Data processing method, recommendation parameter determination method and related products

Abstract

The invention discloses a data processing method, a recommendation parameter determination method and a related product. In the method, a reward value represents improvement of a target effect of a target service, a parameter used for executing the target service is a target parameter, the method is used for determining the reward value under the condition that a target value of the target parameter is determined by enabling the reward value to meet a preset condition, and the method comprises the steps of obtaining an experimental index set of an experimental object set; obtaining an experiment effect quantity set and an initial weight set of the experiment effect quantity set; under the condition that the target experiment index in the experiment index set is smaller than the expected index, the initial weight corresponding to the target experiment index in the initial weight set is increased, and an optimized weight set is obtained; and based on the optimized weight set, performing weighted summation on the experimental effect variables in the experimental effect variable set to obtain a reward value.

Description

Technical Field

[0001] The present application relates to the technical field of big data analysis, and in particular to a data processing method, a method for determining recommended parameters, and related products. Background Art

[0002] In real-life scenarios, we often encounter situations where we need to determine parameters for the target business. Specifically, the target business can be executed through target parameters, and the target parameters can control the execution effect of the target business. In order to achieve the expected execution effect of the target business, it is necessary to determine the target value for the target parameter. Since the execution effect of the target business can be reflected by the business indicators of the target business, the execution effect of the target business can be improved by adjusting the target parameters to improve the business indicators of the target business. However, when adjusting the target parameters, the indicator exchange phenomenon is prone to occur. The indicator exchange phenomenon refers to the decrease of another part of the business indicators of the target business while improving the business indicators of the target business. The occurrence of the indicator exchange phenomenon will affect the efficiency of determining the target value of the target parameter. Therefore, it is of great significance to reduce the probability of the indicator exchange phenomenon in the process of determining the target value of the target parameter. Summary of the invention

[0003] The present application provides a data processing method, a method for determining recommended parameters and related products to reduce the probability of indicator exchange phenomenon, wherein the related products include a data processing device, a device for determining recommended parameters, an electronic device, a computer-readable storage medium and a computer program product.

[0004] In a first aspect, a data processing method is provided, in which a reward value represents an improvement in a target effect of a target business, a parameter for executing the target business is a target parameter, and the method is used to determine the reward value when the target value of the target parameter is determined by making the reward value satisfy a preset condition, and the method includes:

[0005] Acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated by executing the target business on the experimental objects in the experimental object set;

[0006] Acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the experimental effect sizes represent the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the experimental effect sizes;

[0007] When the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to obtain an optimized weight set, the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0008] Based on the optimization weight set, the experimental effect quantities in the experimental effect quantity set are weightedly summed to obtain the reward value.

[0009] In combination with any embodiment of the present application, when the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased, and before obtaining the optimized weight set, the method further includes:

[0010] Acquire a control indicator set of a control object set, wherein the control indicator set is a set of control indicators, wherein the control indicators are business indicators generated by executing the target business on control objects in the control object set, wherein the control objects are objects matched with the experimental objects, and the control indicators in the control indicator set are obtained when the target parameters are initial values;

[0011] For the experimental indicator in the experimental indicator set, the expected indicator of the experimental indicator is determined based on the control indicator in the control indicator set corresponding to the experimental indicator.

[0012] In combination with any implementation manner of the present application, the step of obtaining a set of experimental effect sizes includes:

[0013] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0014] Dividing the control indicator set into n control indicator groups according to the preset method, the control indicators in the same control indicator group have a matching relationship, and the control indicators in different control indicator groups do not have a matching relationship, and the control indicator groups correspond to the experimental indicator groups one by one;

[0015] For the experimental indicator group and the control indicator group having a corresponding relationship, the experimental effect size set is determined based on the difference between the experimental indicator group and the control indicator group.

[0016] In combination with any embodiment of the present application, the step of determining the experimental effect size set based on the difference between the experimental indicator group and the control indicator group includes:

[0017] Obtaining n expected coefficients of the n control indicator groups, wherein the expected coefficients correspond one to one to the control indicator groups;

[0018] For the control indicator group and the expected coefficient having a corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups;

[0019] For the experimental indicator group and the expected indicator group having a corresponding relationship, determine a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, wherein the upper limit value of the indicator is the product of the mean value of the indicator in the expected indicator group and an upper limit coefficient, and the upper limit coefficient is greater than 1;

[0020] When the first difference is a non-positive number, a first difference between the experimental indicator group and the expected indicator group is determined as the experimental effect size.

[0021] In combination with any embodiment of the present application, after determining a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, the method further includes:

[0022] When the first difference is a positive number, determining a product of the first difference and a first penalty coefficient to obtain a second difference, wherein the first penalty coefficient is a positive number less than or equal to 1;

[0023] obtaining the experimental effect size based on the difference between the first difference and the second difference;

[0024] The experimental effect size set is obtained based on the experimental effect size.

[0025] In combination with any embodiment of the present application, obtaining the experimental effect size based on the difference between the first difference and the second difference includes:

[0026] Obtaining the effect size to be confirmed based on the difference between the first difference and the second difference;

[0027] When the effect size to be confirmed is a negative number, determining the product of the effect size to be confirmed and a second penalty coefficient to obtain the experimental effect size, and the second penalty coefficient is greater than 1;

[0028] When the effect size to be confirmed is greater than the first preset value, the product of the effect size to be confirmed and the third penalty coefficient is determined to obtain the experimental effect size, and the second penalty coefficient is a positive number less than 1.

[0029] In combination with any implementation manner of the present application, obtaining the effect size to be confirmed based on the difference between the first difference and the second difference includes:

[0030] Determine the difference between the first difference and the second difference to obtain a second difference;

[0031] Determine a discrete value based on the indicators in the experimental indicator group, wherein the discrete value represents the degree of discreteness of the indicators in the experimental indicator group;

[0032] The effect size to be confirmed is obtained based on the ratio of the second difference to the discrete value.

[0033] In combination with any implementation manner of the present application, the determining of discrete values ​​based on the indicators in the experimental indicator group includes:

[0034] Determining a first standard deviation of an indicator within the experimental indicator group;

[0035] determining a second standard deviation of the index within the control index group;

[0036] The first standard deviation and the second standard deviation are weighted averaged to obtain the discrete value.

[0037] In combination with any implementation manner of the present application, obtaining n expected coefficients includes:

[0038] Dividing the n experimental indicator groups into an experimental offensive group and an experimental defensive group, wherein the indicators in the experimental offensive group match the target effect, and the experimental defensive group is an indicator group other than the experimental offensive group among the n experimental indicator groups;

[0039] Determining the expected coefficient of the experimental offensive group to be a second preset value;

[0040] Determine that the expected coefficient of the experimental defense group is a third preset value, wherein the third preset value is a positive number and is less than the second preset value;

[0041] The n expected coefficients are obtained based on the second preset value and the third preset value.

[0042] In combination with any implementation manner of the present application, the determining the expected indicator of the experimental indicator based on the control indicator in the control indicator set corresponding to the experimental indicator includes:

[0043] For the experimental index and the control index having a corresponding relationship, determining that there is a corresponding relationship between the experimental index and an index in the expected index group corresponding to the control index;

[0044] An indicator in the expected indicator group corresponding to the experimental indicator is determined as the expected indicator of the experimental indicator.

[0045] In combination with any implementation manner of the present application, the step of increasing the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set includes:

[0046] Determine the absolute value of the difference between the target experimental index and the expected index;

[0047] Determining an increase coefficient of a target weight corresponding to the target experimental indicator based on the absolute value, wherein the increase coefficient is greater than 1 and is positively correlated with the absolute value;

[0048] Determine the product of the increase coefficient and the target weight to obtain an optimization weight corresponding to the target experimental indicator;

[0049] The optimized weight set is obtained based on the optimized weight and the weights in the initial weight set except the target weight.

[0050] In combination with any implementation manner of the present application, before determining the increase coefficient of the target weight corresponding to the target experimental indicator based on the absolute value, the method includes:

[0051] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0052] Determine a target experimental indicator group including the target experimental indicator from the n experimental indicator groups;

[0053] The weight corresponding to the experimental indicator in the target experimental indicator group in the initial weight set is determined as the target weight.

[0054] In combination with any implementation manner of the present application, the preset condition is to maximize the reward value. After obtaining the reward value, the method further includes:

[0055] The reward value is maximized by Bayesian optimization, and the value of the target parameter is determined as the target value.

[0056] In combination with any implementation manner of the present application, the step of performing weighted summation on the effect sizes in the experimental effect size set based on the optimization weight set to obtain the reward value includes:

[0057] Normalizing the weights in the optimized weight set to obtain a normalized weight set;

[0058] Based on the normalized weight set, the effect sizes in the experimental effect size set are weighted summed to obtain the reward value.

[0059] In a second aspect, a method for determining a recommended parameter is provided, the method comprising:

[0060] Acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated after recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority;

[0061] Acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the effect sizes;

[0062] In the case where the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to obtain an optimized weight set, the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0063] Based on the optimization weight set, weighted summing the experimental effect quantities in the experimental effect quantity set is performed to obtain a reward value, wherein the reward value represents an improvement in a target effect of a service of recommending multimedia content, wherein the service of recommending multimedia content is a service of recommending multimedia content according to a recommendation priority of the multimedia content, wherein the recommendation priority is determined based on a recommendation parameter;

[0064] By making the reward value satisfy a preset condition, a target value of the recommended parameter is determined.

[0065] In a third aspect, a data processing device is provided. In a method executed by the data processing device, a reward value represents an improvement in a target effect of a target business, a parameter used to execute the target business is a target parameter, and the data processing device is used to determine the reward value when the target value of the target parameter is determined by making the reward value satisfy a preset condition. The data processing device includes:

[0066] An acquisition unit, configured to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated by executing the target business on the experimental objects in the experimental object set;

[0067] The acquisition unit is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, and the experimental effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the experimental effect sizes;

[0068] an increasing unit, for increasing the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0069] A processing unit is used to perform weighted summation on the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain the reward value.

[0070] In conjunction with any embodiment of the present application, the acquisition unit is further used to acquire a control indicator set of a control object set, the control indicator set is a set of control indicators, the control indicators are business indicators generated by executing the target business on the control objects in the control object set, the control objects are objects matched with the experimental objects, and the control indicators in the control indicator set are obtained when the target parameters are initial values;

[0071] The processing unit is further configured to determine, for the experimental indicator in the experimental indicator set, the expected indicator of the experimental indicator based on the control indicator in the control indicator set corresponding to the experimental indicator.

[0072] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0073] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0074] Dividing the control indicator set into n control indicator groups according to the preset method, the control indicators in the same control indicator group have a matching relationship, and the control indicators in different control indicator groups do not have a matching relationship, and the control indicator groups correspond to the experimental indicator groups one by one;

[0075] For the experimental indicator group and the control indicator group having a corresponding relationship, the experimental effect size set is determined based on the difference between the experimental indicator group and the control indicator group.

