的方式中，卷积神经网络模型可以为VggNet卷积神经网络模型、GoogLeNet卷积神经网络模型等。
在本发明实施例中，卷积神经网络模型为AlexNet卷积神经网络模型。
进一步地，卷积神经网络模型用于进行N分类，将待识别图像输入至卷积神经网络模型得到分类结果，根据N分类与M类监测结果的对照关系，确定分类结果对应的监测结果，其中，N大于M。
在本发明实施例中，图6示例性的示出了一种卷积神经网络模型的示意图，如图6所示，AlexNet包含5个卷积层(conv)，分别为c1、c2、c3、c4和c5；3个全连接层(fully connected)，分别f1、f2和f3；模型输出为1000个数字值对应1000个分类，通过softmax函数将输出结果转换为0-1之间的小数-对应业务量状态结果多个分类的概率。
示例性的，N分类与M类监测结果的对照关系可以是运维人员根据经验预设的值，例如，N分类包括(m1，……，m1000)，将检测结果分为5中类型，包括业务正常、业务轻微异常、业务普通异常、业务重大异常和业务严重异常。其中，(m1，m2，……，m200)对应业务正常，(m201，m202，……，m400)对应业务轻微异常，(m401，m402，……，m600)对应业务普通异常，(m601，m602，……，m800)对应业务重大异常，(m801，m802，……，m1000)对应业务严重异常。
在一种可实现的方式中，针对不同类型的业务指标，可根据业务指标对应的权重确定出综合监测结果，根据综合检测结果确定第一业务数据是否异常。
具体的，将待识别图像输入至卷积神经网络模型中，确定预设时段内的业务数据在业务指标下的监测结果之后，根据各业务指标的预设权重和预设时段内的业务数据在各业务指标下的监测结果，确定预设时段内的业务数据的综合监测结果。
结合上述技术方案举例来说，对检测结果类型也可以预设权重，例如，业务正常权重为0.9、业务轻微异常权重为0.8、业务普通异常权重为0.6、业务重大异常权重为0.3和业务严重异常权重为0.1。
业务指标的预设权重可以为，业务成功率权重为0.5，业务平均耗时权重为0.3，业务交易量权重为0.2。
结合实例1进行举例，现针对预设时段的第一业务数据的3个业务指标(业务交易量、业务平均耗时和业务成功率)，根据卷积神经网络模型确定出的监测结果分别为，业务普通异常、业务轻微异常、业务正常，则可以根据预设权重得到综合监测结果为，z＝(0.6*0.2)+(0.8*0.3)+(0.9*0.5)＝0.81，若0.81大于异常阈值，则确定第一业务数据为正常数据。
需要说明的是，预设权重和异常阈值是人为根据经验设置的，在此不做具体限定。
示例性的，针对卷积神经网络模型确定出的异常业务数据，若运维人员在确定异常业务数据为误异常时，对卷积神经网络模型进行优化。
具体的，在确定误异常的监测结果的数量大于数量阈值时，根据各误异常对应的待识别图像及矫正标签对所述卷积神经网络模型进行训练，得到更新后的卷积神经网络模型。
举例来说，卷积神经网络模型确定出的异常业务数据a1，运维人员在确定异常业务数据a1为误异常时，将业务数据a1进行标记，在确定误异常的业务数据的数量(如1001，包括a1，……，a1001)大于1000(数量阈值)时，则将业务数据a1，……，a1001，作为训练样本，对卷积神经网络模型进行优化训练，来增加卷积神经网络模型的识别准信性，减少卷积神经网络模型确定误异常的业务数据的概率。
在本发明实施例中，在确定出业务数据异常后，发出告警，以指示用户发生异常的业务数据，具体告警方法可以为语音播报等，在此不做具体限定。
基于相同的技术构思，图7示例性的示出了本发明实施例提供的一种业务数据的监测装置的结构示意图，该装置可以执行业务数据的监测方法的流程。
如图7所示，该装置具体包括：
