Application-based data processing method, device, computer equipment, and storage medium

Abstract

The present application relates to an application-based data processing method, apparatus, computer device, and storage medium. The method includes: obtaining a test data set and a control data set; filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set; determining a first variance of a first cumulative distribution function corresponding to the target test data, and a second variance of a second cumulative distribution function corresponding to the target control data; determining a first target dispersion degree of the quantile of the test data set at a preset value quantile based on the first variance; determining a second target dispersion degree of the quantile of the control data set at a preset value quantile based on the second variance; and determining the difference between the test data set and the control data set using the first target dispersion degree and the second target dispersion degree. Using this method, the accuracy of the determined data difference can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an application-based data processing method, apparatus, computer device, and storage medium. Background Art

[0002] With the development of science and technology, the iterative updates of applications are becoming more and more frequent. Therefore, R&D personnel can use the AB testing platform to determine the effectiveness of the iterative updates of applications. For example, taking an instant messaging application as an example, the original version of the instant messaging application can be used as experimental group A, and the updated instant messaging application can be used as experimental group B, so that users can be randomly guided to experimental group A or experimental group B. The AB testing platform can record the application operation data generated by users when using the original version of the instant messaging application to obtain A test data, and record the application operation data generated by users when using the updated version of the instant messaging application to obtain B test data. The A test data and the B test data are analyzed for differences to obtain difference results, and then the degree of superiority or inferiority between the updated version of the instant messaging application and the original version of the instant messaging application can be determined based on the difference results.

[0003] Currently, the two-sample t-test is commonly used to test whether the means of two groups of experimental data are significantly different, and the degree of difference is used to determine the degree of superiority. However, the mean is very sensitive to singular values ​​and asymmetric or highly skewed distributions. Therefore, when the two groups of experimental data include singular values, are asymmetric, or highly skewed, the difference determined by the mean is not accurate. Summary of the Invention

[0004] Based on this, it is necessary to provide an application-based data processing method, apparatus, computer device and storage medium that can improve the accuracy of the determined data differences in order to address the above technical issues.

[0005] A data processing method based on an application program, the method comprising:

[0006] Obtain test and control datasets corresponding to the target application;

[0007] Filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set;

[0008] determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data;

[0009] Determining, based on the first variance, a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile;

[0010] determining, based on the second variance, a second target dispersion between a quantile of the reference data set at a preset value quantile and a mean of the quantiles of the reference data set at the preset value quantile;

[0011] The difference between the test data set and the control data set is determined by using the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0012] A data processing device based on an application program, the device comprising:

[0013] a variance determination module configured to obtain a test data set and a control data set corresponding to a target application; filter target test data corresponding to a preset value from the test data set, and filter target control data corresponding to a preset value from the control data set; and determine a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data;

[0014] a dispersion determination module, configured to determine, based on the first variance, a first target dispersion degree between a quantile of the test data set at a preset value quantile and a mean of the quantiles of the test data set at the preset value quantile; and to determine, based on the second variance, a second target dispersion degree between a quantile of the control data set at a preset value quantile and the mean of the quantiles of the control data set at the preset value quantile;

[0015] A difference determination module is used to determine the difference between the test data set and the control data set based on the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0016] In one embodiment, the variance determination module further includes a data acquisition module configured to determine an updated version of the target application and an original version of the target application; acquire multiple first performance data generated when the updated version of the target application is running, and obtain a test data set based on the multiple first performance data; acquire multiple second performance data generated when the original version of the target application is running, and obtain a control data set based on the multiple second performance data.

[0017] In one embodiment, the variance determination module further includes a data screening module, which is configured to obtain a preset value and determine an interval range of the preset value; determine a first cumulative distribution function corresponding to the test data set and a second cumulative distribution function corresponding to the control data; screen out at least one target test data from the test data set so that the value of the first cumulative distribution function is within the interval range; and screen out at least one target control data from the control data set so that the value of the second cumulative distribution function is within the interval range.

[0018] In one embodiment, the variance determination module is further used to resample the test data set a preset number of times to obtain a first data set, and to resample the control data set a preset number of times to obtain a second data set; for each first data group in the first data set, it is determined that the data in the current first data group is less than or equal to the first probability value of the target test data, and based on each first probability value, the sum of the first probability values ​​corresponding to the current first data group is obtained; based on each sum of the first probability values, the first variance of the first cumulative distribution function corresponding to the target test data is obtained; for each second data group in the second data set, it is determined that the data in the current second data group is less than or equal to the second probability value of the target control data, and based on each second probability value, the sum of the second probability values ​​corresponding to the current second data group is obtained; based on the sum of each second probability value, the second variance of the second cumulative distribution function corresponding to the target control data is obtained.

[0019] In one embodiment, the variance determination module is further used to perform a logarithmic transformation on each of the test data in the test data set to obtain a corresponding logarithmic test data set, and divide the logarithmic test data set into a plurality of intermediate data groups; wherein the amount of logarithmic test data included in each of the intermediate data groups is not completely consistent; a first data group is obtained by integrating the plurality of intermediate data groups; an arrangement order of each logarithmic test data in the logarithmic test data set is updated, and the next round of resampling processing is entered, and the step of dividing the logarithmic test data set into a plurality of intermediate data groups is returned to continue execution until a preset number of resampling times is reached and the first data group is obtained.

[0020] In one embodiment, the variance determination module is further used to perform an averaging operation on the sum of each first probability value to obtain a first probability mean; determine a first difference between the sum of each first probability value and the first probability mean to obtain a first variance of a first cumulative distribution function corresponding to the target test data.

[0021] In one embodiment, the dispersion determination module further includes a first determination module for determining, based on the first variance, a first intermediate dispersion degree between the quantile of the test data set at the target test data and the quantile mean of the test data set at the target test data; and based on the first intermediate dispersion degree, determining a first target dispersion degree between the quantile of the test data set at the preset value quantile and the quantile mean of the test data set at the preset value quantile.

[0022] In one embodiment, the first determination module is further used to determine a corresponding first upper quantile and a first lower quantile based on the first variance; wherein, the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile satisfies a preset confidence condition; determine a second difference between the first upper quantile and the target test data, and determine a third difference between the target test data and the first lower quantile; when the second difference is greater than the third difference, determine a corresponding first intermediate dispersion degree based on the second difference; when the second difference is less than the third difference, determine a corresponding first intermediate dispersion degree based on the third difference.

[0023] In one embodiment, the first determination module is further used to perform linear interpolation processing on a data pair consisting of the first cumulative distribution function and the corresponding first intermediate dispersion degree to obtain a first corresponding relationship between the first cumulative distribution function and the first intermediate dispersion degree; based on the first corresponding relationship and the preset value, determine a first target dispersion degree between the quantile of the test data set at the preset value quantile line and the quantile mean of the test data set at the preset value quantile line.

[0024] In one embodiment, the dispersion determination module further includes a second determination module for determining, based on the second variance, a second intermediate dispersion degree between the quantile of the control data set at the target control data and the quantile mean of the control data set at the target control data; and based on the second intermediate dispersion degree, determining a second target dispersion degree between the quantile of the control data set at the preset value quantile and the quantile mean of the control data set at the preset value quantile.

[0025] In one embodiment, the difference determination module also includes a probability value determination module, which is used to determine the preset value quantile of the test data set and the preset value quantile of the control data set; based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set, determine the true probability value that there is a difference between the preset value quantile of the test data set and the preset value quantile of the control data set; when the true probability value is greater than or equal to a preset threshold, determine the difference between the test data set and the control data set based on the preset value quantile of the test data set and the preset value quantile of the control data set.

[0026] In one embodiment, the probability value determination module is further used to determine a fourth difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the difference between the preset value quantile of the test data set and the preset value quantile of the control data set; determine the standard deviation of the fourth difference based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set; and determine a true probability value of a difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the fourth difference and the standard deviation of the fourth difference.

[0027] In one embodiment, the test data set and the control data set are performance data collected for different versions of the target application; the difference determination module is further used to restore the updated version of the target application to the original version of the target application when it is determined based on the difference between the test data set and the control data set that the performance of the original version of the target application is better than the performance of the updated version of the target application; and to update the original version of the target application when it is determined based on the difference between the test data set and the control data set that the performance of the updated version of the target application is better than the performance of the original version of the target application.