[0076] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0077] Obtaining n expected coefficients of the n control indicator groups, wherein the expected coefficients correspond one to one to the control indicator groups;

[0078] For the control indicator group and the expected coefficient having a corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups;

[0079] For the experimental indicator group and the expected indicator group having a corresponding relationship, determine a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, wherein the upper limit value of the indicator is the product of the mean value of the indicator in the expected indicator group and an upper limit coefficient, and the upper limit coefficient is greater than 1;

[0080] When the first difference is a non-positive number, a first difference between the experimental indicator group and the expected indicator group is determined as the experimental effect size.

[0081] In combination with any implementation manner of the present application, the acquisition unit is further used to:

[0082] When the first difference is a positive number, determining a product of the first difference and a first penalty coefficient to obtain a second difference, wherein the first penalty coefficient is a positive number less than or equal to 1;

[0083] obtaining the experimental effect size based on the difference between the first difference and the second difference;

[0084] The experimental effect size set is obtained based on the experimental effect size.

[0085] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0086] Obtaining the effect size to be confirmed based on the difference between the first difference and the second difference;

[0087] When the effect size to be confirmed is a negative number, determining the product of the effect size to be confirmed and a second penalty coefficient to obtain the experimental effect size, and the second penalty coefficient is greater than 1;

[0088] When the effect size to be confirmed is greater than the first preset value, the product of the effect size to be confirmed and the third penalty coefficient is determined to obtain the experimental effect size, and the second penalty coefficient is a positive number less than 1.

[0089] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0090] determining a difference between the first difference and the second difference to obtain a second difference;

[0091] Determine a discrete value based on the indicators in the experimental indicator group, wherein the discrete value represents the degree of discreteness of the indicators in the experimental indicator group;

[0092] The effect size to be confirmed is obtained based on the ratio of the second difference to the discrete value.

[0093] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0094] Determining a first standard deviation of an indicator within the experimental indicator group;

[0095] Determining a second standard deviation of the index within the control index group;

[0096] The first standard deviation and the second standard deviation are weighted averaged to obtain the discrete value.

[0097] In combination with any implementation manner of the present application, the acquisition unit is specifically used to:

[0098] Divide the n experimental indicator groups into an experimental offensive group and an experimental defensive group, wherein the indicators in the experimental offensive group match the target effect, and the experimental defensive group is an indicator group among the n experimental indicator groups except the experimental offensive group;

[0099] Determining the expected coefficient of the experimental offensive group to be a second preset value;

[0100] Determine that the expected coefficient of the experimental defense group is a third preset value, wherein the third preset value is a positive number and is less than the second preset value;

[0101] The n expected coefficients are obtained based on the second preset value and the third preset value.

[0102] In conjunction with any implementation manner of the present application, the processing unit is specifically used to:

[0103] For the experimental index and the control index having a corresponding relationship, determining that there is a corresponding relationship between the experimental index and an index in the expected index group corresponding to the control index;

[0104] An indicator in the expected indicator group corresponding to the experimental indicator is determined as the expected indicator of the experimental indicator.

[0105] In conjunction with any implementation manner of the present application, the processing unit is specifically used to:

[0106] Determine the absolute value of the difference between the target experimental index and the expected index;

[0107] Determining an increase coefficient of a target weight corresponding to the target experimental indicator based on the absolute value, wherein the increase coefficient is greater than 1 and is positively correlated with the absolute value;

[0108] Determine the product of the increase coefficient and the target weight to obtain an optimization weight corresponding to the target experimental indicator;

[0109] The optimized weight set is obtained based on the optimized weight and the weights in the initial weight set except the target weight.

[0110] In combination with any embodiment of the present application, the processing unit is further used for:

[0111] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0112] Determine a target experimental indicator group including the target experimental indicator from the n experimental indicator groups;

[0113] The weight corresponding to the experimental indicator in the target experimental indicator group in the initial weight set is determined as the target weight.

[0114] In combination with any implementation manner of the present application, the preset condition is to maximize the reward value, and the processing unit is further used to:

[0115] The reward value is maximized by Bayesian optimization, and the value of the target parameter is determined as the target value.

[0116] In conjunction with any implementation manner of the present application, the processing unit is specifically used to:

[0117] Normalizing the weights in the optimized weight set to obtain a normalized weight set;

[0118] Based on the normalized weight set, the effect sizes in the experimental effect size set are weighted summed to obtain the reward value.

[0119] In a fourth aspect, a device for determining a recommended parameter is provided, the device for determining a recommended parameter comprising:

[0120] An acquisition unit, configured to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated after recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority;

[0121] The acquisition unit is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the effect sizes;

[0122] an increasing unit, configured to increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0123] a processing unit, configured to perform weighted summation of the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain a reward value, wherein the reward value represents an improvement in a target effect of a service of recommending multimedia content, wherein the service of recommending multimedia content is a service of recommending multimedia content according to a recommendation priority of the multimedia content, wherein the recommendation priority is determined based on a recommendation parameter;

[0124] A determination unit is used to determine a target value of the recommendation parameter by making the reward value meet a preset condition.

[0125] In a fifth aspect, an electronic device is provided, comprising: a processor and a memory, the memory being used to store computer program code, the computer program code comprising computer instructions, and when the processor executes the computer instructions, the electronic device executes the first aspect and any implementation manner thereof, or the electronic device executes the second aspect.

[0126] In a sixth aspect, another electronic device is provided, comprising: a processor, a sending device, an input device, an output device and a memory, wherein the memory is used to store computer program code, and the computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes the first aspect and any implementation manner thereof, or the electronic device executes the second aspect.

[0127] In the seventh aspect, a computer-readable storage medium is provided, in which a computer program is stored. The computer program includes program instructions. When the program instructions are executed by a processor, the processor is caused to execute the first aspect and any implementation manner thereof, or the processor is caused to execute the second aspect.

[0128] In an eighth aspect, a computer program product is provided, which includes a computer program or instructions, and when the computer program or instructions are run on a computer, the computer is caused to execute the above-mentioned first aspect and any embodiment thereof, or the computer is caused to execute the above-mentioned second aspect.

[0129] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application.

[0130] In the present application, the reward value represents the improvement of the target effect of the target business, and the parameter used to execute the target business is the target parameter. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a business indicator generated by executing the target business on the experimental objects in the experimental object set. The experimental effect quantity set is a set of experimental effect quantities of the experimental indicators, wherein the experimental effect quantity represents the improvement of the target effect by the experimental indicators, and the initial weight set is a set of initial weights of the experimental effect quantities. After the data processing device obtains the experimental indicator set, the experimental effect quantity set and the initial weight set of the experimental effect quantity set of the experimental object set, when the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to increase the negative effect of the experimental effect quantity of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the process of determining the target value of the target parameter by making the reward value meet the preset conditions. That is, in the iterative optimization process of determining the target value of the target parameter by making the reward value meet the preset conditions, if the target experimental indicator is smaller than the expected indicator, it means that the target experimental indicator is lower than the expected indicator. The data processing device then increases the initial weight corresponding to the target experimental indicator in the initial weight set to improve the target experimental indicator, thereby reducing the probability of indicator exchange, thereby improving the efficiency of achieving the desired improvement in the target effect by adjusting the target parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0131] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the background technology, the drawings required for use in the embodiments of the present application or the background technology will be described below.

[0132] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0133] Figure 1 A flowchart of a data processing method provided in an embodiment of the present application;

[0134] Figure 2A schematic diagram of a process for obtaining a reward value provided in an embodiment of the present application;

[0135] Figure 3 A schematic diagram of a process for determining a recommended parameter provided in an embodiment of the present application;

[0136] Figure 4 A schematic diagram of the structure of a data processing device provided in an embodiment of the present application;

[0137] Figure 5 A schematic diagram of the structure of a device for determining recommended parameters provided in an embodiment of the present application;

[0138] Figure 6 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0139] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0140] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0141] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0142] The execution subject of the data processing method in the embodiment of the present application is a data processing device, wherein the data processing device can be any electronic device that can execute the technical solution disclosed in the embodiment of the method of the present application. Optionally, the data processing device can be one of the following: a computer, a server. The reward value represents the improvement of the target effect of the target business, and the parameter used to execute the target business is the target parameter. The method executed by the data processing device is used to determine the reward value when the target value of the target parameter is determined by making the reward value meet the preset conditions.

[0143] In a possible implementation scenario, the target service is to recommend multimedia content, and the target parameter for executing the target service is a parameter for determining the recommendation priority of the multimedia content, wherein the multimedia content includes one or more of the following: images, audio, text, and video. For example, the multimedia content is content including text, and for another example, the multimedia content may be content including text and images, and for another example, the multimedia content is content including audio, text, and video. Specifically, the recommendation priority of the multimedia content can be determined by weighted summing up the various attributes of the multimedia content based on the fusion formula, wherein the attributes include: the popularity of the multimedia content, the proportion of advertising content in the multimedia content, and the number of clicks on the multimedia content. By weighted summing up the various attributes of the multimedia content through the fusion formula, a score for characterizing the recommendation priority of the multimedia content can be obtained, and the higher the score, the higher the recommendation priority of the multimedia content. After determining the recommendation priority of each multimedia content, each multimedia content can be recommended based on the recommendation priority of each multimedia content. Since the recommendation priority will affect the recommendation priority of multimedia content, and thus affect the recommendation of multimedia content, and by adjusting the weights of each attribute in the fusion formula, the recommendation priority of multimedia content can be changed, the weights of each attribute in the fusion formula are parameters for recommending multimedia content, that is, the weights of each attribute in the fusion formula are target parameters for executing the target task. The target effect is the effect produced by executing the target business. For example, in the case where the target business is to recommend multimedia content, the target effect can be the number of clicks on the multimedia content, and the target effect can also be the browsing time of the multimedia content. The reward value is used to characterize the improvement of the target effect. For example, in the case where the target effect is the number of clicks on the multimedia content, the reward value is the improvement of the number of clicks on the multimedia content. In the case where the target effect is the browsing time of the multimedia content, the reward value is the improvement of the browsing time of the multimedia content. Since the target parameter affects the execution effect of the target business, the target parameter affects the target effect. Based on this, by making the reward value meet the preset conditions, the specific value of the target parameter can be determined, that is, the target value. Because the reward value represents the improvement of the target effect of the target business, the reward value meeting the preset conditions indicates that the improvement of the target effect of the target business has reached the expectation. For example, the target effect is the improvement of the number of clicks on the multimedia content, and the preset condition is to maximize the reward value. Then the reward value meeting the preset conditions indicates that the increase in the number of clicks on the multimedia content is the largest. At this time, the expectation is to maximize the increase in the number of clicks on the multimedia content. Among them, the preset conditions can be any conditions. For example, the preset condition is to maximize the reward value. For another example, the preset condition is that the reward value is not less than a certain value. Optionally, in the case where there is a functional relationship between the reward value and the target parameter, the target value can be obtained by solving the functional relationship by taking the reward value satisfying the preset conditions as a constraint condition.Optionally, when there is no functional relationship between the reward value and the target parameter, the target value can be determined based on the Bayesian optimization method by taking the reward value satisfying the preset condition as a constraint condition. Since the target value is determined by making the reward value satisfy the preset condition, when the value of the target parameter is the target value, executing the target business based on the target parameter can achieve the expected improvement in the target effect. Based on this, the data processing device determines the reward value by executing the data processing method, and can further determine the target value of the target parameter based on the reward value.