获取模块710，用于获取预设时段内的第一业务数据以及与所述预设时段关联的历史时段的第二业务数据；所述第一业务数据和所述第二业务数据针对同一业务指标；
处理模块720，用于生成表征所述第一业务数据和所述第二业务数据的待 识别图像；
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果；所述卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的；所述历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
可选的，所述业务指标为以下至少一种：业务交易量、业务平均耗时和业务成功率；
所述历史时段包括所述预设时段的环比时段和/或所述预设时段的同比时段。
可选的，所述处理模块720具体用于：
以时间为横坐标、业务指标为纵坐标生成坐标系；
确定所述第一业务数据在所述坐标系下的第一曲线和所述第二业务数据在所述坐标系下的第二曲线，得到曲线图；
将所述曲线图进行预处理，确定所述待识别图像。
可选的，所述处理模块720具体用于：
将所述曲线图生成图片；
将所述图片的分辨率缩放为预设分辨率，根据下述公式(1)确定所述待识别图像中任一像素点的像素值，得到所述待识别图像；
Figure PCTCN2021139697-appb-000014
其中，f(P)为所述待识别图像内任一像素点P的像素值，(x，y)为像素点P的坐标值，(x1，y1)为所述图片中位于像素点P左下角的相邻像素点Q11的坐标值；(x1，y2)为所述图片中位于像素点P左上角的相邻像素点Q12的坐标值；(x2，y1)为所述图片中位于像素点P右下角Q21的相邻像素点的坐标值；(x2，y2)为所述图片中位于像素点P右上角的相邻像素点Q22的坐标值；f(Q11)为像素点Q11的像素值，f(Q12)为像素点Q12的像素值，f(Q21)为像素点Q21的像素值，f(Q22)为像素点Q22的像素值。
可选的，所述处理模块720还用于：
根据各业务指标的预设权重和所述预设时段内的业务数据在各业务指标下的监测结果，确定所述预设时段内的业务数据的综合监测结果。
可选的，所述卷积神经网络模型用于进行N分类；
所述处理模块720具体用于：
将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内业务数据的监测结果，包括：
将所述待识别图像输入至卷积神经网络模型得到分类结果；
根据N分类与M类监测结果的对照关系，确定所述分类结果对应的监测结果，其中，N大于M。
可选的，所述处理模块720还用于：
在确定误异常的监测结果的数量大于数量阈值时，根据各误异常对应的待识别图像及矫正标签对所述卷积神经网络模型进行训练，得到更新后的卷积神经网络模型。
基于相同的技术构思，本发明实施例还提供一种计算机设备，包括：
存储器，用于存储程序指令；
处理器，用于调用所述存储器中存储的程序指令，按照获得的程序执行上述业务数据的监测方法。
基于相同的技术构思，本发明实施例还提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于使计算机执行上述业务数据的监测方法。
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请是参照根据本申请的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
显然，本领域的技术人员可以对本申请进行各种改动和变型而不脱离本申请的精神和范围。这样，倘若本申请的这些修改和变型属于本申请权利要求及其等同技术的范围之内，则本申请也意图包含这些改动和变型在内。