[0028] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0029] Obtain test and control datasets corresponding to the target application;

[0030] Filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set;

[0031] determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data;

[0032] Determining, based on the first variance, a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile;

[0033] determining, based on the second variance, a second target dispersion between a quantile of the reference data set at a preset value quantile and a mean of the quantiles of the reference data set at the preset value quantile;

[0034] The difference between the test data set and the control data set is determined by using the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0036] Obtain test and control datasets corresponding to the target application;

[0037] Filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set;

[0038] determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data;

[0039] Determining, based on the first variance, a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile;

[0040] determining, based on the second variance, a second target dispersion between a quantile of the reference data set at a preset value quantile and a mean of the quantiles of the reference data set at the preset value quantile;

[0041] The difference between the test data set and the control data set is determined by using the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0042] A computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the following steps: obtaining a test data set and a control data set corresponding to a target application; filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set; determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data; determining a first target dispersion between a quantile of the test data set at a preset value quantile and a mean of the quantiles of the test data set at the preset value quantile based on the first variance; determining a second target dispersion between a quantile of the control data set at the preset value quantile and a mean of the quantiles of the control data set at the preset value quantile based on the second variance; and determining a difference between the test data set and the control data set using the first target dispersion and the second target dispersion; wherein the difference is used to optimize the target application.

[0043] The above-mentioned application-based data processing method, apparatus, computer device, storage medium, and computer program, by obtaining a test data set and a control data set corresponding to a target application, can filter out target test data corresponding to a preset value from the test data set, and filter out corresponding target control data from the control data set. By determining the target test data, the first variance of the first cumulative distribution function corresponding to the target test data can be determined. By determining the target control data, the second variance of the second cumulative distribution function corresponding to the target control data can be determined. Thus, based on the first variance, the corresponding first target dispersion can be determined, and based on the second variance, the corresponding second target dispersion can be determined. By determining the first target dispersion and the second target dispersion, the difference between the test data set and the control data set can be determined based on the first target dispersion and the second target dispersion, and the target application can be subsequently optimized based on this difference. Because the present application determines the difference between the test data set and the control data set using quantiles, compared to the traditional method of determining the difference using the mean, the present application can reduce the difference errors caused by singular values, asymmetric, and highly skewed distributions, thereby making the determined difference results more accurate.

[0044] In addition, since the present application uses the first variance and the second variance to determine the first target dispersion degree and the second target dispersion degree, compared with the traditional method of estimating the first target dispersion degree and the second target dispersion degree through density function, the present application not only improves the accuracy of the determined first target dispersion degree and the second target dispersion degree, but also reduces the amount of calculation required to determine the first target dispersion degree and the second target dispersion degree, thereby improving the data processing efficiency based on the application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 An application environment diagram of a data processing method based on an application program in one embodiment;

[0046] Figure 2 is a flowchart of a data processing method based on an application program in one embodiment;

[0047] Figure 3 Schematic diagram of data set collection in one embodiment;

[0048] Figure 4 is a schematic diagram of resampling in one embodiment;

[0049] Figure 5 A schematic flow chart of the steps of determining the first variance and the second variance in one embodiment;

[0050] Figure 6 Schematic diagram of a test platform in one embodiment;

[0051] Figure 7 is a flowchart of a data processing method based on an application program in a specific embodiment;

[0052] Figure 8 is a structural block diagram of a data processing device based on an application program in one embodiment;

[0053] Figure 9 is a structural block diagram of a data processing device based on an application program in another embodiment;

[0054] Figure 10 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0056] Figure 1 FIG. 1 is an application environment diagram of a data processing method based on an application program in an embodiment. Figure 1 , the application-based data processing method is applied to an application-based data processing system 100. The application-based data processing system 100 includes a terminal 102 and a server 104. The terminal 102 can be used alone to execute the application-based data processing method of the present application, and the terminal 102 and the server 104 can also be used in conjunction to execute the application-based data processing method of the present application. Taking the terminal 102 and the server 104 being used in conjunction to execute the application-based data processing method of the present application as an example, the terminal 102 can run a test data processing application, and the terminal 102 can obtain the operating data of two applications through the test data processing application, and use the operating data of one of the applications as a test data set, and the operating data of the other application as a control data set. The terminal 102 can send the control data set and the test data set to the server 104 through the test data processing application, so that the server 104 determines the difference between the test data set and the control data set, and returns the difference to the terminal 102 for corresponding display.

[0057] The server 104 may be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services. The terminal 102 may be, but is not limited to, a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, in-vehicle terminal, smart TV, etc. The terminal 102 and the server 104 may be connected directly or indirectly via wired or wireless communication, which is not limited in this application.

[0058] In one embodiment, Figure 2 As shown, a data processing method based on an application is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, and the computer device can be the above Figure 1 The terminal 102 or server 104 in the application program includes the following steps:

[0059] Step S202: Acquire a test data set and a control data set corresponding to the target application.

[0060] Among them, the target application refers to the application to be performance evaluated. For example, when the R&D personnel updates a certain function of the instant messaging application, in order to determine the update effect of the function, the instant messaging application can be used as the target application. For another example, when the R&D personnel updates the usage method of a certain virtual prop of the game application, in order to determine the update effect of the virtual prop, the game application can be used as the target application. For another example, when the R&D personnel updates the playback method of the multimedia playback application, in order to determine the update effect of the playback method, the multimedia playback application can be used as the target application. It is easy to understand that the target application is not only limited to instant messaging applications, game applications and multimedia playback applications, but can also be other applications that need to be performance evaluated, and the embodiments of this application will not be repeated here.

[0061] The test data set may include multiple test data, which refers to data generated when the updated version of the target application is running. The control data set may include multiple control data, which refers to data generated when the original version of the target application is running.

[0062] Specifically, when an AB test needs to be conducted on a target application, the computer device can determine the updated version and the original version corresponding to the target application, and obtain the data generated by the original version of the application during operation to obtain a test data set, and obtain the data generated by the updated version of the application during operation to obtain a control data set.

[0063] In one embodiment, reference Figure 3 The updated version of the target application can be used as the test group in an AB experiment, while the original version of the target application can be used as the control group. Users can be randomly assigned to either experimental group A or experimental group B. The data generated by experimental groups A and B during operation, such as user page views, object retention time, average viewing time per person, and basic interaction rate, can then be collected. Furthermore, the computer device uses the collected data for each evaluation metric corresponding to experimental group A as the test data set, and the data for each evaluation metric corresponding to experimental group B as the control data set. Figure 3 A schematic diagram of data set collection in one embodiment is shown.

[0064] In one embodiment, both the test dataset and the control dataset may include multiple data under multiple evaluation indicators. For example, both the test dataset and the control dataset may include the number of likes, forwardings, comments, document downloads, video views, consultations, camera transmission delay, page response time, application opening time, etc. When the test dataset and the control dataset include multiple different types of data, the computer device may compare the test dataset and the control dataset from the perspective of multiple evaluation indicator dimensions to determine the differences between the test dataset and the control dataset. For example, the computer device may compare the object retention time in the test dataset with the object retention time in the control dataset from the perspective of the object retention time evaluation indicator dimension to obtain a comparison result.

[0065] In one embodiment, obtaining a test data set and a control data set corresponding to a target application includes: determining an updated version of the target application and an original version of the target application; obtaining a plurality of first performance data generated when the updated version of the target application is running, and obtaining a test data set based on the plurality of first performance data; obtaining a plurality of second performance data generated when the original version of the target application is running, and obtaining a control data set based on the plurality of second performance data.

[0066] Specifically, the computer device may determine an updated version of a target application and an original version of the target application, and obtain multiple first performance data generated when the updated version of the target application is running, as well as multiple second performance data generated when the original version of the target application is running, and use the multiple first performance data as a test data set and the multiple second performance data as a second data set. The performance data refers to data reflecting the performance of the target application, such as the response time of the target application, the application opening time, etc.

[0067] In one embodiment, when the test data set and the control data set are obtained, the computer device may delete or filter the data in the test data set and the control data set to remove abnormal data or garbled data in the data set.

[0068] Step S204 , filtering out target test data corresponding to a preset value from the test data set, and filtering out target control data corresponding to a preset value from the control data set.

[0069] The preset value refers to the θ in the θ quantile of the random variable. For example, the preset value can be 1% of the 1% quantile, or 5% of the 5% quantile. The quantile can also be called the quantile. For any random variable or distribution, the quantile can be defined as follows: Assume that the cumulative distribution function of a random variable X is F(x), then the θ quantile of the distribution or random variable is the value that satisfies the condition inf{x|F(x)≥θ}, where 0<θ<1 and inf{} represents the infimum of the function value.

[0070] Specifically, because the mean sometimes does not well reflect the difference between the test dataset and the control dataset, the quantile line can be used to determine the difference between the test dataset and the control dataset. However, due to the influence of errors, the calculated result of "the quantile line of the test dataset is different from the quantile line of the control dataset" is not accurate. Therefore, before using the quantile line to determine the difference between the test dataset and the control dataset, it is also necessary to determine the probability value of the condition "the quantile line of the test dataset is different from the quantile line of the control dataset" being true.

[0071] In an embodiment of the present application, the probability value of the condition that "there is a difference between the quantiles of the test data set and the quantiles of the control data set" is true can be determined by a first target dispersion degree between the quantile of the test data set at the preset value quantile line and the quantile mean of the test data set at the preset value quantile line, and a second target dispersion degree between the quantile of the control data set at the preset value quantile line and the quantile mean of the control data set at the preset value quantile line.