[0144] It should be understood that the method embodiment of the present application can also be implemented by a processor executing a computer program code. The following describes the embodiment of the present application in conjunction with the drawings in the embodiment of the present application. Figure 1 , Figure 1 A flowchart of a data processing method provided in an embodiment of the present application.

[0145] 101. Obtain a set of experimental indicators for a set of experimental objects.

[0146] In the embodiment of the present application, the experimental object set is a set of experimental objects, wherein the number of experimental objects in the experimental object set is greater than 1. The experimental object is the execution object of the target business. For example, if the target business is to recommend multimedia content, then the experimental object is the recommended object of the multimedia content. For another example, if the target business is to place advertisements, then the experimental object is the object of advertisement placement. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a business indicator generated by executing the target business on the experimental objects in the experimental object set. For example, if the target business is to recommend multimedia content, then by recommending multimedia content to the object, the business indicators that can be generated include: the number of clicks on multimedia content, the browsing time of multimedia content, the exposure of multimedia content, the hiding rate of multimedia content, and the traffic of new multimedia content, wherein the hiding rate refers to the proportion of hidden or deleted multimedia content, the traffic of new multimedia content refers to the traffic of new multimedia content, and the new multimedia content is the latest released multimedia content. The traffic includes the number of clicks on multimedia content and the number of interactions with multimedia content. The number of interactions is the number of interactive behaviors generated for multimedia content, and the interactive behaviors include: forwarding, liking, and commenting.

[0147] In an implementation method of obtaining an experimental indicator set, a data processing device receives an experimental indicator set input by a user through an input component, wherein the input component includes: a keyboard, a mouse, a touch screen, a touch pad, and an audio input device.

[0148] In another implementation of obtaining the experimental indicator set, the data processing device receives the experimental indicator set sent by the terminal. Optionally, the terminal can be any of the following: a mobile phone, a computer, a tablet computer, and a server.

[0149] 102. Obtain an experimental effect size set and an initial weight set of the experimental effect size set.

[0150] In the embodiment of the present application, the set of experimental effect sizes is a set of experimental effect sizes of experimental indicators, and the experimental effect sizes of experimental indicators represent the improvement of the target effect by the experimental indicators. For example, the target business is to recommend multimedia content, and the target effect is to increase the number of interactions. The experimental effect size of the experimental indicator represents the increase in the number of interactions brought about by the experimental indicator. For example, the experimental indicator is the browsing time of multimedia content. Since the longer the browsing time of multimedia content, the higher the probability of generating interaction with the multimedia content, the browsing time of multimedia content contributes to the increase in the number of interactions of multimedia content. The experimental effect size of the browsing time of multimedia content represents the increase in the number of interactions brought about by the browsing time of multimedia content.

[0151] In the embodiment of the present application, the initial weight set is a set of initial weights of the experimental effect size. Since each experimental indicator contributes to improving the target effect, and the experimental effect size of the experimental indicator represents the improvement of the target effect by the experimental indicator, the improvement of the target effect can be determined by determining the sum of the improvement of all experimental indicators to the target effect, and the reward value can also be determined. Because different experimental indicators contribute differently to improving the target effect, the weights of the experimental effect sizes of different experimental indicators are different, so each experimental indicator has an initial weight, and the initial weight set is the set of initial weights. Based on the initial weights in the initial weight set, the experimental effect sizes of all experimental indicators are weighted and summed to obtain a reward value.

[0152] 103. When the target experimental indicator in the experimental indicator set is less than the expected indicator, increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set.

[0153] In the embodiment of the present application, the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation. For example, the target business is to recommend multimedia content, and the business indicators generated by the target business include the browsing time of the multimedia content and the number of clicks on the multimedia content. The target effect is to increase the number of interactions of the multimedia content, and the expectation is that the number of interactions of the multimedia content is increased to the expected value. When the number of interactions of the multimedia content is the expected value, the browsing time of the multimedia content is the first value, and the number of clicks on the multimedia content is the second value, then for the business indicator of browsing time, the expected indicator is the first value, and for the business indicator of clicks, the expected indicator is the second value.

[0154] If the target experimental index is less than the expected index, it means that the target experimental index has not achieved the expected improvement in the target effect, which may lead to the target effect not achieving the expected improvement. Because the experimental effect size of the target experimental index is positively correlated with the target difference, where the target difference is the difference between the target experimental index and the expected index, when the target experimental index is less than the expected index, the experimental effect size of the target experimental index is negative. At this time, if the reward value is obtained by weighted summing the experimental effect sizes of all experimental indicators, then the contribution of the experimental effect size of the target experimental index to the reward value is negative, that is, the experimental effect size of the target experimental index has a negative effect on the improvement of the target effect, which leads to the reward value not meeting the preset conditions. In this way, if you want the reward value to meet the preset conditions, you need to increase the target experimental index so that the target experimental index is not less than the expected index. Therefore, when the target experimental indicator in the experimental indicator set is smaller than the expected indicator, the data processing device increases the initial weight corresponding to the target experimental indicator in the initial weight set to increase the negative effect of the experimental effect amount of the target experimental indicator on the improvement of the target effect, and then improves the target experimental indicator by making the reward value meet the preset conditions, so that the target experimental indicator is not less than the expected indicator.

[0155] By adjusting the parameters of the target business, the execution effect of the target business can be adjusted, which can lead to changes in the business indicators generated by executing the target business. However, when there is no clear functional relationship between the adjustment of the parameters of the target business and the change of the business indicators, while adjusting the parameters of the target business to improve some business indicators, other business indicators may decrease. For example, the target business is to recommend multimedia content, and the target parameter used to execute the target business is the weight of the fusion formula used to determine the recommendation priority of the multimedia content, where the weight is the weight of each attribute of the multimedia content in the fusion formula. If the object likes multimedia content with high popularity but does not like multimedia content with many advertisements, and the attributes of the multimedia content include: the popularity of the multimedia content, and the proportion of advertising content in the multimedia content. By increasing the weight of the attribute of popularity and reducing the weight of the attribute of the proportion of advertising content in the multimedia content, the recommendation priority of the multimedia content with high popularity can be higher, and then the multimedia content with high popularity can be recommended to the object with a greater probability, so that when the object likes the multimedia content with high popularity, the probability of the multimedia content being clicked by the object can be increased, thereby increasing the number of clicks on the multimedia content. However, reducing the weight of the proportion of advertising content in multimedia content may increase the recommendation priority of multimedia content with a large proportion of advertising content, which in turn may lead to a greater probability of multimedia content with a large proportion of advertising content being recommended to the object. Therefore, if the object does not like multimedia content with many advertisements, it will lead to a decrease in the object's browsing time of the multimedia content after clicking to view the multimedia content. At this time, by increasing the weight of the attribute of popularity and reducing the weight of the attribute of the proportion of advertising content in multimedia content, the business indicator of the number of clicks on multimedia content can be increased, but it will also lead to a decrease in the business indicator of the browsing time of multimedia content. For the sake of convenience, the phenomenon of improving some business indicators of the target business while other business indicators decrease is referred to as the indicator exchange phenomenon below.

[0156] Although the business indicators of the target business all contribute to the improvement of the target effect, the business indicators can be divided into indicators that match the target effect and indicators that do not match the target effect. For example, if the target business is to recommend multimedia content, and the improvement of the target effect is to increase the browsing time of multimedia content, then the business indicators that match the duration are indicators that match the target effect, such as the average browsing time of multimedia content, the total browsing time of multimedia content, and the browsing time of new multimedia content. Business indicators that do not match the duration are indicators that do not match the target effect, such as the exposure of multimedia content, the proportion of advertising content in multimedia content, and the popularity of multimedia content. In order to improve the target effect, it is usually chosen to improve the indicators that match the target effect by adjusting the target parameters of the target business. However, due to the existence of indicator exchange, while the indicators that match the target effect are improved, the indicators that do not match the target effect may decrease. Indicators that do not match the target effect also contribute to the improvement of the target effect, thereby reducing the efficiency of achieving the desired improvement in the target effect by adjusting the target parameters.

[0157] In the embodiments of the present application, the target value of the target parameter determined by making the reward value satisfy the preset conditions as described above refers to making the reward value satisfy the preset conditions, thereby making the target effect reach the expected improvement, that is, determining the target value of the target parameter by making the target effect reach the expected improvement. In the process of determining the target value of the target parameter by making the reward value satisfy the preset conditions, it is necessary to adjust the target parameter multiple times, and determine a new reward value based on the adjusted target parameter until the reward value satisfies the preset conditions and determines the target value of the target parameter. That is, the process of determining the target value of the target parameter by making the reward value satisfy the preset conditions is an iterative optimization process.

[0158] In the embodiment of the present application, when the target experimental indicator in the experimental indicator set is less than the expected indicator, the data processing device increases the initial weight corresponding to the target experimental indicator in the initial weight set to increase the negative effect of the experimental effect amount of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the iterative optimization process. That is, in the iterative optimization process of determining the target value of the target parameter by making the reward value meet the preset conditions, if the target experimental indicator is less than the expected indicator, it means that the target experimental indicator is lower than the expected indicator, and the data processing device then improves the target experimental indicator in real time by executing step 103, thereby reducing the probability of the indicator exchange phenomenon, thereby improving the efficiency of achieving the desired improvement in the target effect by adjusting the target parameters. It should be understood that the above-mentioned target experimental indicator can be any indicator that is less than the expected indicator among all experimental indicators.

[0159] In one possible implementation, the data processing device increases the initial weight corresponding to the target experimental index in the initial weight set by performing the following to obtain an optimized weight set: determine the absolute value of the difference between the target experimental index and the expected index. Determine the increase coefficient of the target weight corresponding to the target experimental index based on the absolute value, wherein the increase coefficient is greater than 1 and the increase coefficient is positively correlated with the absolute value. Optionally, the absolute value is used as the increase coefficient. Optionally, the increase coefficient is determined by the following formula: Increase coefficient = (1 + K p *e(t))…Formula (1), where K p is a number greater than 1, e(t) is an absolute value, optional, K p is 3. The product of the increase coefficient and the target weight is determined to obtain the optimization weight corresponding to the target experimental index. Based on the optimization weight and the weights in the initial weight set except the target weight, the optimization weight set is obtained.

