A business processing method, device, equipment and computer storage medium

Abstract

The present application provides a business processing method, device, equipment and computer storage medium, which relates to the field of computer technology and is used to automatically estimate the probability of an account's operation intention to transfer electronic resources for a specific operation. The method includes: using the account feature data of each historical account in the historical data and the correlation between the conversion operation results of each historical account in their respective non-trial periods, based on the account feature data of the account to be predicted, estimating the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period, wherein the target behavior includes the operation that can be used after the conversion operation of transferring electronic resources in the non-trial period; the account feature data includes the attribute feature data of the account and the behavior feature data of the account in the trial period of the target behavior. The method can automatically estimate the probability threshold of the operation intention of the account to be predicted without relying on manual work, thereby improving the efficiency of estimating the probability threshold of the operation intention of the account to perform the conversion operation.

Description

Technical Field

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

[0002] In the related art, for designated operations with a trial period, an account can use the above-mentioned specific operations during the trial period without transferring electronic resources, but the account can continue to use the above-mentioned designated operations after transferring electronic resources for the above-mentioned specific operations during the non-trial period; however, when estimating whether an account will transfer electronic resources for the above-mentioned designated operations during the non-trial period, the operation data corresponding to the account is often manually analyzed, and whether the account will transfer electronic resources during the non-trial period is estimated based on the analysis results. Therefore, how to automate the composition of estimating whether an account will transfer electronic resources for a specific operation and perform business processing based on the estimation results has become an issue that needs to be considered. Summary of the invention

[0003] The embodiments of the present application provide a business processing method, apparatus, device and computer storage medium for automatically estimating at least the probability of an account's intention to transfer electronic resources for a specific operation.

[0004] In a first aspect, the present application provides a business processing method, including:

[0005] Acquire account characteristic data of the account to be predicted, the account characteristic data including attribute characteristic data of the account to be predicted and behavior characteristic data of the account to be predicted in the trial period of the target behavior, the attribute characteristic data is used to characterize the estimated value of the electronic resource holding amount of the account to be predicted, the behavior characteristic data is used to characterize the account's preference for the target behavior, the behavior characteristic data characterizes the account's preference for the target behavior, the target behavior includes an operation that can be used after a conversion operation of transferring electronic resources is performed in a non-trial period, and the non-trial period is a period after the end of the trial period;

[0006] By using the correlation between the account characteristic data of each historical account and the conversion operation results of each historical account in the respective non-trial period in the historical data, based on the account characteristic data of the account to be predicted, the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period is estimated, wherein the conversion operation result includes the result of whether the conversion operation is performed, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in the respective trial period;

[0007] Business processing is performed on the account to be predicted according to the operation intention probability value of the account to be predicted performing a conversion operation in a non-trial period.

[0008] In a possible implementation, the target application for implementing the target behavior includes an accelerator, the target behavior includes an operation of accelerating a game application through the accelerator, and the behavior characteristic data further includes at least one of the following data:

[0009] Game access information of the account accessing the game application during the trial period;

[0010] Information about similar applications accessed by the account during the trial period, where the similar applications include other accelerators besides the accelerator.

[0011] In a possible implementation manner, the target behavior operation information includes at least one of the following information:

[0012] Information about changes in the network efficiency of the game application after the account accelerates the game application through the accelerator during the trial period;

[0013] The acceleration duration information of the account using the accelerator to accelerate the game application during the trial period;

[0014] The success rate of the account accelerating the game application through the accelerator during the trial period.

[0015] In a possible implementation, after the account accelerates the game application through the accelerator during the trial period, the network efficiency change information of the game application includes at least one of the following information:

[0016] The average network delay of the game application after the account accelerates the game application through the accelerator during the trial period, wherein the average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application;

[0017] The ratio of the second average network delay of the game application in the first sub-period of the account to the second average network delay of the game application in the second sub-period, wherein the average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application, the first sub-period is a sub-period of the trial period, the second sub-period is a sub-period of the trial period, and the first sub-period is an adjacent period after the end of the second sub-period.