[0072] Furthermore, since neither the first target dispersion degree nor the second target dispersion degree can be directly calculated using a formula, it is necessary to estimate the first target dispersion degree and the second target dispersion degree using preset values. In the process of estimating the first target dispersion degree and the second target dispersion degree using preset values, since the preset value and the first target dispersion degree have a meaningful correspondence relationship, and the preset value and the second target dispersion degree also have a meaningful correspondence relationship, and when the interval in which the preset value is located is sufficiently small, a linear relationship can be used to approximate the meaningful correspondence relationship. Therefore, in order to make the preset value and the first target dispersion degree or the second target dispersion degree have a linear relationship, and thus obtain the first target dispersion degree and the second target dispersion degree through this linear relationship, the computer device can cover the preset value with a small interval, and determine the target test data and the target control data through this small interval.

[0073] In one embodiment, target test data corresponding to preset values ​​are filtered out from a test data set, and target control data corresponding to preset values ​​are filtered out from a control data set, including: obtaining the preset values ​​and determining an interval range of the preset values; determining a first cumulative distribution function corresponding to the test data set and a second cumulative distribution function corresponding to the control data; filtering out at least one target test data from the test data set so that the value of the first cumulative distribution function is within the interval range; filtering out at least one target control data from the control data set so that the value of the second cumulative distribution function is within the interval range.

[0074] Specifically, the user can determine the preset value based on actual needs. For example, when calculating the 1% quantile, the preset value can be set to 1%. For another example, when calculating the 5% quantile, the preset value can be set to 5%. Furthermore, the computer device can determine the interval range of the preset value according to the preset interval determination rule. For example, when the preset value is 5%, the corresponding interval range can be determined as 5%-a<=5%<=5%+a. Where a is a sufficiently small value.

[0075] Furthermore, the computer device may determine a first cumulative distribution function corresponding to the test data set, and filter out at least one target test data from the test data set such that the value of the first cumulative distribution function falls within an interval range. The computer device may determine a second cumulative distribution function corresponding to the control data set, and filter out at least one target test data from the test data set such that the value of the second cumulative distribution function falls within an interval range.

[0076] In one embodiment, the first cumulative distribution function corresponding to the test data set is in, is the test data in the test data set. At this time, for the given preset value θ k , find satisfaction of in, The upper limit of the preset value range. The lower limit of the preset value range, satisfying of That is the target test data. Accordingly, the second cumulative distribution function can be determined as in, is the control data in the control data set. At this time, for the given preset value θ k , find satisfaction of in, The upper limit of the preset value range. The lower limit of the preset value range, satisfying of This is the target control data.

[0077] In the above embodiment, by determining the first cumulative distribution function and the second cumulative distribution function, the corresponding target test data can be determined based on the first cumulative distribution function, and the corresponding target control data can be determined based on the second cumulative distribution function, so that the difference between the test data set and the control data set can be subsequently determined based on the target test data and the target control data.

[0078] Step S206 , determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data.

[0079] The Cumulative Distribution Function (CDF), also known as the distribution function, is the integral of the probability density function, which can fully describe the probability distribution of a random variable X. The cumulative distribution function can be defined as F(x) = P(X≤x). Correspondingly, with the target test data The corresponding first cumulative distribution function can be defined as The first cumulative distribution function corresponding to the test data set is is any test data in the test data set. That is, the first cumulative distribution function corresponding to the test data set refers to the probability distribution function that describes the random sample obtained by taking the test data in the test data set as a random sample of the cumulative distribution function F(x); the first cumulative distribution function corresponding to the target test data refers to the probability distribution function that describes the random sample that is less than or equal to the target test data by taking the test data in the test data set as a random sample of the cumulative distribution function F(x).

[0080] Specifically, when a test data set and a control data set are obtained, the computer device may perform a preset number of resampling processes on the test data set to obtain a first data set including a plurality of first data groups, and perform a preset number of resampling processes on the control data set to obtain a second data set including a plurality of second data groups. Furthermore, for each first data group, the computer device may use the data in the current first data group as a random sample of the cumulative distribution function to obtain a probability distribution function that describes whether the random sample is less than or equal to the target test data, and for each second data group, the computer device may use the data in the current second data group as a random sample of the cumulative distribution function to obtain a probability distribution function that describes whether the random sample is less than or equal to the target control data.

[0081] The computer device can perform an averaging operation on the probability distribution function corresponding to each first data set describing that the random sample is less than or equal to the target test data to obtain the corresponding first probability mean, and obtain the first variance of the first cumulative distribution function of the test data set at the target test data based on the probability distribution function corresponding to each first data set describing that the random sample is less than or equal to the target test data and the first probability mean.

[0082] Correspondingly, the computer device can perform an averaging operation on the probability distribution function corresponding to each second data set describing that the random sample is less than or equal to the target control data to obtain the corresponding second probability mean, and obtain the second variance of the second cumulative distribution function of the control data set at the target control data based on the probability distribution function corresponding to each second data set describing that the random sample is less than or equal to the target control data and the second probability mean.

[0083] It is easy to understand that when there are multiple target test data, the computer device can determine a first variance of a first cumulative distribution function of the test data set at each target test data. When there are multiple target control data, the computer device can determine a second variance of a second cumulative distribution function of the control data set at each target control data.

[0084] In one embodiment, a test data set is resampled a preset number of times to obtain a first data set, including: performing a logarithmic transformation on each test data in the test data set to obtain a corresponding logarithmic test data set; dividing the logarithmic test data set into multiple intermediate data groups, and combining the multiple intermediate data groups to obtain a first data set; wherein the amount of logarithmic test data included in each intermediate data group is not completely consistent; updating the arrangement order of each logarithmic test data in the logarithmic test data set, and entering the next round of resampling processing, returning to the step of dividing the logarithmic test data set into multiple intermediate data groups and continuing to execute until the preset number of resampling times is reached, thereby obtaining the first data set.

[0085] Specifically, since the quantile line is invariant to the log transformation, that is, the θ quantile line of the original distribution is equal to the θ quantile line after taking the log of the original distribution, the corresponding formula can be expressed as log(μ f(x),θ )=μ f(log(x)),θ , where F(x) represents the original distribution, μ f(x),θ Represents the theta quantile of the original distribution, μ f(log(x)),θ represents the θ quantile after taking the logarithm of the original distribution. Therefore, in order to reduce the time complexity of the algorithm, reduce the amount of calculation, and improve the efficiency of application-based data processing, each test data in the test data set can be logarithmically transformed to obtain a logarithmic test data set containing multiple logarithmic test data.

[0086] Further, refer to Figure 4 , the computer device enters the first round of resampling processing, randomly divides the logarithmic test data in the logarithmic test data set into multiple intermediate data groups, and combines the intermediate data groups to obtain a first data group corresponding to the first round of resampling processing. The number of intermediate data groups can be a preset value, for example, it can be S groups, so that the computer device can randomly divide the multiple logarithmic test data into S groups of intermediate data groups. It is worth noting that the number of logarithmic test data included in each group of intermediate data groups may not be completely consistent. For example, the first group of intermediate data groups may include 8 logarithmic test data, the second group of intermediate data groups may include 15 logarithmic test data, and so on.

[0087] Furthermore, the computer device updates the order of the logarithmic test data in the logarithmic test data set and enters a second round of resampling processing. The logarithmic test data set after the order of the first round of resampling is divided into multiple intermediate data groups. Based on the multiple intermediate data groups, a first data group corresponding to the second round of resampling processing is obtained. This process is repeated until a predetermined number of resampling times is reached, thereby obtaining a first data group set including multiple first data groups. Figure 4 A schematic diagram of resampling in one embodiment is shown.

[0088] In one embodiment, the computer device may take the logarithmic value of each test data and retain a preset number of digits, for example, retaining d significant digits after the decimal point, thereby obtaining corresponding logarithmic test data.

[0089] In one embodiment, the computer device uses the plurality of intermediate data groups as the first data group.

[0090] In one embodiment, the computer device takes the log of each test data and retains d decimal places to obtain the logarithmic test data Among them, T is the test data set, K T represents the number of unique values ​​after the decimal point of the logarithmic test data set. Furthermore, the computer device performs a preset number of resampling processes on the logarithmic test data set. In each round of resampling, each logarithmic test data is randomly divided into s buckets, thereby obtaining a first data set V = {v1,…,v s}.in, For the first data group, v in the i-th bucket i Logarithmic test data included The number of

[0091] In one embodiment, the computer device may refer to the resampling method of the test data set for a preset number of times and perform corresponding resampling processing on the control data set to obtain multiple second data groups, thereby combining the multiple second data groups to obtain a second data group set corresponding to the control data set.