[0160] The above-mentioned target weight is the weight corresponding to the target experimental index in the initial weight concentration, that is, the weight of the experimental effect amount corresponding to the target experimental index is the target weight. After determining the increase coefficient of the target weight, the data processing device obtains the optimized weight by determining the product of the increase coefficient and the target weight, thereby increasing the initial weight, wherein the increase coefficient is positively correlated with the absolute value. Because the absolute value is the absolute value of the difference between the target experimental index and the expected index, the larger the absolute value, the farther the target experimental index is from the expected index. Therefore, in the case where the increase coefficient is positively correlated with the absolute value, the increase coefficient is determined based on the absolute value, and the target weight can be dynamically adjusted based on the gap between the target experimental index and the expected index. Optionally, the above-mentioned absolute value is the ratio of the relative difference to the expected index, wherein the relative difference is the difference between the target experimental index and the expected index. At this time, the absolute value can better reflect the relative gap between the target experimental index and the expected index.

[0161] As an optional implementation, before the data processing device determines the increase coefficient of the target experimental effect size corresponding to the target experimental indicator based on the absolute value, it also determines the target weight by executing the following steps: the experimental indicator set is divided into n experimental indicator groups according to a preset method, and the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship. For example, the experimental indicator set includes the total browsing time of multimedia content, the average browsing time of multimedia content, the browsing time of videos in multimedia content, the total clicks on multimedia content, and the average clicks on multimedia content. Then it can be determined that the total browsing time, the average browsing time, and the browsing time of videos in multimedia content have a matching relationship, and it can be determined that the total clicks and the average clicks have a matching relationship, and then the total browsing time of multimedia content, the average browsing time of multimedia content, and the browsing time of videos in multimedia content can be divided into the same experimental indicator group, and the total clicks on multimedia content and the average clicks on multimedia content can be divided into the same experimental indicator group. A target experimental indicator group including the target experimental indicator is determined from the n experimental indicator groups. Optionally, the target experimental indicator is a representative indicator in the target experimental indicator group, wherein the representative indicator is an indicator that can reflect the overall effect of the target business. For example, the target business is to recommend multimedia content, and the target experimental indicator group includes the total browsing time of multimedia content, the average browsing time of multimedia content, and the browsing time of videos in multimedia content. Since multimedia content includes video, image, and text, the browsing time of videos in multimedia content only reflects the browsing time of videos in multimedia content, but does not reflect the browsing time of images in multimedia content and the browsing time of text in multimedia content. Therefore, the browsing time of videos in multimedia content cannot reflect the overall effect of recommending multimedia content. On the contrary, the total browsing time of multimedia content and the average browsing time of multimedia content can both reflect the overall effect of recommending multimedia content, and then the total browsing time of multimedia content or the average browsing time of multimedia content can be determined as a representative indicator. The weight corresponding to the experimental indicator in the target experimental indicator group in the initial weight set is determined as the target weight, that is, the initial weights of the experimental effect quantities of the experimental indicators in the target experimental indicator group are all target weights.

[0162] 104. Based on the above optimization weight set, weighted sum is performed on the above experimental effect quantities in the above experimental effect quantity set to obtain the above reward value.

[0163] In an embodiment of the present application, the weights in the optimization weight set correspond one-to-one to the experimental effect sizes in the experimental effect size set. The data processing device uses the weights in the optimization weight set as the weights of the corresponding experimental effect sizes in the experimental effect size set, and performs weighted summation of the experimental effect sizes in the experimental effect size set to obtain a reward value.

[0164] In one possible implementation, the data processing device normalizes the weights in the optimization weight set to obtain a normalized weight set, so that the sum of the weights in the normalized weight set is 1. Based on the normalized weight set, the experimental effect sizes in the experimental effect size set are weighted and summed to obtain the reward value.

[0165] In the embodiment of the present application, the reward value represents the improvement of the target effect of the target business, and the parameter used to execute the target business is the target parameter. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a business indicator generated by executing the target business on the experimental objects in the experimental object set. The experimental effect quantity set is a set of experimental effect quantities of the experimental indicators, wherein the experimental effect quantity represents the improvement of the target effect by the experimental indicators, and the initial weight set is a set of initial weights of the experimental effect quantities. After the data processing device obtains the experimental indicator set, the experimental effect quantity set and the initial weight set of the experimental effect quantity set of the experimental object set, when the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to increase the negative effect of the experimental effect quantity of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the process of determining the target value of the target parameter by making the reward value meet the preset conditions. That is, in the iterative optimization process of determining the target value of the target parameter by making the reward value meet the preset conditions, if the target experimental indicator is smaller than the expected indicator, it means that the target experimental indicator is lower than the expected indicator. The data processing device then increases the initial weight corresponding to the target experimental indicator in the initial weight set to improve the target experimental indicator, thereby reducing the probability of indicator exchange, thereby improving the efficiency of achieving the desired improvement in the target effect by adjusting the target parameters.

[0166] As an optional implementation manner, before executing step 103, the data processing device further executes the following steps:

[0167] 2001. Obtain a control indicator set of a control object set.

[0168] In the embodiment of the present application, the control object set is a set of control objects, wherein the number of control objects in the control object set is greater than 1. The control object is an execution object of the target service. For example, if the target service is to recommend multimedia content, then the control object is a recommendation object of the multimedia content. For another example, if the target service is to place an advertisement, then the control object is a placement object of the advertisement.

[0169] The control object is an object for comparative experiment with the experimental object. Specifically, by conducting comparative experiments between the experimental objects in the experimental object set and the control objects in the control object set, the experimental effect size of the experimental indicator generated by the execution of the target business on the experimental objects in the experimental object set is determined. In one possible implementation method, at the initial moment, the experimental object set and the control object set are determined. At the initial moment, the value of the target parameter is the initial value. The experimental indicator set and the control indicator set at the initial moment are obtained, that is, when the value of the target parameter is the initial value, the experimental objects in the experimental object set and the control objects in the control object set respectively perform the target task, and the experimental indicator set and the control indicator set at the initial moment can be generated. After the initial moment, for the control objects in the control object set, the value of the target parameter remains unchanged, and for the experimental objects in the experimental object set, the value of the target parameter is continuously optimized by making the reward value meet the preset conditions, until the reward value meets the preset conditions to obtain the target value of the target parameter.

[0170] Optionally, matching the experimental subject with the control subject means that the characteristics of the experimental subject match the characteristics of the control subject. For example, if the characteristics of the experimental subject and the characteristics of the control subject both include hobbies, then the hobbies of the experimental subject are the same as those of the control subject.

[0171] In the embodiment of the present application, the control indicator set is a set of control indicators, wherein the control indicator is a business indicator generated by executing the target business on the control object in the control object set. For example, the target business is to recommend multimedia content. Then, by recommending multimedia content to the object, the business indicators that can be generated include: the number of clicks on the multimedia content, the browsing time of the multimedia content, the exposure of the multimedia content, the hiding rate of the multimedia content, and the traffic of new multimedia content, wherein the hiding rate refers to the proportion of hidden or deleted multimedia content, the traffic of new multimedia content refers to the traffic of new multimedia content, and the new multimedia content is the latest multimedia content released. The traffic includes the number of clicks on the multimedia content and the number of interactions with the multimedia content. The number of interactions is the number of interactive behaviors generated for the multimedia content, and the interactive behaviors include: forwarding, liking, and commenting. It should be understood that there is a corresponding relationship between the experimental indicator set and the control indicator set. Specifically, the indicators included in the experimental indicator set are the same as those included in the control indicator set, but the values ​​of the indicators in the experimental indicator set may be different from the values ​​of the indicators in the control indicator set.

[0172] In an embodiment of the present application, the control indicators in the control indicator set are obtained when the target parameters used to execute the target business are initial values. Specifically, when the target parameters are initial values, the control indicator set can be obtained by executing the target business on the control objects in the control object set.

[0173] 2002. For the experimental indicator in the experimental indicator set, determine the expected indicator of the experimental indicator based on the control indicator in the control indicator set corresponding to the experimental indicator.

[0174] Since the control indicators in the control indicator set are obtained when the target parameters for executing the target business are initial values, and the target values ​​are obtained by adjusting on the basis of the initial values, the data processing device can determine the expected indicators of the experimental indicators based on the control indicators in the control indicator set corresponding to the experimental indicators. For example, the experimental indicator set includes the following experimental indicators: browsing time of multimedia content, click volume of multimedia content, then based on the browsing time of multimedia content in the control indicator set, the expected indicators of browsing time of multimedia content in the experimental indicator set can be determined, and based on the click volume of multimedia content in the control indicator set, the expected indicators of click volume of multimedia content in the experimental indicator set can be determined.

[0175] In a possible implementation, the control indicator set is divided into n control indicator groups according to a preset method, the control indicators in the same control indicator group have a matching relationship, the control indicators in different control indicator groups do not have a matching relationship, and the control indicator group corresponds to the experimental indicator group one by one. Obtain n expected coefficients of n control indicator groups, wherein the expected coefficient corresponds to the control indicator group one by one, and the expected coefficient represents the expectation of the improvement of the indicators in the control indicator group. For example, the control indicator group includes the average browsing time of multimedia content and the total browsing time of multimedia content, then the expected coefficient of the browsing time of multimedia content represents the expectation of the improvement of the average browsing time of multimedia content and the expectation of the improvement of the total browsing time of multimedia content. For the control indicator group and the expected coefficient with a corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups. For the experimental indicator and the control indicator with a corresponding relationship, determine that there is a corresponding relationship between the experimental indicator and the indicator corresponding to the control indicator in the expected indicator group, and determine that the indicator corresponding to the experimental indicator in the expected indicator group is the expected indicator of the experimental indicator.

[0176] For example, the experimental indicator set includes the following experimental indicators: browsing time A of multimedia content and click volume B of multimedia content, and the control indicator set includes the following control indicators: browsing time C of multimedia content and click volume D of multimedia content, wherein the browsing time A of multimedia content corresponds to the browsing time C of multimedia content, and the click volume B of multimedia content corresponds to the click volume D of multimedia content. The expected indicator group includes the following indicators: browsing time E of multimedia content and click volume F of multimedia content, wherein the browsing time E of multimedia content is obtained by multiplying the browsing time C of multimedia content with the expected coefficient, and the click volume F of multimedia content is obtained by multiplying the click volume D of multimedia content with the expected coefficient, that is, the browsing time C of multimedia content corresponds to the browsing time E of multimedia content, and the click volume D of multimedia content corresponds to the click volume F of multimedia content. At this time, the browsing time A of multimedia content corresponds to the browsing time E of multimedia content, and the click volume B of multimedia content corresponds to the click volume F of multimedia content. The browsing time E of the multimedia content is the expected indicator of the browsing time A of the multimedia content, and the click volume F of the multimedia content is the expected indicator of the click volume B of the multimedia content.