Claims (10)

  1. 一种业务数据的监测方法，其特征在于，包括：
    获取预设时段内的第一业务数据以及与所述预设时段关联的历史时段的第二业务数据；所述第一业务数据和所述第二业务数据针对同一业务指标；
    生成表征所述第一业务数据和所述第二业务数据的待识别图像；
    将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果；所述卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的；所述历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
  2. 如权利要求1所述的方法，其特征在于，所述业务指标为以下至少一种：业务交易量、业务平均耗时和业务成功率；
    所述历史时段包括所述预设时段的环比时段和/或所述预设时段的同比时段。
  3. 如权利要求1所述的方法，其特征在于，生成表征所述第一业务数据和所述第二业务数据的待识别图像，包括：
    以时间为横坐标、业务指标为纵坐标生成坐标系；
    确定所述第一业务数据在所述坐标系下的第一曲线和所述第二业务数据在所述坐标系下的第二曲线，得到曲线图；
    将所述曲线图进行预处理，确定所述待识别图像。
  4. 如权利要求1所述的方法，其特征在于，将所述曲线图进行预处理，确定所述待识别图像，包括：
    将所述曲线图生成图片；
    将所述图片的分辨率缩放为预设分辨率，根据下述公式(1)确定所述待识别图像中任一像素点的像素值，得到所述待识别图像；
    Figure PCTCN2021139697-appb-100001
    其中，f(P)为所述待识别图像内任一像素点P的像素值，(x，y)为像素点P的坐标值，(x1，y1)为所述图片中位于像素点P左下角的相邻像素点Q11的坐标值；(x1，y2)为所述图片中位于像素点P左上角的相邻像素点Q12的坐标值；(x2，y1)为所述图片中位于像素点P右下角Q21的相邻像素点的坐标值；(x2，y2)为所述图片中位于像素点P右上角的相邻像素点Q22的坐标值；f(Q11)为像素点Q11的像素值，f(Q12)为像素点Q12的像素值，f(Q21)为像素点Q21的像素值，f(Q22)为像素点Q22的像素值。
  5. 如权利要求1所述的方法，其特征在于，其特征在于，所述将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果之后，还包括：
    根据各业务指标的预设权重和所述预设时段内的业务数据在各业务指标 下的监测结果，确定所述预设时段内的业务数据的综合监测结果。
  6. 如权利要求1所述的方法，其特征在于，所述卷积神经网络模型用于进行N分类；
    将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内业务数据的监测结果，包括：
    将所述待识别图像输入至卷积神经网络模型得到分类结果；
    根据N分类与M类监测结果的对照关系，确定所述分类结果对应的监测结果，其中，N大于M。
  7. 如权利要求1至6任一项所述的方法，其特征在于，还包括：
    在确定误异常的监测结果的数量大于数量阈值时，根据各误异常对应的待识别图像及矫正标签对所述卷积神经网络模型进行训练，得到更新后的卷积神经网络模型。
  8. 一种业务数据的监测装置，其特征在于，包括：
    获取模块，用于获取预设时段内的第一业务数据以及与所述预设时段关联的历史时段的第二业务数据；所述第一业务数据和所述第二业务数据针对同一业务指标；
    处理模块，用于生成表征所述第一业务数据和所述第二业务数据的待识别图像；
    将所述待识别图像输入至卷积神经网络模型中，确定所述预设时段内的业务数据在所述业务指标下的监测结果；所述卷积神经网络模型是根据具有历史监测结果标签的历史识别图像训练得到的；所述历史监测结果是根据历史识别图像中第一业务数据与第二业务数据的关系确定的。
  9. 一种计算机设备，其特征在于，包括：
    存储器，用于存储程序指令；
    处理器，用于调用所述存储器中存储的程序指令，按照获得的程序执行权利要求1至7任一项所述的方法。
  10. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于使计算机执行权利要求1至7任一项所述的方法。
PCT/CN2021/139697 2021-05-31 2021-12-20 一种业务数据的监测方法及装置 WO2022252573A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110597265.4A CN113326879B (zh) 2021-05-31 2021-05-31 一种业务数据的监测方法及装置
CN202110597265.4 2021-05-31

Publications (1)

Publication Number Publication Date
WO2022252573A1 true WO2022252573A1 (zh) 2022-12-08

Family

ID=77422580

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/139697 WO2022252573A1 (zh) 2021-05-31 2021-12-20 一种业务数据的监测方法及装置

Country Status (2)

Country Link
CN (1) CN113326879B (zh)
WO (1) WO2022252573A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113326879B (zh) * 2021-05-31 2024-07-19 深圳前海微众银行股份有限公司 一种业务数据的监测方法及装置
CN114401310B (zh) * 2021-12-31 2024-05-31 苏州市瑞川尔自动化设备有限公司 可视化云业务数据优化方法及服务器
CN116150221B (zh) * 2022-10-09 2023-07-14 浙江博观瑞思科技有限公司 服务于企业电商运营管理的信息交互方法及系统
CN116091202A (zh) * 2022-12-29 2023-05-09 北京君航微金信息科技有限公司 金融业务监控预警方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10445738B1 (en) * 2018-11-13 2019-10-15 Capital One Services, Llc Detecting a transaction volume anomaly
CN111199018A (zh) * 2019-12-27 2020-05-26 东软集团股份有限公司 异常数据检测方法、装置、存储介质及电子设备
CN111382020A (zh) * 2020-03-06 2020-07-07 中国工商银行股份有限公司 交易流量监控方法及系统
CN111404832A (zh) * 2019-01-02 2020-07-10 中国移动通信有限公司研究院 一种基于连续tcp链接的业务分类方法和装置
CN112612671A (zh) * 2020-12-16 2021-04-06 深圳前海微众银行股份有限公司 系统监控方法、装置、设备及存储介质
CN113326879A (zh) * 2021-05-31 2021-08-31 深圳前海微众银行股份有限公司 一种业务数据的监测方法及装置