[0092] In the above embodiment, by performing logarithmic transformation on both the test data set and the control data set, the amount of calculation for subsequently determining the first target dispersion degree and the second target dispersion degree can be reduced, thereby improving the data processing efficiency based on the application program.

[0093] In one embodiment, reference Figure 5 , determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data, comprising:

[0094] S502, resampling the test data set a preset number of times to obtain a first data set, and resampling the control data set a preset number of times to obtain a second data set;

[0095] S504, for each first data group in the first data group set, determining a first probability value that the data in the current first data group is less than or equal to the target test data, and obtaining a sum of the first probability values ​​corresponding to the current first data group based on the first probability values;

[0096] S506, obtaining a first variance of a first cumulative distribution function corresponding to the target test data according to the sum of each first probability value;

[0097] S508, for each second data group in the second data group set, determining a second probability value that the data in the current second data group is less than or equal to the target control data, and obtaining a sum of the second probability values ​​corresponding to the current second data group based on each second probability value;

[0098] S510 , obtaining a second variance of a second cumulative distribution function corresponding to the target control data according to the sum of each second probability value.

[0099] Specifically, when the test data set is resampled a preset number of times to obtain a first data set, for each first data set in the first data set, the computer device can determine that the data in the current first data set is less than or equal to a first probability value of the target test data, and sum the first probability values ​​to obtain the sum of the first probability values ​​corresponding to the current first data set. Correspondingly, when the control data set is resampled a preset number of times to obtain a second data set, for each second data set in the second data set, the computer device can determine that the data in the current second data set is less than or equal to a second probability value of the target control data, and sum the second probability values ​​to obtain the sum of the second probability values ​​corresponding to the current second data set.

[0100] Furthermore, when the sum of the first probability values ​​corresponding to each first data group is obtained, the computer device may obtain a first variance of the first cumulative distribution function of the test data set at the target test data based on the sum of each first probability value. When the sum of the second probability values ​​is obtained, the computer device may obtain a second variance of the second cumulative distribution function of the control data set at the target control data based on the sum of each second probability value.

[0101] In one embodiment, the computer device may determine the sum of the first probability values ​​using the following formula:

[0102]

[0103] in, represents the sum of the first probability values ​​corresponding to the i-th first data group, a represents the a-th intermediate data group in the i-th first data group, and s represents the total number of intermediate data groups included in the i-th first data group. Represents the number of logarithmic test data included in the ath intermediate data group in the i-th first data group. represents the jth logarithmic test data in the i-th first data group, K T is the total number of logarithmic test data included in the i-th first data group. Represents the target test data. It is easy to understand that when the sum of the first probability values ​​is calculated using the above formula, the sum of the first probability values ​​can be expressed as a probability distribution function that describes whether the random sample obtained by taking the data in the first data group as a random sample of the cumulative distribution function is less than or equal to the target control data.

[0104] Accordingly, the computer device may determine the sum of the second probability values ​​using the following formula:

[0105]

[0106] in, represents the sum of the second probability values ​​corresponding to the i-th second data group, a represents the a-th intermediate data group in the i-th second data group, and s represents the total number of intermediate data groups included in the i-th second data group. Represents the number of logarithmic test data included in the ath intermediate data group in the i-th second data group. represents the jth logarithmic test data in the i-th second data group, K C is the total number of logarithmic test data included in the i-th second data group. Represents the target control data. It is easy to understand that when the sum of the second probability values ​​is calculated using the above formula, the sum of the second probability values ​​can be expressed as a probability distribution function that describes whether the random sample is less than or equal to the target control data, using the data in the second data group as a random sample of the cumulative distribution function.

[0107] In the above embodiment, by resampling both the test data set and the control data set, the first variance and the second variance can be obtained through the resampling results, and the true degree of difference between the test data set and the control data set can be accurately determined based on the first variance and the second variance.

[0108] In one embodiment, a first variance of a first cumulative distribution function corresponding to the target test data is obtained based on the sum of each first probability value, including: averaging the sum of each first probability value to obtain a first probability mean; determining a first difference between the sum of each first probability value and the first probability mean to obtain a first variance of the cumulative distribution function of the test data set at the target test data.

[0109] Specifically, when the sum of the first probability values ​​corresponding to each first data set is obtained, the computer device may perform a sum operation on the sum of the first probability values ​​to obtain a total probability value, and divide the total probability value by the total number of first data sets included in the first data set to obtain a first probability mean corresponding to the first data set. Further, the computer device determines a first difference between the sum of the first probability values ​​and the first probability mean to obtain a first variance of the first cumulative distribution function corresponding to the target test data, with the test data set being a random sample from the cumulative distribution function. For example, the computer device may determine a first difference between the sum of each first probability value and the first probability mean, and average the first differences to obtain the first variance.

[0110] In one embodiment, the computer device may determine the first variance of the cumulative distribution function corresponding to the target test data using the following formula:

[0111]

[0112]

[0113] in, Represents the test data set in the target test data B represents the total number of first data sets included in the first data set. Represents the sum of the first probability values ​​corresponding to the i-th first data group.

[0114] In one embodiment, after obtaining the sum of the second probability values ​​corresponding to each second data group, the computer device may average the sums of the second probability values ​​to obtain a second probability mean. The computer device then determines a fifth difference between the sum of the second probability values ​​and the second probability mean, and obtains a second variance of the second cumulative distribution function corresponding to the target control data based on the fifth difference. It will be readily understood that the computer device may determine the second variance based on the specific implementation for determining the first variance.

[0115] In the above embodiment, by determining the first variance and the second variance, the first target dispersion degree and the second target dispersion degree can be estimated based on the first variance and the second variance.

[0116] Step S208 : determining a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile based on the first variance.

[0117] Specifically, in order to determine the probability value of the condition "there is a difference between the quantile corresponding to the test data set and the quantile corresponding to the control data set" being true, the computer device also needs to determine a first target dispersion degree between the quantile of the test data set at the preset value quantile and the mean of the quantiles of the test data set at the preset value quantile, and determine a second target dispersion degree between the quantile of the control data set at the preset value quantile and the mean of the quantiles of the control data set at the preset value quantile. Subsequently, based on the first target dispersion degree and the second target dispersion degree, it can be determined whether the condition "there is a difference between the preset value quantile of the test data and the preset value quantile of the control data" is true. If it is determined to be true, the difference between the test data and the control data is determined based on the difference between the preset value quantile of the test data and the preset value quantile of the control data. In this way, the impact of errors on the determined difference can be reduced, thereby improving the accuracy of data processing.

[0118] In one embodiment, based on the first variance, determining a first target dispersion degree between the quantile of the test data set at the preset value quantile and the quantile mean of the test data set at the preset value quantile includes: based on the first variance, determining a first intermediate dispersion degree between the quantile of the test data set at the target test data and the quantile mean of the test data set at the target test data; based on the first intermediate dispersion degree, determining a first target dispersion degree between the quantile of the test data set at the preset value quantile and the quantile mean of the test data set at the preset value quantile.

[0119] The first intermediate dispersion reflects the degree of dispersion between the quantile of the test dataset at the target test data and its mean. In a specific embodiment, the first intermediate dispersion may be the standard deviation of the quantile of the test dataset at the target test data. The first target dispersion reflects the degree of dispersion between the quantile of the test dataset at the preset value quantile and its mean. In a specific embodiment, the first target dispersion may be the standard deviation of the quantile of the test dataset at the preset value quantile.

[0120] Specifically, since the first target dispersion degree is difficult to calculate using a formula, and when the interval of the preset value is small enough, the preset value will have a linear relationship with the first target dispersion degree, therefore, the computer device can estimate the first target dispersion degree through the linear relationship between the preset value and the first target dispersion degree. k The range is and is the first cumulative probability distribution corresponding to the target test data. Therefore, the computer device can estimate the first target dispersion by comparing the first cumulative distribution function with the target test data. For example, the computer device can determine a target function that is linearly related to the first cumulative distribution function, and use the value of the target function as the value of the first target dispersion.

[0121] Furthermore, since the first intermediate dispersion degree between the quantile of the test data set at the target test data and the mean of the quantile of the test data set at the target test data can be determined by the first cumulative distribution function, the first intermediate dispersion degree can be used as the above-mentioned objective function, thereby estimating the value of the first target dispersion degree by the value of the first intermediate dispersion degree.

[0122] In this embodiment, by determining the first intermediate dispersion degree, the first target dispersion degree can be estimated by the first intermediate dispersion degree, so that the difference between the test dataset and the control dataset can be determined based on the first target dispersion degree.