[0177] Optionally, the data processing device obtains n expected coefficients by performing the following steps: n experimental indicator groups are divided into an experimental attack group and an experimental defense group, wherein the indicators in the experimental attack group match the target effect, for example, the target business is to recommend multimedia content, and the target effect is the browsing time of the multimedia content, then the indicators in the experimental attack group are all indicators related to the browsing time, such as the total browsing time of the multimedia content, the average browsing time of the multimedia content, and the browsing time of the video in the multimedia content. The experimental defense group is an indicator group other than the experimental attack group in the n experimental indicator groups. Determine the expected coefficient of the experimental attack group as the second preset value. Determine the expected coefficient of the experimental defense group as the third preset value, wherein the third preset value is a positive number, and the third preset value is less than the second preset value. Based on the second preset value and the third preset value, n expected coefficients are obtained. Since the indicators in the experimental attack group match the target effect, and the indicators in the experimental defense group do not match the target effect, in order to achieve the desired improvement in the target effect, the improvement of the indicators in the experimental attack group should be greater than the improvement of the indicators in the experimental defense group, that is, the expected coefficient of the experimental attack group should be greater than the expected coefficient of the experimental defense group. Therefore, the second preset value is greater than the third preset value.

[0178] Optionally, the expectation for the index in the experimental defense group is that the index does not decrease, so that the index in the experimental defense group does not have a negative effect on the improvement of the target effect. Therefore, the third preset value is 0, or the third preset value is a small positive number, such as 0.001.

[0179] Optionally, when the n experimental indicator groups are divided into the experimental offensive group and the experimental defensive group, the data processing device divides the n control indicator groups into the control offensive group and the control defensive group, wherein the indicators in the control offensive group match the target effect, and the control defensive group is the indicator group other than the control offensive group among the n control indicator groups. In this way, by multiplying the expected coefficient of the experimental offensive group with the indicator in the control offensive group, the expected indicator of the indicator in the experimental offensive group can be obtained, and by multiplying the expected coefficient of the experimental defensive group with the indicator in the control defensive group, the expected indicator of the indicator in the experimental defensive group can be obtained.

[0180] In this embodiment, after obtaining the control indicator set of the control object set, the data processing device can determine the expected indicators of the experimental indicators in the experimental indicator set based on the control indicators in the control indicator set corresponding to the experimental indicators, thereby determining the reward value through comparative experiments, and then determining the target value of the target parameter.

[0181] As an optional implementation, the data processing device obtains the experimental effect size set by executing the following steps:

[0182] 3001. Divide the above experimental indicator set into n experimental indicator groups according to a preset method.

[0183] In the embodiment of the present application, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship. Optionally, after obtaining n experimental indicator groups, the data processing device performs data standardization on the indicators in each experimental indicator group. Specifically, the mean of the indicators in each experimental indicator group and the standard deviation of the indicators in each experimental indicator group are determined, and then the indicators in each experimental indicator group are standardized by the following formula: nx = (x-mean1) / std1... Formula (2), wherein nx is the experimental indicator after data standardization, x is the experimental indicator before data standardization, mean1 is the mean of the indicators in the experimental indicator group, and std1 is the standard deviation of the indicators in the experimental indicator group.

[0184] 3002. Divide the above-mentioned control indicator set into n control indicator groups according to the above-mentioned preset method.

[0185] In the embodiment of the present application, the control indicators in the same control indicator group have a matching relationship, the control indicators in different control indicator groups do not have a matching relationship, and the control indicator group corresponds to the experimental indicator group one by one. Optionally, after obtaining n control indicator groups, the data processing device performs data standardization on the indicators in each control indicator group, specifically, determines the mean of the indicators in each control indicator group and the standard deviation of the indicators in each control indicator group, and then performs data standardization on the indicators in each control indicator group by the following formula: ny = (y-mean2) / std2 ... Formula (3), wherein ny is the control indicator after data standardization, y is the control indicator before data standardization, mean2 is the mean of the indicators in the control indicator group, and std2 is the standard deviation of the indicators in the control indicator group.

[0186] In step 3001 and step 3002, by grouping the experimental indicator set and the control indicator set, indicators with matching relationships can be grouped into the same group. Different indicators in the same group represent the effect of the target business in the same dimension. For example, if the target business is to recommend multimedia content, the average browsing time of multimedia content and the total browsing time of multimedia content both represent the effect of recommending multimedia content in the dimension of browsing time. Therefore, when determining the reward value in the future, if the indicators in the same group are considered as a whole and the indicators in different groups are considered separately, the relationship between the indicators can be used more accurately to determine the reward value, thereby improving the accuracy of the reward value.

[0187] 3003. For the experimental indicator group and the control indicator group having a corresponding relationship, determine the experimental effect size set based on the difference between the experimental indicator group and the control indicator group.

[0188] Since the control indicator group is obtained when the target parameter is the initial value, and the experimental indicator group is obtained by adjusting the target parameter based on the initial value, the difference between the experimental indicator group and the control indicator group can represent the improvement of the business indicator of the target business by adjusting the target parameter. Because the improvement of the business indicator of the target business contributes to the improvement of the target effect, the data processing device can determine the experimental effect size set based on the difference between the experimental indicator group and the control indicator group. Specifically, the effect size of the experimental indicator is determined based on the difference between the experimental indicator in the experimental indicator group and the control indicator in the control indicator group. Optionally, the effect size of the experimental indicator is positively correlated with the difference.

[0189] In a possible implementation, the data processing device determines the experimental effect size set based on the difference between the experimental indicator group and the control indicator group by executing the following steps: obtaining n expected coefficients of n control indicator groups. For the control indicator group and the expected coefficient with corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups. For the experimental indicator group and the expected indicator group with corresponding relationship, determine the first difference between the mean of the indicator in the experimental indicator group and the upper limit of the indicator, the upper limit of the indicator is the product of the mean of the indicator in the expected indicator group and the upper limit coefficient, wherein the upper limit coefficient is greater than 1. In the case where the first difference is a non-positive number, determine the first difference between the experimental indicator group and the expected indicator group as the experimental effect size, and optionally, the first difference is the difference between the mean of the indicator in the experimental indicator group and the indicator in the expected indicator group. In the case where the first difference is a positive number, determine the product of the first difference and the first penalty coefficient to obtain the second difference, wherein the first penalty coefficient is a positive number less than or equal to 1. Based on the difference between the first difference and the second difference, obtain the experimental effect size. The experimental effect size set is obtained based on the experimental effect size.

[0190] The first difference is a non-positive number, which means that the mean of the indicators in the experimental indicator group is not greater than the upper limit of the indicator. Since the upper limit of the indicator is the product of the mean of the indicators in the expected indicator group and the upper limit coefficient, it means that the mean of the indicators in the experimental indicator group is not much larger than the mean of the indicators in the expected indicator group, which means that the indicators in the experimental indicator group do not exceed the expected indicators by much. On the contrary, the first difference is a positive number, which means that the indicators in the experimental indicator group exceed the expected indicators by much. Considering that the improvement of the business indicators of the target business requires resources, where resources include hardware resources and traffic resources, for example, the target business is to recommend multimedia content, and improving the business indicators of the target business requires more equipment to execute the business of recommending multimedia content, and also requires more network traffic resources to recommend multimedia content. If the indicators in the experimental indicator group exceed the expected indicators by a large margin, it means that although the indicators in the experimental indicator group have reached the expected indicators, the improvement of the indicators beyond the expected indicators consumes additional resources, which means that a waste of resources is generated at this time. Therefore, when the first difference is a non-positive number, the data processing device determines the first difference between the experimental index group and the expected index group as the experimental effect size, and then after obtaining the reward value based on the experimental effect size, the experimental index in the experimental index group is close to the expected index by making the reward value meet the preset condition. When the first difference is a positive number, the data processing device determines the product of the first difference and the first penalty coefficient to obtain the second difference, wherein the first penalty coefficient is a positive number less than or equal to 1. Based on the difference between the first difference and the second difference, the experimental effect size is obtained. Since the second difference is a positive number, the experimental effect size is obtained based on the difference between the first difference and the second difference, which can make the experimental effect size smaller than the first difference. Then, after obtaining the reward value based on the experimental effect size, the experimental index in the experimental index group is lower than the upper limit of the index by making the reward value meet the preset condition, thereby reducing the waste of resources. Optionally, the data processing device uses the difference between the first difference and the second difference as the experimental effect size. Finally, after obtaining the experimental effect size, the experimental effect size set can be obtained based on the experimental effect size.

[0191] Optionally, the first difference is obtained by the following formula: mean_diff = (M1-M2)-a*RELU(M1-bM2), wherein mean_diff represents the first difference, M1 represents the mean of the indicators in the experimental indicator group, M2 represents the mean of the indicators in the control indicator group, a represents the first penalty coefficient, and b represents the upper limit coefficient.

[0192] Optionally, the data processing device obtains the experimental effect size based on the difference between the first difference and the second difference by executing the following steps: obtain the effect size to be confirmed based on the difference between the first difference and the second difference, and optionally, the data processing device determines the difference between the first difference and the second difference as the effect size to be confirmed. When the effect size to be confirmed is a negative number, the product of the effect size to be confirmed and the second penalty coefficient is determined to obtain the experimental effect size, wherein the second penalty coefficient is greater than 1. When the effect size to be confirmed is greater than the first preset value, the product of the effect size to be confirmed and the third penalty coefficient is determined to obtain the experimental effect size, and the second penalty coefficient is a positive number less than 1.

[0193] If the effect size to be confirmed is a negative number, it means that the effect size of the experimental indicator in the experimental indicator group has a negative effect on the improvement of the target effect, which in turn causes the reward value to fail to meet the preset conditions. Therefore, when the effect size to be confirmed is a negative number, the data processing device determines the product of the effect size to be confirmed and the second penalty coefficient to obtain the experimental effect size. At this time, the negative effect of the experimental effect size of the experimental indicator in the experimental indicator group on the improvement of the target effect can be increased, and then the experimental indicator in the experimental indicator group can be improved by making the reward value meet the preset conditions.

[0194] If the effect size to be confirmed is greater than the first preset value, it means that the effect size of the experimental indicator in the experimental indicator group exceeds expectations by a large margin. Since the increase in the effect size requires consumption of resources (this principle is the same as the increase in the indicators mentioned above), the effect size of the experimental indicators in the experimental indicator group exceeds expectations by a large margin, indicating that the experimental indicator group has generated a waste of resources. Therefore, when the effect size to be confirmed is greater than the first preset value, the data processing device determines the product of the effect size to be confirmed and the third penalty coefficient to obtain the experimental effect size, wherein the second penalty coefficient is a positive number less than 1, thereby reducing the experimental effect size of the indicators in the experimental indicator group, thereby reducing the waste of resources.