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10275284B2 (en) * 2016-06-16 2019-04-30 Vmware, Inc. Datacenter resource allocation based on estimated capacity metric
CN107967624A (zh) * 2017-11-24 2018-04-27 四川长虹电器股份有限公司 一种业务活动销售预测模型的评价方法
TW201931265A (zh) * 2017-12-29 2019-08-01 林俊良 用於金融商品價格圖像處理之運算設備及其方法
CN109327439B (zh) * 2018-09-29 2021-04-23 武汉极意网络科技有限公司 业务请求数据的风险识别方法、装置、存储介质及设备
CN110059688B (zh) * 2019-03-19 2024-05-28 平安科技（深圳）有限公司 图片信息识别方法、装置、计算机设备和存储介质
CN110097451B (zh) * 2019-04-01 2023-10-03 中国银联股份有限公司 一种银行业务的监控方法及装置
CN110390465A (zh) * 2019-06-18 2019-10-29 深圳壹账通智能科技有限公司 业务数据的风控分析处理方法、装置和计算机设备
CN111310995A (zh) * 2020-02-12 2020-06-19 支付宝(杭州)信息技术有限公司 基于卷积神经网络的业务预测方法以及装置
CN111429161B (zh) * 2020-04-10 2023-10-10 杭州网易再顾科技有限公司 特征提取方法、特征提取装置、存储介质及电子设备
CN111798042A (zh) * 2020-06-19 2020-10-20 光合信号(深圳)信息科技有限公司 一种外汇市场预测方法及系统
CN111709190A (zh) * 2020-06-24 2020-09-25 国电联合动力技术有限公司 风电机组运行数据图像识别方法及装置
CN111953548B (zh) * 2020-08-21 2022-12-13 中国建设银行股份有限公司 一种业务影响判定方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10445738B1 (en) * 2018-11-13 2019-10-15 Capital One Services, Llc Detecting a transaction volume anomaly
CN111404832A (zh) * 2019-01-02 2020-07-10 中国移动通信有限公司研究院 一种基于连续tcp链接的业务分类方法和装置
CN111199018A (zh) * 2019-12-27 2020-05-26 东软集团股份有限公司 异常数据检测方法、装置、存储介质及电子设备
CN111382020A (zh) * 2020-03-06 2020-07-07 中国工商银行股份有限公司 交易流量监控方法及系统
CN112612671A (zh) * 2020-12-16 2021-04-06 深圳前海微众银行股份有限公司 系统监控方法、装置、设备及存储介质
CN113326879A (zh) * 2021-05-31 2021-08-31 深圳前海微众银行股份有限公司 一种业务数据的监测方法及装置

Also Published As

Publication number Publication date
CN113326879A (zh) 2021-08-31
CN113326879B (zh) 2024-07-19

Similar Documents

Publication Publication Date Title
WO2022252573A1 (zh) 一种业务数据的监测方法及装置
CN112633384B (zh) 基于图像识别模型的对象识别方法、装置和电子设备
WO2020062702A9 (zh) 短信发送的方法、装置、计算机设备和存储介质
CN111814910B (zh) 异常检测方法、装置、电子设备及存储介质
WO2021180009A1 (zh) 一种数据检测方法及装置
US10643113B2 (en) Return mail services
US20230049656A1 (en) Method of processing image, electronic device, and medium
CN113299073A (zh) 识别车辆违章停车的方法、装置、设备和存储介质
CN114969040A (zh) 一种数据展示方法、装置、电子设备及存储介质
CN113822110B (zh) 一种目标检测的方法和装置
KR20230051994A (ko) 엣지 클라우드 기반의 정보처리 종합물류 관제시스템
CN114138741A (zh) 一种历史数据分析平台
CN113158922A (zh) 基于yolo神经网络的车流量统计方法、装置及设备
CN111597889A (zh) 一种视频中目标移动的检测方法、装置及系统
CN115801361A (zh) 一种网络安全运维能力评估方法及系统
CN115278217A (zh) 图像画面检测方法、装置、电子设备及存储介质
CN113761082A (zh) 一种数据可视化方法、装置和系统
CN111047433B (zh) 用户数异常原因的分析方法、装置、服务器和存储介质
CN112991308A (zh) 一种图像质量的确定方法、装置、电子设备和介质
CN112115316A (zh) 一种分箱方法、装置、电子设备及存储介质
US11875354B2 (en) System and methods for identifying counterfeit checks using dimensional reduction of serial numbers
US11810164B1 (en) Computerized time-series analysis for inference of correlated input modifications
CN114494754B (zh) 车间杂物的检测方法、装置、终端及存储介质
KR20220106718A (ko) 이미지 인식 방법, 장치, 전자 기기, 저장 매체 및 컴퓨터 프로그램
CN107341430B (zh) 监控装置、监控方法及计数方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21943925

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 02.04.2024)