[0123] In one embodiment, based on the first variance, determining a first intermediate degree of dispersion between the quantile of the test data set at the target test data and the quantile mean of the test data set at the target test data includes: determining a corresponding first upper quantile and a first lower quantile according to the first variance; wherein the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile satisfies a preset confidence condition; determining a second difference between the first upper quantile and the target test data, and determining a third difference between the target test data and the first lower quantile; when the second difference is greater than the third difference, determining the corresponding first intermediate degree of dispersion according to the second difference; when the second difference is less than the third difference, determining the corresponding first intermediate degree of dispersion according to the third difference.

[0124] Specifically, to determine the first intermediate dispersion, the computer device may determine the first upper quantile and the first lower quantile based on the first variance, wherein the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile satisfies a preset confidence condition. For example, the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile is (1-α), that is, Where α is the confidence level, is the first lower quantile, is the first upper quantile, μ θ is the quantile line corresponding to the test data set, that is, μ θ is the θ quantile line of the test data set.

[0125] Furthermore, the computer device may determine a second difference between the first upper quantile and the target test data, and a third difference between the target test data and the first lower quantile. When the second difference is greater than the third difference, the computer device may determine a corresponding first intermediate dispersion based on the second difference. When the second difference is less than the third difference, the computer device may determine a corresponding first intermediate dispersion based on the third difference. When the second difference is equal to the third difference, the computer device may determine the first intermediate dispersion based on the second difference or the third difference.

[0126] In one embodiment, the first lower quantile line can be determined by the following formula:

[0127]

[0128] in, is the first variance. Target test data. is the 1- / 2 quantile of the standard normal distribution. is the first cumulative distribution function corresponding to the target test data. is the cumulative distribution function of the test data set.

[0129] Accordingly, the computer device can determine the first upper quantile line by the following formula:

[0130]

[0131] In one embodiment, the computer device may determine the first intermediate dispersion level by the following formula:

[0132]

[0133] In the above embodiment, by determining the upper quantile and the lower quantile, the corresponding intermediate dispersion degree can be determined based on the upper quantile and the lower quantile, so that the difference between the test dataset and the control dataset can be determined based on the intermediate dispersion degree.

[0134] In one embodiment, based on the first intermediate dispersion degree, a first target dispersion degree between the quantile of the test data set at the preset value quantile and the mean of the quantiles of the test data set at the preset value quantile is determined, including: performing linear interpolation processing on a data pair consisting of a first cumulative distribution function and a corresponding first intermediate dispersion degree to obtain a first corresponding relationship between the first cumulative distribution function and the first intermediate dispersion degree; and determining, according to the first corresponding relationship and a preset value, the first target dispersion degree between the quantile of the test data set at the preset value quantile and the mean of the quantiles of the test data set at the preset value quantile.

[0135] Specifically, because the first cumulative distribution function and the first intermediate dispersion are linearly correlated, and when the preset value lies within the small interval defined by the first cumulative distribution function, the preset value is linearly correlated with the first target dispersion. Therefore, linear interpolation can be performed on data pairs consisting of the first cumulative distribution function and the corresponding first intermediate dispersion to obtain a first correspondence between the first cumulative distribution function and the first intermediate dispersion. The preset value can then be substituted into this first correspondence to obtain the first target dispersion.

[0136] For example, for the linear interpolation formula: Computer equipment can be used to in As x0 in the linear interpolation formula, As y0, in As x1, As y1, and the preset value θ k As x, to get the y value, and the y value as the corresponding first target dispersion That is, the y value is used as the variance of the quantile line of the test data set at the preset value quantile line.

[0137] In the above embodiment, by performing linear interpolation processing on the data pairs, the corresponding first target dispersion degree can be quickly obtained based on the linear interpolation result, thereby not only reducing the calculation amount of the first target dispersion degree, but also improving the efficiency of determining the first target dispersion degree.

[0138] Step S210 : determining a second target dispersion between a quantile of the reference data set at a preset value quantile and a quantile mean of the reference data set at the preset value quantile based on the second variance.

[0139] Specifically, in order to determine the probability value of the condition that "there is a difference between the quantile corresponding to the test data set and the quantile corresponding to the control data set" is true, the computer device also needs to determine the second target dispersion degree between the quantile of the control data set at the preset value quantile and the quantile mean of the control data set at the preset value quantile, so that it can subsequently determine whether the condition that "there is a difference between the preset value quantile of the test data and the preset value quantile of the control data" is true based on the first target dispersion degree and the second target dispersion degree, and when it is determined to be true, the difference between the test data and the control data is determined based on the difference between the preset value quantile of the test data and the preset value quantile of the control data.

[0140] In one embodiment, determining a second target dispersion degree between the quantile of the control data set at a preset value quantile and the quantile mean of the control data set at the preset value quantile based on the second variance includes: determining a second intermediate dispersion degree between the quantile of the control data set at the target control data and the quantile mean of the control data set at the target control data based on the second variance; and determining a second target dispersion degree between the quantile of the control data set at the preset value quantile and the quantile mean of the control data set at the preset value quantile based on the second intermediate dispersion degree.

[0141] Specifically, when the second variance of the second cumulative distribution function is determined, the computer device may determine a corresponding second upper quantile and second lower quantile based on the second variance; wherein the probability value of the quantile corresponding to the control data set being between the second upper quantile and the second lower quantile satisfies a preset confidence condition. The computer device determines a sixth difference between the second upper quantile and the target control data, and determines a seventh difference between the target test data and the second lower quantile; when the sixth difference is greater than the seventh difference, the computer device determines a corresponding second intermediate dispersion based on the sixth difference; when the sixth difference is less than the seventh difference, the computer device determines a corresponding second intermediate dispersion based on the seventh difference; and when the sixth difference is the same as the seventh difference, the computer device determines a corresponding second intermediate dispersion based on either the sixth difference or the seventh difference.

[0142] Furthermore, the computer device performs linear interpolation on data pairs consisting of the second cumulative distribution function and the corresponding second intermediate dispersion to obtain a second correspondence between the second cumulative distribution function and the second intermediate dispersion. Based on the second correspondence and a preset value, the computer device determines a second target dispersion between the quantile of the reference data set at the preset value quantile and the mean of the quantiles of the reference data set at the preset value quantile. It will be readily understood that the specific process for determining the second intermediate dispersion can refer to the process for determining the first intermediate dispersion described above, and the specific process for determining the second target dispersion can refer to the process for determining the first target dispersion described above.

[0143] In this embodiment, by determining the second intermediate dispersion degree, the second target dispersion degree can be quickly estimated based on the second intermediate dispersion degree, thereby greatly improving the efficiency of determining the second target dispersion degree.

[0144] Step S212 : determining the difference between the test data set and the control data set by using the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0145] Specifically, by determining the first target dispersion degree and the second target dispersion degree, the true degree of difference between the test data set and the control data set can be determined based on the first target dispersion degree and the second target dispersion degree. Therefore, when it is determined that there is a real difference between the test data set and the control data set, the difference between the test data set and the control data set is determined based on the preset value quantile line of the test data set and the preset value quantile line of the control data set, and the target application is optimized based on the difference.

[0146] In one embodiment, the difference between a test data set and a control data set is determined by using a first target dispersion degree and a second target dispersion degree, including: determining a preset value quantile of the test data set and a preset value quantile of the control data set; determining a true probability value of a difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set; when the true probability value is greater than or equal to a preset threshold, determining the difference between the test data set and the control data set based on the preset value quantile of the test data set and the preset value quantile of the control data set.

[0147] Specifically, the computer device may determine a preset value quantile of the test dataset and a preset value quantile of the control dataset, and determine a true probability value of a difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset based on the first target dispersion degree, the second target dispersion degree, and the difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset. Further, when the true probability value is greater than or equal to a preset threshold, it can be considered that there is indeed a difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset, and the difference is not due to an error. In this case, the computer device determines the difference between the test dataset and the control dataset based on the preset value quantile of the test dataset and the preset value quantile of the control dataset.

[0148] In one embodiment, the computer device may determine the preset value quantile of the test data set using the following formula:

[0149]

[0150] in, is the cumulative distribution function of the test data set, θ k is the default value.

[0151] In one embodiment, the computer device may determine the preset value quantile of the reference data set using the following formula:

[0152]

[0153] in, is the cumulative distribution function of the control data set, θ k is the default value.

[0154] In one embodiment, based on a first target dispersion degree, a second target dispersion degree, a preset value quantile of a test data set, and a preset value quantile of a control data set, determining a true probability value that there is a difference between the preset value quantile of a test data set and the preset value quantile of a control data set includes: determining a fourth difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on a difference between the preset value quantile of the test data set and the preset value quantile of the control data set; determining a standard deviation of the fourth difference based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set; and determining a true probability value that there is a difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the fourth difference and the standard deviation of the fourth difference.

[0155] Specifically, when the preset value quantile of the test data set and the preset value quantile of the control data set are obtained, the computer device can determine the difference between the preset value quantile of the test data set and the preset value quantile of the control data set, and divide the difference by the preset value quantile of the control data set to obtain the fourth difference between the preset value quantile of the test data set and the preset value quantile of the control data set.