[0195] Optionally, the data processing device obtains the effect size to be confirmed based on the difference between the first difference and the second difference by executing the following steps: determining the difference between the first difference and the second difference to obtain the second difference. Determining a discrete value based on the indicators in the experimental indicator group, the discrete value characterizes the degree of discreteness of the indicators in the experimental indicator group. Obtaining the effect size to be confirmed based on the ratio of the second difference to the discrete value.

[0196] If the degree of dispersion of the indicators in the experimental indicator group is high, some of the indicators in the experimental indicator group may have a greater impact on the mean of the indicators in the experimental indicator group, which may lead to the mean of the indicators in the experimental indicator group reaching the expected indicator, but some of the indicators in the experimental indicator group failing to reach the expected indicator, resulting in a large error in the effect size of the indicators in the experimental indicator group determined based on the indicators in the experimental indicator group. Therefore, in order to reduce the influence of the degree of dispersion of the indicators in the experimental indicator group on the accuracy of the effect size of the indicators in the experimental indicator group, after determining the discrete value based on the indicators in the experimental indicator group, the data processing device obtains the effect size to be confirmed based on the ratio of the second difference value to the discrete value, thereby improving the accuracy of the effect size of the indicators in the experimental indicator group.

[0197] In an implementation method for determining discrete values ​​based on indicators in an experimental indicator group, a data processing device determines the first standard deviation of the indicators in the experimental indicator group. Determine the second standard deviation of the indicators in the control indicator group. The first standard deviation and the second standard deviation are weighted averaged to obtain a discrete value. In another implementation method for determining discrete values ​​based on indicators in the experimental indicator group, the data processing device determines the first variance of the indicators in the experimental indicator group. Determine the second variance of the indicators in the control indicator group. The first variance and the second variance are weighted averaged to obtain a discrete value. In yet another implementation method for determining discrete values ​​based on indicators in the experimental indicator group, the data processing device determines the first standard deviation of the indicators in the experimental indicator group as a discrete value. As an optional embodiment, the preset condition is to maximize the reward value, and the data processing device further performs the following steps after obtaining the reward value: maximize the reward value through Bayesian optimization, and determine the value of the target parameter as the target value.

[0198] See also Figure 2 , Figure 2 A schematic diagram of a process for obtaining a reward value provided in an embodiment of the present application. Figure 2As shown, first obtain the current gravity data, specifically, obtain the experimental indicator set and the control indicator set, wherein the gravity data includes the experimental indicator set and the control indicator set. Then group them. Specifically, divide the experimental indicator set into n experimental indicator groups according to a preset method, and divide the control indicator set into n control indicator groups according to a preset method. After obtaining n experimental indicator groups and n control indicator groups, call the mean_std.json script to perform data standardization on the n experimental indicator groups and n control indicator groups respectively. Specifically, the implementation process of calling the mean_std.json script to perform data standardization on the n experimental indicator groups can be seen in formula (2), and the implementation process of calling the mean_std.json script to perform data standardization on the n control indicator groups can be seen in formula (3). After completing data standardization, calculate the effect size of each group of indicators. Specifically, calculate the effect size of each group of experimental indicator groups to obtain the experimental effect size set. The specific implementation process can be seen in steps 3001 to 3003. In addition, after obtaining the current gravity data, the weights of the effect sizes of the indicators in each experimental indicator group are determined by calculating the weights corresponding to each group. As shown Figure 2 As shown, the last weight (i.e., the initial weight of the effect size of the indicator in the experimental indicator group) is obtained by calling the pid_p.json script, and then the p link in the proportional, integral and differential (proportional-integral-derivative, PID) control is realized by the pid_p.json script. Based on the initial weight set, the optimized weight set of the effect size of the indicators in the n experimental indicator groups is obtained, and then the optimized weight set is stored by recording the new weight through the pid_p.json script. Specifically, if the representative indicator in the experimental indicator group is less than the expected indicator, the weight of the effect size of the indicator in the experimental indicator group is increased. The specific implementation process can be referred to in step 103 to obtain the implementation process of the optimized weight set. Finally, the final reward is obtained by weighted summation. Specifically, based on the weight of the effect size of the indicator in each experimental indicator group, the effect size of the indicator in each experimental indicator group is weighted and summed to obtain the reward value, wherein the final reward is the reward value.

[0199] The present application also provides a method for determining recommended parameters. The method is performed by a device for determining recommended parameters (hereinafter referred to as a determining device). The determining device may be any electronic device that can execute the technical solution disclosed in the method embodiment of the present application. Optionally, the determining device may be one of the following: a computer, a server. Figure 3 , Figure 3 A schematic diagram of a process for determining a recommended parameter provided in an embodiment of the present application.

[0200] 301. Obtain a set of experimental indicators of a set of experimental objects.

[0201] In the embodiment of the present application, the experimental indicator set is a set of experimental indicators, and the experimental indicator is a business indicator generated by recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority. For example, the recommendation priority of multimedia content is the position of the multimedia content in the display page, and the higher the recommendation priority of the multimedia content, the more eye-catching the position of the multimedia content in the display page.

[0202] 302. Obtain an experimental effect size set and an initial weight set of the experimental effect size set.

[0203] In the embodiment of the present application, the experimental effect size set is a set of experimental effect sizes of experimental indicators, and the experimental effect size represents the improvement of the experimental indicators on the target effect. The initial weight set is a set of initial weights of the experimental effect sizes. Optionally, the experimental effect size set is obtained based on the initial weight set.

[0204] 303. When the target experimental indicator in the experimental indicator set is less than the expected indicator, increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set.

[0205] In the embodiment of the present application, in the scenario of recommending multimedia content, the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation.

[0206] 304. Based on the above optimization weight set, weighted sum is performed on the above experimental effect quantities in the above experimental effect quantity set to obtain the above reward value.

[0207] In an embodiment of the present application, in the scenario of recommending multimedia content, the reward value represents the improvement of the target effect of the service of recommending multimedia content, and the service of recommending multimedia content is the service of recommending multimedia content according to the recommendation priority of multimedia content, and the recommendation priority is determined based on the recommendation parameter. The method is used to determine the reward value in the case of determining the target value of the recommendation parameter by making the reward value meet the preset condition. In one possible implementation, the recommendation parameter is the weight in the fusion formula described above. In another possible implementation, the recommendation parameter is the boost coefficient of the strategy for recommending multimedia content. For example, in recommending multimedia content, the boost coefficient is to increase the exposure opportunity of specific multimedia content (such as new multimedia content) in the recommendation result. By adjusting the boost coefficient, the ranking score of specific multimedia content can be weighted, thereby ensuring that specific multimedia content has more opportunities to be recommended to the object. In another possible implementation, the recommendation parameter is the scattered length in the recommended multimedia content, wherein the scattered length refers to the number of multimedia content obtained by random sampling in order to increase the diversity of the recommended content during the re-arrangement stage of recommending multimedia content. In the re-ranking phase of recommending multimedia content, in addition to sorting the multimedia content according to the refined ranking score, the diversity score is also considered. The diversity score is intended to increase the diversity of the recommended content and avoid recommending a large amount of similar or repeated multimedia content to the object. To achieve this, the recommendation system will break up the refined multimedia content in the re-ranking phase, that is, obtain a certain number of multimedia content through random sampling, and these multimedia content constitute a scattered length in terms of quantity.

[0208] 305. Determine the target value of the recommended parameter by making the reward value satisfy the preset condition.

[0209] In an embodiment of the present application, in a scenario of recommending multimedia content, the value of the recommendation parameter that makes the reward value meet the preset condition is the target value. Optionally, in the case where there is no functional relationship between the reward value and the target parameter, the reward value satisfies the preset condition based on Bayesian optimization, and the value of the recommendation parameter is determined as the target value. Optionally, in the case where there is a functional relationship between the reward value and the target parameter, the target value can be obtained by solving the functional relationship by taking the reward value satisfying the preset condition as a constraint condition. For example, the target effect is to increase the number of clicks on the multimedia content, and the preset condition is to maximize the reward value. Then, the reward value satisfying the preset condition indicates that the increase in the number of clicks on the multimedia content is the largest. At this time, the expectation is to maximize the increase in the number of clicks on the multimedia content.

[0210] In the embodiment of the present application, the reward value represents the improvement of the target effect of the service of recommending multimedia content, and the service of recommending multimedia content is the service of recommending multimedia content according to the recommendation priority of multimedia content, and the recommendation priority is determined based on the recommendation parameter. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a service indicator generated by recommending multimedia content to the experimental objects in the experimental object set. The experimental effect quantity set is a set of experimental effect quantities of the experimental indicators, wherein the experimental effect quantity represents the improvement of the target effect by the experimental indicators, and the initial weight set is a set of initial weights of the experimental effect quantity. After obtaining the experimental indicator set, the experimental effect quantity set and the initial weight set of the experimental effect quantity set of the experimental object set, the determination device of the recommended parameters, when the target experimental indicator in the experimental indicator set is less than the expected indicator, increases the initial weight corresponding to the target experimental indicator in the initial weight set to increase the negative effect of the experimental effect quantity of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the process of determining the target value of the recommended parameter so that the reward value meets the preset conditions. That is, in the iterative optimization process of determining the target value of the recommended parameter by making the reward value meet the preset conditions, if the target experimental index is less than the expected index, it means that the target experimental index is lower than the expected index, and the determination device of the recommended parameter further increases the target experimental index by increasing the initial weight corresponding to the target experimental index in the initial weight set, thereby reducing the probability of the index exchange phenomenon, thereby improving the efficiency of achieving the expected improvement in the target effect by adjusting the target parameter. After the experimental effect amount in the experimental effect amount set is weighted and summed based on the optimized weight set to obtain the reward value, the target value of the recommended parameter is determined by making the reward value meet the preset conditions, thereby improving the efficiency of determining the value of the recommended parameter that makes the reward value meet the preset conditions, that is, improving the efficiency of determining the value of the recommended parameter that makes the target effect reach the expected value.

[0211] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0212] The method of the embodiment of the present application is described in detail above, and the device of the embodiment of the present application is provided below.

[0213] See also Figure 4 , Figure 4A structural schematic diagram of a data processing device provided in an embodiment of the present application, in a method executed by the data processing device 1, a reward value represents an improvement in a target effect of a target business, a parameter used to execute the target business is a target parameter, and the data processing device 1 is used to determine the reward value when the target value of the target parameter is determined by making the reward value satisfy a preset condition, and the data processing device 1 includes: an acquisition unit 11, an increase unit 12, and a processing unit 13, specifically:

[0214] An acquisition unit 11 is used to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated by executing the target business on the experimental objects in the experimental object set;

[0215] The acquisition unit 11 is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, and the experimental effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the experimental effect sizes;

[0216] An increasing unit 12 is used to increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation, and the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0217] The processing unit 13 is used to perform weighted summation on the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain the reward value.