[0156] Furthermore, the computer device superimposes the square of the first target dispersion degree and the square of the second target dispersion degree, and then performs a square root operation to obtain a first probability median value. Furthermore, the computer device divides the preset value quantile of the test data set by the preset value quantile of the control data set to obtain a quantile ratio. The computer device multiplies the first probability median value by the quantile ratio to obtain a standard deviation of the fourth difference.

[0157] Furthermore, the computer device determines, based on the fourth difference and the standard deviation of the fourth difference, a probability value of the observed data when the condition "there is no difference between the preset value quantile of the test data set and the preset value quantile of the control data set" is met. Based on the probability value of the observed data, the computer device determines a true probability value of a difference between the preset value quantile of the test data set and the preset value quantile of the control data set. For example, the computer device subtracts the probability value of the observed data from 1 to obtain the corresponding true probability value.

[0158] In one embodiment, the computer device can determine the fourth difference Δ between the preset value quantile line of the test data set and the preset value quantile line of the control data set by the following formula: i %:

[0159]

[0160] in, is the preset value quantile line of the test data set, is the preset value quantile line of the control data set.

[0161] In one embodiment, the computer device may determine the standard deviation of the fourth difference by the following formula:

[0162]

[0163] in, is the first target dispersion degree, The second target dispersion degree. is the preset value quantile line of the test data set, is the preset value quantile line of the control data set.

[0164] In one embodiment, the computer device may determine the probability value of observing the data when the condition that "there is no difference between the preset value quantile line of the test data set and the preset value quantile line of the control data set" is met by the following formula:

[0165]

[0166] Among them, Δ i % is the fourth difference, is the standard deviation of the fourth difference, and Φ represents the distribution function of the standard normal distribution.

[0167] In one embodiment, when p k When the value is less than or equal to 0.05, it can be determined that there is a large difference between the preset value quantile line of the test data set and the preset value quantile line of the control data set. At this time, the computer device determines the difference between the test data set and the control data set based on the fourth difference.

[0168] In the above embodiment, by determining the fourth difference and the standard deviation of the fourth difference, the true probability value can be accurately obtained based on the fourth difference and the standard deviation of the fourth difference. Therefore, when the difference between the preset value quantile line of the test data set and the preset value quantile line of the control data set is determined based on the true probability value, and it is not a difference obtained due to an error, the difference between the test data set and the control data set is determined based on the fourth difference, which can greatly improve the accuracy of the determined difference between the test data set and the control data set.

[0169] In one embodiment, the test data set and the control data set are performance data collected for different versions of the target application; the above method also includes: when it is determined based on the difference between the test data set and the control data set that the performance of the original version of the target application is better than the performance of the updated version of the target application, restoring the updated version of the target application to the original version of the target application; when it is determined based on the difference between the test data set and the control data set that the performance of the updated version of the target application is better than the performance of the original version of the target application, updating the original version of the target application.

[0170] Specifically, both the test dataset and the control dataset may be performance data collected for different versions of the target application. For example, the test dataset may be performance data collected for an updated version of the target application, and the control dataset may be performance data collected for an original version of the target application. When, based on a preset value quantile of the test dataset and a preset quantile of the target dataset, it is determined that the data in the test dataset is superior to the data in the control dataset, it can be considered that the performance of the original version of the target application is superior to that of the updated version of the target application. In this case, the computer device restores the updated version of the target application to the original version of the target application.

[0171] For example, when the test data set includes the response time of an updated version of the target application, the control data set includes the response time of the original version of the target application, and based on the preset value quantile of the test data set and the preset quantile of the target data set, it is determined that the probability value of the response time of the updated version of the target application falling into [x1, positive infinity] is greater than the probability value of the response time of the original version of the target application falling into [x1, positive infinity], it can be determined that the performance of the original version of the target application is better than the performance of the updated version of the target application. At this time, the computer device can perform restoration processing on the updated version of the target application.

[0172] Furthermore, when it is determined that the performance of the updated version of the target application is better than the performance of the original version of the target application based on the difference between the test dataset and the control dataset, the computer device may update the original version of the target application.

[0173] In one embodiment, when it is determined that the performance of the updated version of the target application is better than the performance of the original version of the target application based on the difference between the test data set and the control data set, it can be considered that the update direction of the target application is correct, and the computer device can keep the updated version of the target application unchanged.

[0174] Since the embodiment of the present application can determine the true probability value of the difference between the test data set and the control data set, when it is determined that there is a real difference between the test data set and the control data set based on the true probability value, the target application is optimized based on the difference between the test data set and the control data set. Therefore, the accuracy of the optimization processing of the target application can be greatly improved.

[0175] In the above-mentioned application-based data processing method, by obtaining a test data set and a control data set corresponding to the target application, target test data corresponding to a preset value can be screened out from the test data set, and corresponding target control data can be screened out from the control data set. By determining the target test data, the first variance of the first cumulative distribution function corresponding to the target test data can be determined, and by determining the target control data, the second variance of the second cumulative distribution function corresponding to the target control data can be determined. Thus, based on the first variance, the corresponding first target dispersion degree can be determined, and by the second variance, the corresponding second target dispersion degree can be determined. By determining the first target dispersion degree and the second target dispersion degree, the difference between the test data set and the control data set can be determined based on the first target dispersion degree and the second target dispersion degree, so that the target application can be subsequently optimized based on the difference. Because the present application determines the difference between the test data set and the control data set by quantiles, compared to the traditional method of determining the difference by means, the present application can reduce the difference errors caused by singular values, asymmetric and highly skewed distributions, thereby making the determined difference results more accurate.

[0176] In addition, since the present application uses the first variance and the second variance to determine the first target dispersion degree and the second target dispersion degree, compared with the traditional method of estimating the first target dispersion degree and the second target dispersion degree through density function, the present application not only improves the accuracy of the determined first target dispersion degree and the second target dispersion degree, but also reduces the amount of calculation required to determine the first target dispersion degree and the second target dispersion degree, thereby improving the data processing efficiency based on the application.

[0177] This application also provides an application scenario, which applies the above-mentioned application-based data processing method. Specifically, the application of the application-based data processing method in this application scenario is as follows:

[0178] When a developer improves function A in the original version of an instant messaging application to obtain an updated version of the instant messaging application, the developer can obtain the response time of the original version of the instant messaging application and the response time of the updated version of the instant messaging application, and use the response time of the original version of the instant messaging application as a control data set and the response time of the updated version of the instant messaging application as a test data set. Figure 6 , R&D personnel can input the control data set and the test data set into the experimental platform in the computer device, and the experimental platform will process the test data set and the control data set to determine the differences between the test data set and the control data set, and display the differences accordingly, so that the R&D personnel can determine whether to retain the improvement of function A in the instant messaging application based on the displayed differences. Figure 6 A schematic diagram of a test platform in one embodiment is shown.

[0179] This application also provides an application scenario, which applies the above-mentioned application-based data processing method. Specifically, the application of the application-based data processing method in this application scenario is as follows:

[0180] When comparing application A and application B, the computer device can obtain data generated by the execution of application A to obtain test data, and obtain a data set generated by the execution of application B to obtain a reference data set. Furthermore, the computer device can process the test and reference data sets according to the aforementioned application-based data processing method to determine the differences between the test and reference data sets, and display these differences accordingly. This allows developers to determine the performance differences between application A and application B based on the displayed differences. Application A and application B can be of the same or different types.

[0181] The above application scenarios are merely illustrative. It will be understood that the application of the business-related data reporting methods provided in the various embodiments of the present application is not limited to the above scenarios.

[0182] In a specific embodiment, referring to Figure 7 , the above application-based data processing method further includes the following steps:

[0183] S702: Determine the target application of the updated version and the target application of the original version.

[0184] S704, obtaining multiple first performance data generated when the updated version of the target application is running, and obtaining a test data set based on the multiple first performance data; obtaining multiple second performance data generated when the original version of the target application is running, and obtaining a control data set based on the multiple second performance data.

[0185] S706 , obtaining a preset value, determining an interval range of the preset value, and determining a first cumulative distribution function corresponding to the test data set and a second cumulative distribution function corresponding to the control data set.

[0186] S708 , screening out at least one target test data from the test data set so that the value of the first cumulative distribution function is within the interval range, and screening out at least one target control data from the control data set so that the value of the second cumulative distribution function is within the interval range.

[0187] S710 , resampling the test data set a preset number of times to obtain a first data set, and resampling the control data set a preset number of times to obtain a second data set.

[0188] S712, for each first data group in the first data group set, determine the first probability value that the data in the current first data group is less than or equal to the target test data, and based on each first probability value, obtain the sum of the first probability values ​​corresponding to the current first data group; based on the sum of each first probability value, obtain the first variance of the first cumulative distribution function corresponding to the target test data.

[0189] S714. For each second data group in the second data group set, determine a second probability value that the data in the current second data group is less than or equal to the target control data, and based on each second probability value, obtain the sum of the second probability values ​​corresponding to the current second data group; and based on the sum of each second probability value, obtain a second variance of the second cumulative distribution function corresponding to the target control data.

[0190] S716. Based on the first variance, determine a first intermediate dispersion degree between the quantile of the test data set at the target test data and the quantile mean of the test data set at the target test data; based on the first intermediate dispersion degree, determine a first target dispersion degree between the quantile of the test data set at the preset value quantile and the quantile mean of the test data set at the preset value quantile.

[0191] S718. Based on the second variance, determine a second intermediate dispersion degree between the quantile of the control data set at the target control data and the mean of the quantiles of the control data set at the target control data; based on the second intermediate dispersion degree, determine a second target dispersion degree between the quantile of the control data set at the preset value quantile and the mean of the quantiles of the control data set at the preset value quantile.

[0192] S720: Determine a preset value quantile line of the test data set and a preset value quantile line of the control data set.

[0193] S722, based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set, determine a true probability value that there is a difference between the preset value quantile of the test data set and the preset value quantile of the control data set.

[0194] S724 , when the true probability value is greater than or equal to the preset threshold, determining the difference between the test dataset and the control dataset based on the preset value quantile of the test dataset and the preset value quantile of the control dataset.

[0195] S726 , when it is determined based on the difference between the test data set and the control data set that the performance of the original version of the target application is better than that of the updated version of the target application, restore the updated version of the target application to the original version of the target application.

[0196] S728 , when it is determined based on the difference between the test data set and the control data set that the performance of the updated version of the target application is better than the performance of the original version of the target application, the original version of the target application is updated.

[0197] In the above-mentioned application-based data processing method, by obtaining a test data set and a control data set corresponding to the target application, target test data corresponding to a preset value can be screened out from the test data set, and corresponding target control data can be screened out from the control data set. By determining the target test data, the first variance of the first cumulative distribution function corresponding to the target test data can be determined, and by determining the target control data, the second variance of the second cumulative distribution function corresponding to the target control data can be determined. Thus, based on the first variance, the corresponding first target dispersion degree can be determined, and by the second variance, the corresponding second target dispersion degree can be determined. By determining the first target dispersion degree and the second target dispersion degree, the difference between the test data set and the control data set can be determined based on the first target dispersion degree and the second target dispersion degree, so that the target application can be subsequently optimized based on the difference. Because the present application determines the difference between the test data set and the control data set by quantiles, compared to the traditional method of determining the difference by means, the present application can reduce the difference errors caused by singular values, asymmetric and highly skewed distributions, thereby making the determined difference results more accurate.

[0198] It should be understood that although Figure 2 、 Figure 5 、 Figure 7The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 2 、 Figure 5 、 Figure 7 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0199] In one embodiment, Figure 8 As shown, an application-based data processing device 800 is provided. The device can be a software module or a hardware module, or a combination of both, and can be part of a computer device. The device specifically includes: a variance determination module 802, a dispersion determination module 804, and a difference determination module 806, wherein:

[0200] The variance determination module 802 is used to obtain a test data set and a control data set corresponding to the target application; filter out target test data corresponding to a preset value from the test data set, and filter out target control data corresponding to the preset value from the control data set; and determine a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data.

[0201] The dispersion determination module 804 is used to determine a first target dispersion degree between the quantile of the test data set at the preset value quantile and the mean of the quantiles of the test data set at the preset value quantile based on the first variance; and to determine a second target dispersion degree between the quantile of the control data set at the preset value quantile and the mean of the quantiles of the control data set at the preset value quantile based on the second variance.

[0202] The difference determination module 806 is configured to determine the difference between the test data set and the control data set based on the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

[0203] In one embodiment, reference Figure 9The variance determination module 802 further includes a data acquisition module 8021 for determining an updated version of the target application and an original version of the target application; acquiring a plurality of first performance data generated when the updated version of the target application is running, and obtaining a test data set based on the plurality of first performance data; acquiring a plurality of second performance data generated when the original version of the target application is running, and obtaining a control data set based on the plurality of second performance data.

[0204] In one embodiment, the variance determination module 802 also includes a data screening module 8022, which is used to obtain a preset value and determine the interval range of the preset value; determine a first cumulative distribution function corresponding to the test data set and a second cumulative distribution function corresponding to the control data; screen out at least one target test data from the test data set so that the value of the first cumulative distribution function is within the interval range; and screen out at least one target control data from the control data set so that the value of the second cumulative distribution function is within the interval range.

[0205] In one embodiment, the variance determination module 802 is further configured to perform a resampling process on the test data set a preset number of times to obtain a first data set, and to perform a resampling process on the control data set a preset number of times to obtain a second data set; for each first data group in the first data set, determine that the data in the current first data group is less than or equal to a first probability value of the target test data, and obtain the sum of the first probability values ​​corresponding to the current first data group based on each first probability value; obtain the first variance of the first cumulative distribution function corresponding to the target test data based on the sum of each first probability value; for each second data group in the second data set, determine that the data in the current second data group is less than or equal to a second probability value of the target control data, and obtain the sum of the second probability values ​​corresponding to the current second data group based on each second probability value; and obtain the second variance of the second cumulative distribution function corresponding to the target control data based on the sum of each second probability value.

[0206] In one embodiment, the variance determination module 802 is further configured to perform a logarithmic transformation on each test data in the test data set to obtain a corresponding logarithmic test data set, and divide the logarithmic test data set into a plurality of intermediate data groups; wherein the amount of logarithmic test data included in each intermediate data group is not completely consistent; a first data group is obtained by integrating the plurality of intermediate data groups; an arrangement order of each logarithmic test data in the logarithmic test data set is updated, and the next round of resampling processing is entered, and the step of dividing the logarithmic test data set into a plurality of intermediate data groups is returned to continue execution until a preset number of resampling times is reached and the first data group is obtained.

[0207] In one embodiment, the variance determination module 802 is further used to perform an averaging operation on the sum of the first probability values ​​to obtain a first probability mean; determine a first difference between the sum of the first probability values ​​and the first probability mean to obtain a first variance of the first cumulative distribution function corresponding to the target test data.

[0208] In one embodiment, the dispersion determination module 804 also includes a first determination module 8041, which is used to determine, based on the first variance, a first intermediate dispersion degree between the quantile of the test data set at the target test data and the quantile mean of the test data set at the target test data; based on the first intermediate dispersion degree, determine a first target dispersion degree between the quantile of the test data set at the preset value quantile and the quantile mean of the test data set at the preset value quantile.

[0209] In one embodiment, the first determination module 8041 is further used to determine the corresponding first upper quantile and first lower quantile based on the first variance; wherein, the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile satisfies a preset confidence condition; determine the second difference between the first upper quantile and the target test data, and determine the third difference between the target test data and the first lower quantile; when the second difference is greater than the third difference, determine the corresponding first intermediate dispersion degree based on the second difference; when the second difference is less than the third difference, determine the corresponding first intermediate dispersion degree based on the third difference.

[0210] In one embodiment, the first determination module 8041 is further used to perform linear interpolation processing on a data pair consisting of a first cumulative distribution function and a corresponding first intermediate dispersion degree to obtain a first corresponding relationship between the first cumulative distribution function and the first intermediate dispersion degree; based on the first corresponding relationship and a preset value, determine a first target dispersion degree between the quantile of the test data set at the preset value quantile line and the quantile mean of the test data set at the preset value quantile line.

[0211] In one embodiment, the dispersion determination module 804 also includes a second determination module 8042, which is used to determine, based on the second variance, a second intermediate dispersion degree between the quantile of the control data set at the target control data and the quantile mean of the control data set at the target control data; based on the second intermediate dispersion degree, determine a second target dispersion degree between the quantile of the control data set at the preset value quantile and the quantile mean of the control data set at the preset value quantile.

[0212] In one embodiment, the difference determination module 806 also includes a probability value determination module 8061, which is used to determine the preset value quantile of the test data set and the preset value quantile of the control data set; based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set, determine the true probability value that there is a difference between the preset value quantile of the test data set and the preset value quantile of the control data set; when the true probability value is greater than or equal to the preset threshold value, determine the difference between the test data set and the control data set based on the preset value quantile of the test data set and the preset value quantile of the control data set.

[0213] In one embodiment, the probability value determination module 8061 is also used to determine the fourth difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the difference between the preset value quantile of the test data set and the preset value quantile of the control data set; determine the standard deviation of the fourth difference based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set; determine the true probability value of the difference between the preset value quantile of the test data set and the preset value quantile of the control data set based on the fourth difference and the standard deviation of the fourth difference.