[0218] In conjunction with any embodiment of the present application, the acquisition unit 11 is further used to acquire a control indicator set of a control object set, the control indicator set is a set of control indicators, the control indicators are business indicators generated by executing the target business on the control objects in the control object set, the control objects are objects matched with the experimental objects, and the control indicators in the control indicator set are obtained when the target parameters are initial values;

[0219] The processing unit 13 is further configured to determine, for the experimental indicator in the experimental indicator set, the expected indicator of the experimental indicator based on the control indicator in the control indicator set corresponding to the experimental indicator.

[0220] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0221] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0222] Dividing the control indicator set into n control indicator groups according to the preset method, the control indicators in the same control indicator group have a matching relationship, and the control indicators in different control indicator groups do not have a matching relationship, and the control indicator groups correspond to the experimental indicator groups one by one;

[0223] For the experimental indicator group and the control indicator group having a corresponding relationship, the experimental effect size set is determined based on the difference between the experimental indicator group and the control indicator group.

[0224] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0225] Obtaining n expected coefficients of the n control indicator groups, wherein the expected coefficients correspond one to one to the control indicator groups;

[0226] For the control indicator group and the expected coefficient having a corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups;

[0227] For the experimental indicator group and the expected indicator group having a corresponding relationship, determine a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, wherein the upper limit value of the indicator is the product of the mean value of the indicator in the expected indicator group and an upper limit coefficient, and the upper limit coefficient is greater than 1;

[0228] When the first difference is a non-positive number, a first difference between the experimental indicator group and the expected indicator group is determined as the experimental effect size.

[0229] In combination with any embodiment of the present application, the acquisition unit 11 is further used for:

[0230] When the first difference is a positive number, determining a product of the first difference and a first penalty coefficient to obtain a second difference, wherein the first penalty coefficient is a positive number less than or equal to 1;

[0231] obtaining the experimental effect size based on the difference between the first difference and the second difference;

[0232] The experimental effect size set is obtained based on the experimental effect size.

[0233] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0234] Obtaining the effect size to be confirmed based on the difference between the first difference and the second difference;

[0235] When the effect size to be confirmed is a negative number, determining the product of the effect size to be confirmed and a second penalty coefficient to obtain the experimental effect size, and the second penalty coefficient is greater than 1;

[0236] When the effect size to be confirmed is greater than the first preset value, the product of the effect size to be confirmed and the third penalty coefficient is determined to obtain the experimental effect size, and the second penalty coefficient is a positive number less than 1.

[0237] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0238] determining a difference between the first difference and the second difference to obtain a second difference;

[0239] Determine a discrete value based on the indicators in the experimental indicator group, wherein the discrete value represents the degree of discreteness of the indicators in the experimental indicator group;

[0240] The effect size to be confirmed is obtained based on the ratio of the second difference to the discrete value.

[0241] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0242] Determining a first standard deviation of an indicator within the experimental indicator group;

[0243] Determining a second standard deviation of the index within the control index group;

[0244] The first standard deviation and the second standard deviation are weighted averaged to obtain the discrete value.

[0245] In combination with any embodiment of the present application, the acquisition unit 11 is specifically used for:

[0246] Divide the n experimental indicator groups into an experimental offensive group and an experimental defensive group, wherein the indicators in the experimental offensive group match the target effect, and the experimental defensive group is an indicator group among the n experimental indicator groups except the experimental offensive group;

[0247] Determining the expected coefficient of the experimental offensive group to be a second preset value;

[0248] Determine that the expected coefficient of the experimental defense group is a third preset value, wherein the third preset value is a positive number and is less than the second preset value;

[0249] The n expected coefficients are obtained based on the second preset value and the third preset value.

[0250] In combination with any embodiment of the present application, the processing unit 13 is specifically used for:

[0251] For the experimental index and the control index having a corresponding relationship, determining that there is a corresponding relationship between the experimental index and an index in the expected index group corresponding to the control index;

[0252] An indicator in the expected indicator group corresponding to the experimental indicator is determined as the expected indicator of the experimental indicator.

[0253] In combination with any embodiment of the present application, the processing unit 13 is specifically used for:

[0254] Determine the absolute value of the difference between the target experimental index and the expected index;

[0255] Determining an increase coefficient of a target weight corresponding to the target experimental indicator based on the absolute value, wherein the increase coefficient is greater than 1 and is positively correlated with the absolute value;

[0256] Determine the product of the increase coefficient and the target weight to obtain an optimization weight corresponding to the target experimental indicator;

[0257] The optimized weight set is obtained based on the optimized weight and the weights in the initial weight set except the target weight.

[0258] In combination with any embodiment of the present application, the processing unit 13 is further configured to:

[0259] Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship;

[0260] Determine a target experimental indicator group including the target experimental indicator from the n experimental indicator groups;

[0261] The weight corresponding to the experimental indicator in the target experimental indicator group in the initial weight set is determined as the target weight.

[0262] In combination with any embodiment of the present application, the preset condition is to maximize the reward value, and the processing unit 13 is further used to:

[0263] The reward value is maximized by Bayesian optimization, and the value of the target parameter is determined as the target value.

[0264] In combination with any embodiment of the present application, the processing unit 13 is specifically used for:

[0265] Normalizing the weights in the optimized weight set to obtain a normalized weight set;

[0266] Based on the normalized weight set, the effect sizes in the experimental effect size set are weighted summed to obtain the reward value.

[0267] In the embodiment of the present application, the reward value represents the improvement of the target effect of the target business, and the parameter used to execute the target business is the target parameter. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a business indicator generated by executing the target business on the experimental objects in the experimental object set. The experimental effect quantity set is a set of experimental effect quantities of the experimental indicators, wherein the experimental effect quantity represents the improvement of the target effect by the experimental indicators, and the initial weight set is a set of initial weights of the experimental effect quantities. After the data processing device obtains the experimental indicator set, the experimental effect quantity set and the initial weight set of the experimental effect quantity set of the experimental object set, when the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to increase the negative effect of the experimental effect quantity of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the process of determining the target value of the target parameter by making the reward value meet the preset conditions. That is, in the iterative optimization process of determining the target value of the target parameter by making the reward value meet the preset conditions, if the target experimental indicator is smaller than the expected indicator, it means that the target experimental indicator is lower than the expected indicator. The data processing device then increases the initial weight corresponding to the target experimental indicator in the initial weight set to improve the target experimental indicator, thereby reducing the probability of indicator exchange, thereby improving the efficiency of achieving the desired improvement in the target effect by adjusting the target parameters.

[0268] See also Figure 5 , Figure 5 A schematic diagram of a structure of a device for determining a recommended parameter provided in an embodiment of the present application. The device 2 for determining a recommended parameter includes: an acquisition unit 21, an increase unit 22, a processing unit 23, and a determination unit 24. Specifically:

[0269] An acquisition unit 21 is used to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated after recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority;

[0270] The acquisition unit 21 is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, and the effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the effect sizes;

[0271] An increasing unit 22 is used to increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation, and the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator;

[0272] The processing unit 23 is configured to perform weighted summation of the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain a reward value, wherein the reward value represents the improvement of the target effect of the service of recommending multimedia content, wherein the service of recommending multimedia content is a service of recommending multimedia content according to the recommendation priority of the multimedia content, and the recommendation priority is determined based on the recommendation parameter;

[0273] The determination unit 24 is configured to determine a target value of the recommendation parameter by making the reward value satisfy a preset condition.

[0274] In the embodiment of the present application, the reward value represents the improvement of the target effect of the service of recommending multimedia content, and the service of recommending multimedia content is the service of recommending multimedia content according to the recommendation priority of multimedia content, and the recommendation priority is determined based on the recommendation parameter. The experimental indicator set is a set of experimental indicators, wherein the experimental indicator is a service indicator generated by recommending multimedia content to the experimental objects in the experimental object set. The experimental effect quantity set is a set of experimental effect quantities of the experimental indicators, wherein the experimental effect quantity represents the improvement of the target effect by the experimental indicators, and the initial weight set is a set of initial weights of the experimental effect quantity. After obtaining the experimental indicator set, the experimental effect quantity set and the initial weight set of the experimental effect quantity set of the experimental object set, the determination device of the recommended parameters, when the target experimental indicator in the experimental indicator set is less than the expected indicator, increases the initial weight corresponding to the target experimental indicator in the initial weight set to increase the negative effect of the experimental effect quantity of the target experimental indicator on the improvement of the target effect, thereby improving the target experimental indicator in the process of determining the target value of the recommended parameter so that the reward value meets the preset conditions. That is, in the iterative optimization process of determining the target value of the recommended parameter by making the reward value meet the preset conditions, if the target experimental index is less than the expected index, it means that the target experimental index is lower than the expected index, and the determination device of the recommended parameter further increases the target experimental index by increasing the initial weight corresponding to the target experimental index in the initial weight set, thereby reducing the probability of the index exchange phenomenon, thereby improving the efficiency of achieving the expected improvement in the target effect by adjusting the target parameter. After the experimental effect amount in the experimental effect amount set is weighted and summed based on the optimized weight set to obtain the reward value, the target value of the recommended parameter is determined by making the reward value meet the preset conditions, thereby improving the efficiency of determining the value of the recommended parameter that makes the reward value meet the preset conditions, that is, improving the efficiency of determining the value of the recommended parameter that makes the target effect reach the expected value.

[0275] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0276] Figure 6A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. The electronic device 3 includes a processor 31 and a memory 32. Optionally, the electronic device 3 also includes an input device 33 and an output device 34. The processor 31, the memory 32, the input device 33 and the output device 34 are coupled via a connector, and the connector includes various interfaces, transmission lines or buses, etc., which are not limited in the embodiments of the present application. It should be understood that in each embodiment of the present application, coupling refers to mutual connection in a specific manner, including direct connection or indirect connection through other devices, for example, it can be connected through various interfaces, transmission lines, buses, etc.

[0277] The processor 31 may include one or more processors, for example, one or more central processing units (CPUs). When the processor is a CPU, the CPU may be a single-core CPU or a multi-core CPU. Optionally, the processor 31 may be a processor set consisting of multiple CPUs, and the multiple processors are coupled to each other through one or more buses. Optionally, the processor may also be other types of processors, etc., which are not limited in the embodiments of the present application.

[0278] The memory 32 can be used to store computer program instructions and various computer program codes including the program code for executing the program code of the present application. Optionally, the memory includes but is not limited to random access memory (RAM), read-only memory (ROM), erasable programmable read only memory (EPROM), or portable read only memory (CD-ROM), which is used for related instructions and data.

[0279] The input device 33 is used to input data and / or signals, and the output device 34 is used to output data and / or signals. The input device 33 and the output device 34 can be independent devices or an integrated device.