[0214] In one embodiment, the test data set and the control data set are performance data collected for different versions of the target application; the difference determination module 808 is also used to restore the updated version of the target application to the original version of the target application when it is determined based on the difference between the test data set and the control data set that the performance of the original version of the target application is better than the performance of the updated version of the target application; and update the original version of the target application when it is determined based on the difference between the test data set and the control data set that the performance of the updated version of the target application is better than the performance of the original version of the target application.

[0215] For the specific definition of the application-based data processing device, please refer to the definition of the application-based data processing method above, and will not be repeated here. The various modules in the above-mentioned application-based data processing device can be implemented in whole or in part through software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above-mentioned modules.

[0216] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 10As shown. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the operating data of the application. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, an application-based data processing method is implemented.

[0217] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0218] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0219] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of each of the above-described method embodiments.

[0220] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0221] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0222] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A data processing method based on an application, characterized in that: The method comprises: Obtain test and control datasets corresponding to the target application; Filtering target test data corresponding to a preset value from the test data set, and filtering target control data corresponding to a preset value from the control data set; the preset value refers to θ in the quantile line of the random variable θ; determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data; Determining, based on the first variance, a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile; determining, based on the second variance, a second target dispersion between a quantile of the reference data set at a preset value quantile and a mean of the quantiles of the reference data set at the preset value quantile; The difference between the test data set and the control data set is determined by using the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

2. The method according to claim 1, characterized in that The step of obtaining a test data set and a control data set corresponding to the target application includes: determining a target application for the updated version and a target application for the original version; Acquire a plurality of first performance data generated when the updated version of the target application is running, and obtain a test data set based on the plurality of first performance data; A plurality of second performance data generated when the original version of the target application is running is obtained, and a comparison data set is obtained based on the plurality of second performance data.

3. The method according to claim 1, characterized in that The step of screening out target test data corresponding to a preset value from the test data set, and screening out target control data corresponding to a preset value from the control data set, comprises: Obtaining a preset value and determining an interval range of the preset value; determining a first cumulative distribution function corresponding to the test data set and a second cumulative distribution function corresponding to the control data; Filtering at least one target test data from the test data set so that the value of the first cumulative distribution function is within the interval range; At least one target control data set whose value of the second cumulative distribution function is within the interval is selected from the control data set.

4. The method according to claim 1, wherein Determining a first variance of a first cumulative distribution function corresponding to the target test data and a second variance of a second cumulative distribution function corresponding to the target control data includes: Resampling the test data set a preset number of times to obtain a first data set, and resampling the control data set a preset number of times to obtain a second data set; For each first data group in the first data group set, determining a first probability value that data in the current first data group is less than or equal to the target test data, and obtaining a sum of first probability values ​​corresponding to the current first data group based on each of the first probability values; Obtaining a first variance of a first cumulative distribution function corresponding to the target test data according to the sum of each of the first probability values; For each second data group in the second data group set, determining a second probability value that data in the current second data group is less than or equal to the target control data, and obtaining a sum of second probability values ​​corresponding to the current second data group based on each second probability value; A second variance of a second cumulative distribution function corresponding to the target control data is obtained according to the sum of each of the second probability values.

5. The method according to claim 4, characterized in that The resampling process is performed on the test data set for a preset number of times to obtain a first data set, including: Performing a logarithmic transformation on each of the test data sets to obtain a corresponding logarithmic test data set, and dividing the logarithmic test data set into a plurality of intermediate data groups; wherein the amount of logarithmic test data included in each of the intermediate data groups is not completely consistent; Combining the plurality of intermediate data groups to obtain a first data group; Update the arrangement order of each logarithmic test data in the logarithmic test data set, enter the next round of resampling processing, return to the step of dividing the logarithmic test data set into multiple intermediate data groups and continue executing until the preset number of resampling times is reached, and obtain the first data set.

6. The method according to claim 5, characterized in that Obtaining a first variance of a first cumulative distribution function corresponding to the target test data according to the sum of each of the first probability values ​​includes: Performing an averaging operation on the sum of the first probability values ​​to obtain a first probability mean; A first difference between the sum of the first probability values ​​and the first probability mean is determined to obtain a first variance of a first cumulative distribution function corresponding to the target test data.

7. The method according to claim 1, characterized in that Determining, based on the first variance, a first target dispersion between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile includes: Determining, based on the first variance, a first intermediate dispersion between a quantile of the test data set at the target test data and a mean of the quantiles of the test data set at the target test data; Based on the first intermediate dispersion degree, a first target dispersion degree between a quantile of the test data set at a preset value quantile and a quantile mean of the test data set at the preset value quantile is determined.

8. The method according to claim 7, characterized in that Determining, based on the first variance, a first intermediate dispersion between a quantile of the test data set at the target test data and a quantile mean of the test data set at the target test data includes: Determining a corresponding first upper quantile and a first lower quantile based on the first variance; wherein the probability value of the quantile corresponding to the test data set being between the first upper quantile and the first lower quantile satisfies a preset confidence condition; determining a second difference between the first upper quantile and the target test data, and determining a third difference between the target test data and the first lower quantile; When the second difference is greater than the third difference, determining a corresponding first intermediate dispersion degree according to the second difference; When the second difference is smaller than the third difference, a corresponding first intermediate dispersion degree is determined according to the third difference.

9. The method according to claim 7, characterized in that Determining, based on the first intermediate dispersion, a first target dispersion between a quantile of the test data set at a preset value quantile and a mean of quantiles of the test data set at the preset value quantile includes: performing linear interpolation processing on a data pair consisting of the first cumulative distribution function and the corresponding first intermediate dispersion degree to obtain a first corresponding relationship between the first cumulative distribution function and the first intermediate dispersion degree; A first target dispersion degree between a quantile of the test data set at the preset value quantile and a quantile mean of the test data set at the preset value quantile is determined according to the first corresponding relationship and the preset value.

10. The method according to claim 1, characterized in that Determining, based on the second variance, a second target dispersion between a quantile of the control data set at a preset value quantile and a quantile mean of the control data set at the preset value quantile includes: determining, based on the second variance, a second intermediate dispersion between a quantile of the control data set at the target control data and a mean of the quantiles of the control data set at the target control data; Based on the second intermediate dispersion degree, a second target dispersion degree between a quantile of the reference data set at a preset value quantile and a mean of the quantiles of the reference data set at the preset value quantile is determined.

11. The method according to claim 1, wherein The determining the difference between the test data set and the control data set by using the first target dispersion degree and the second target dispersion degree includes: Determining a preset value quantile line of the test data set and a preset value quantile line of the control data set; Determining a true probability value of a difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test dataset, and the preset value quantile of the control dataset; When the true probability value is greater than or equal to a preset threshold, the difference between the test dataset and the control dataset is determined according to a preset value quantile of the test dataset and a preset value quantile of the control dataset.

12. The method according to claim 11, characterized in that The determining, based on the first target dispersion degree, the second target dispersion degree, the preset value quantile of the test data set, and the preset value quantile of the control data set, a true probability value of a difference between the preset value quantile of the test data set and the preset value quantile of the control data set includes: determining a fourth difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset based on a difference between the preset value quantile of the test dataset and the preset value quantile of the control dataset; determining a standard deviation of the fourth difference based on the first target dispersion level, the second target dispersion level, a preset value quantile of the test data set, and a preset value quantile of the control data set; A true probability value of a difference between the preset value quantile of the test data set and the preset value quantile of the control data set is determined based on the fourth difference and the standard deviation of the fourth difference.

13. The method according to any one of claims 1 to 11, characterized in that The test data set and the control data set are performance data collected for different versions of target applications; the method further includes: When it is determined that the performance of the original version of the target application is better than the performance of the updated version of the target application based on the difference between the test data set and the control data set, restoring the updated version of the target application to the original version of the target application; When it is determined that the performance of the updated version of the target application is better than the performance of the original version of the target application based on the difference between the test data set and the control data set, the original version of the target application is updated.

14. A data processing device based on an application, characterized in that: The device comprises: a variance determination module configured to obtain a test data set and a control data set corresponding to a target application; filter target test data corresponding to a preset value from the test data set, and filter target control data corresponding to a preset value from the control data set; determine a first variance of a first cumulative distribution function corresponding to the target test data, and a second variance of a second cumulative distribution function corresponding to the target control data; the preset value being θ in a quantile of a random variable θ; a dispersion determination module, configured to determine, based on the first variance, a first target dispersion degree between a quantile of the test data set at a preset value quantile and a mean of the quantiles of the test data set at the preset value quantile; and to determine, based on the second variance, a second target dispersion degree between a quantile of the control data set at a preset value quantile and the mean of the quantiles of the control data set at the preset value quantile; A difference determination module is used to determine the difference between the test data set and the control data set based on the first target dispersion degree and the second target dispersion degree; wherein the difference is used to optimize the target application.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 13 are implemented.

16. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

Images