[0280] It can be understood that in the embodiment of the present application, the memory 32 can be used not only to store relevant instructions, but also to store relevant data. The embodiment of the present application does not limit the specific data stored in the memory.

[0281] Understandably, Figure 6Only a simplified design of an electronic device is shown. In practical applications, the electronic device may also include other necessary components, including but not limited to any number of input / output devices, processors, memories, etc., and all electronic devices that can implement the embodiments of the present application are within the protection scope of the present application.

[0282] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0283] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. Those skilled in the art can also clearly understand that the descriptions of the various embodiments of the present application have different focuses. For the convenience and brevity of description, the same or similar parts may not be repeated in different embodiments. Therefore, for parts not described or not described in detail in a certain embodiment, refer to the records of other embodiments.

[0284] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or assemblies can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0285] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0286] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0287] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0288] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by a computer program to instruct the relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned method embodiments. The aforementioned storage medium includes: a read-only memory (ROM) or a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

Claims

1. A data processing method, characterized in that: The reward value represents the improvement of the target effect of the target business, the parameter used to execute the target business is the target parameter, and the method is used to determine the reward value when the target value of the target parameter is determined by making the reward value meet a preset condition, and the method includes: Acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated by executing the target business on the experimental objects in the experimental object set; Acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the experimental effect sizes represent the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the experimental effect sizes; When the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to obtain an optimized weight set, the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator; Based on the optimization weight set, the experimental effect quantities in the experimental effect quantity set are weightedly summed to obtain the reward value.

2. The method according to claim 1, characterized in that When the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased, and before obtaining the optimized weight set, the method further includes: Acquire a control indicator set of a control object set, wherein the control indicator set is a set of control indicators, wherein the control indicators are business indicators generated by executing the target business on control objects in the control object set, wherein the control objects are objects matched with the experimental objects, and the control indicators in the control indicator set are obtained when the target parameters are initial values; For the experimental indicator in the experimental indicator set, the expected indicator of the experimental indicator is determined based on the control indicator in the control indicator set corresponding to the experimental indicator.

3. The method according to claim 2, characterized in that The step of obtaining the experimental effect size set includes: Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship; Dividing the control indicator set into n control indicator groups according to the preset method, the control indicators in the same control indicator group have a matching relationship, and the control indicators in different control indicator groups do not have a matching relationship, and the control indicator groups correspond to the experimental indicator groups one by one; For the experimental indicator group and the control indicator group having a corresponding relationship, the experimental effect size set is determined based on the difference between the experimental indicator group and the control indicator group.

4. The method according to claim 3, characterized in that The step of determining the experimental effect size set based on the difference between the experimental indicator group and the control indicator group includes: Obtaining n expected coefficients of the n control indicator groups, wherein the expected coefficients correspond one to one to the control indicator groups; For the control indicator group and the expected coefficient having a corresponding relationship, determine the product of the indicator in the control indicator group and the expected coefficient to obtain n expected indicator groups; For the experimental indicator group and the expected indicator group having a corresponding relationship, determine a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, wherein the upper limit value of the indicator is the product of the mean value of the indicator in the expected indicator group and an upper limit coefficient, and the upper limit coefficient is greater than 1; When the first difference is a non-positive number, a first difference between the experimental indicator group and the expected indicator group is determined as the experimental effect size.

5. The method according to claim 4, characterized in that After determining a first difference between a mean value of an indicator in the experimental indicator group and an upper limit value of the indicator, the method further includes: When the first difference is a positive number, determining a product of the first difference and a first penalty coefficient to obtain a second difference, wherein the first penalty coefficient is a positive number less than or equal to 1; obtaining the experimental effect size based on the difference between the first difference and the second difference; The experimental effect size set is obtained based on the experimental effect size.

6. The method according to claim 4, characterized in that The obtaining the experimental effect size based on the difference between the first difference and the second difference comprises: Obtaining the effect size to be confirmed based on the difference between the first difference and the second difference; When the effect size to be confirmed is a negative number, determining the product of the effect size to be confirmed and a second penalty coefficient to obtain the experimental effect size, and the second penalty coefficient is greater than 1; When the effect size to be confirmed is greater than the first preset value, the product of the effect size to be confirmed and the third penalty coefficient is determined to obtain the experimental effect size, and the second penalty coefficient is a positive number less than 1.

7. The method according to claim 6, characterized in that The obtaining the effect size to be confirmed based on the difference between the first difference and the second difference comprises: determining a difference between the first difference and the second difference to obtain a second difference; Determine a discrete value based on the indicators in the experimental indicator group, wherein the discrete value represents the degree of discreteness of the indicators in the experimental indicator group; The effect size to be confirmed is obtained based on the ratio of the second difference to the discrete value.

8. The method according to claim 7, characterized in that The determining of discrete values ​​based on the indicators in the experimental indicator group includes: Determining a first standard deviation of an indicator within the experimental indicator group; Determining a second standard deviation of the index within the control index group; The first standard deviation and the second standard deviation are weighted averaged to obtain the discrete value.

9. The method according to claim 4, characterized in that Get n expected coefficients, including: Dividing the n experimental indicator groups into an experimental offensive group and an experimental defensive group, wherein the indicators in the experimental offensive group match the target effect, and the experimental defensive group is an indicator group other than the experimental offensive group among the n experimental indicator groups; Determining the expected coefficient of the experimental offensive group to be a second preset value; Determine that the expected coefficient of the experimental defense group is a third preset value, wherein the third preset value is a positive number and is less than the second preset value; The n expected coefficients are obtained based on the second preset value and the third preset value.

10. The method according to claim 4, characterized in that The determining the expected indicator of the experimental indicator based on the control indicator in the control indicator set corresponding to the experimental indicator includes: For the experimental index and the control index having a corresponding relationship, determining that there is a corresponding relationship between the experimental index and an index in the expected index group corresponding to the control index; An indicator in the expected indicator group corresponding to the experimental indicator is determined as the expected indicator of the experimental indicator.

11. The method according to any one of claims 1 to 10, characterized in that The step of increasing the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set includes: Determine the absolute value of the difference between the target experimental index and the expected index; Determining an increase coefficient of a target weight corresponding to the target experimental indicator based on the absolute value, wherein the increase coefficient is greater than 1 and is positively correlated with the absolute value; Determine the product of the increase coefficient and the target weight to obtain an optimization weight corresponding to the target experimental indicator; The optimized weight set is obtained based on the optimized weight and the weights in the initial weight set except the target weight.

12. The method according to claim 11, characterized in that Before determining the increase coefficient of the target weight corresponding to the target experimental indicator based on the absolute value, the method includes: Dividing the experimental indicator set into n experimental indicator groups according to a preset method, the experimental indicators in the same experimental indicator group have a matching relationship, and the experimental indicators in different experimental indicator groups do not have a matching relationship; Determine a target experimental indicator group including the target experimental indicator from the n experimental indicator groups; The weight corresponding to the experimental indicator in the target experimental indicator group in the initial weight set is determined as the target weight.

13. The method according to any one of claims 1 to 10, characterized in that The preset condition is to maximize the reward value. After obtaining the reward value, the method further includes: The reward value is maximized by Bayesian optimization, and the value of the target parameter is determined as the target value.

14. The method according to any one of claims 1 to 10, characterized in that The step of performing weighted summation on the effect amounts in the experimental effect amount set based on the optimization weight set to obtain the reward value includes: Normalizing the weights in the optimized weight set to obtain a normalized weight set; Based on the normalized weight set, the effect sizes in the experimental effect size set are weighted summed to obtain the reward value.

15. A method for determining a recommended parameter, characterized in that: The method comprises: Acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated after recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority; Acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the effect sizes; In the case where the target experimental indicator in the experimental indicator set is less than the expected indicator, the initial weight corresponding to the target experimental indicator in the initial weight set is increased to obtain an optimized weight set, the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator; Based on the optimization weight set, weighted summing the experimental effect quantities in the experimental effect quantity set is performed to obtain a reward value, wherein the reward value represents an improvement in a target effect of a service of recommending multimedia content, wherein the service of recommending multimedia content is a service of recommending multimedia content according to a recommendation priority of the multimedia content, wherein the recommendation priority is determined based on a recommendation parameter; By making the reward value satisfy a preset condition, a target value of the recommended parameter is determined.

16. A data processing device, characterized in that: The reward value represents the improvement of the target effect of the target business, the parameter used to execute the target business is the target parameter, and the data processing device is used to determine the reward value when the target value of the target parameter is determined by making the reward value meet a preset condition, and the data processing device includes: An acquisition unit, configured to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated by executing the target business on the experimental objects in the experimental object set; The acquisition unit is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, and the experimental effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the experimental effect sizes; an increasing unit, configured to increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the target business when the target effect of the target business reaches the expectation, the effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator; A processing unit is used to perform weighted summation on the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain the reward value.

17. A device for determining recommended parameters, characterized in that: The recommended parameter determination device comprises: An acquisition unit, configured to acquire an experimental indicator set of an experimental object set, wherein the experimental indicator set is a set of experimental indicators, and the experimental indicators are business indicators generated after recommending multimedia content to the experimental objects in the experimental object set according to the recommendation priority; The acquisition unit is used to acquire an experimental effect size set and an initial weight set of the experimental effect size set, wherein the experimental effect size set is a set of experimental effect sizes of the experimental indicators, the effect size represents the improvement of the experimental indicators on the target effect, and the initial weight set is a set of initial weights of the effect sizes; an increasing unit, configured to increase the initial weight corresponding to the target experimental indicator in the initial weight set to obtain an optimized weight set when the target experimental indicator in the experimental indicator set is less than the expected indicator, wherein the expected indicator is a business indicator generated by the business of recommending multimedia content when the target effect reaches the expectation, the experimental effect size of the target experimental indicator is positively correlated with the target difference, and the target difference is the difference between the target experimental indicator and the expected indicator; a processing unit, configured to perform weighted summation of the experimental effect quantities in the experimental effect quantity set based on the optimization weight set to obtain a reward value, wherein the reward value represents an improvement in a target effect of a service of recommending multimedia content, wherein the service of recommending multimedia content is a service of recommending multimedia content according to a recommendation priority of the multimedia content, wherein the recommendation priority is determined based on a recommendation parameter; A determination unit is used to determine a target value of the recommendation parameter by making the reward value meet a preset condition.

18. An electronic device, characterized in that: include: A processor and a memory, wherein the memory is used to store computer program code, wherein the computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes the method as claimed in any one of claims 1 to 14, or the electronic device executes the method as claimed in claim 15.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor is caused to execute the method described in any one of claims 1 to 14, or the processor is caused to execute the method described in any one of claim 15.

20. A computer program product, characterized in that The computer program product includes a computer program or instructions; when the computer program or instructions are run on a computer, the computer is enabled to execute the method described in any one of claims 1 to 14, or the computer is enabled to execute the method described in any one of claim 15.

Images