[0018] In a possible implementation manner, the acceleration duration information of the account using the accelerator to accelerate the game application during the trial period includes at least the following information:

[0019] The total duration of acceleration of the game application by the account through the accelerator during the trial period;

[0020] the total duration of acceleration of the account in a third sub-period, where the third sub-period is a sub-period of the trial period;

[0021] The ratio of the total acceleration duration of the account in the fourth sub-period to the total acceleration duration in the fifth sub-period, wherein the fourth sub-period is a sub-period of the trial period, the fifth sub-period is a sub-period of the trial period, and the fourth sub-period is an adjacent period after the end of the fifth sub-period.

[0022] In a possible implementation manner, the attribute characteristic data further includes at least one of the following data used to characterize the network preference degree of the third-party account:

[0023] Group information of the third-party account in an application corresponding to the third-party account;

[0024] The account access information of the third-party account accessing the application corresponding to the third-party account during a specified period of time.

[0025] In a second aspect of the present application, a service processing device is provided, including:

[0026] A feature extraction unit, used to obtain account feature data of the account to be predicted, the account feature data including attribute feature data of the account to be predicted and behavior feature data of the account to be predicted in a trial period of a target behavior, the attribute feature data being used to characterize an estimated value of the electronic resource holding amount of the account to be predicted, the behavior feature data being used to characterize a preference of the account for the target behavior, the behavior feature data characterizing a preference of the account to be predicted for the target behavior, the target behavior including an operation that can be used after a conversion operation of transferring electronic resources is performed in a non-trial period, the non-trial period being a period after the end of the trial period;

[0027] An operation intention prediction unit is used to estimate the operation intention probability value of the account to be predicted to perform a conversion operation in the non-trial period based on the account characteristic data of the account to be predicted, using the account characteristic data of each historical account in the historical data and the correlation degree of the conversion operation results of each historical account in the respective non-trial period, wherein the conversion operation result includes the result of whether the conversion operation is performed, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in the respective trial period;

[0028] The business processing unit is used to perform business processing on the account to be predicted according to the operation intention probability value of the account to be predicted to perform conversion operation in the non-trial period.

[0029] In a possible implementation, the operation intention prediction unit is specifically used to:

[0030] The trained operation intention prediction model is used, the account feature data of the account to be predicted is input, and the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period output by the operation intention prediction model is obtained. The operation intention prediction model is based on an ensemble learning method, and is trained by using the account feature data of each historical account and the operation data of whether each historical account performs the conversion operation in its respective non-trial period as training samples. The trained operation intention prediction model includes multiple classifiers and a decision maker, wherein:

[0031] Each classifier makes a judgment based on the account feature data of the account to be predicted and outputs a first classification result, wherein the first classification result is used to indicate the degree of association between the account feature data and the conversion operation result in the non-trial period;

[0032] The decision maker obtains the operation intention probability value of the account to be predicted to perform the conversion operation in a non-trial period according to the first classification results output by each classifier.

[0033] In a possible implementation manner, the operation intention prediction unit is further used to:

[0034] The trained operation intention prediction model is obtained by:

[0035] Extract the account feature data of the historical accounts in each training sample and the operation data of whether the historical accounts have been converted during the non-trial period;

[0036] Determine a second classification result of each training sample according to the operation data of whether the historical account in each training sample performs the conversion operation in the non-trial period, wherein the second classification result is used to indicate the operation result of whether each historical account performs the conversion operation in the respective non-trial period;

[0037] The operation intention prediction model obtained through historical training or the initial operation intention prediction model is adjusted by using the mapping relationship between the account feature data of the historical accounts in each training sample and the second classification result to obtain the trained operation intention prediction model.

[0038] In a possible implementation manner, the operation intention prediction unit is further used to:

[0039] Perform at least one of the following data processing on the historical data, and determine the historical data after the data processing as the training sample:

[0040] Discarding historical data of the conversion operation during the trial period;

[0041] Discarding historical data of the conversion operation performed by setting the resource conversion mode during the non-trial period;

[0042] Discard historical data whose total duration of the target behavior is less than the duration threshold.

[0043] In a possible implementation, the behavior characteristic data includes at least one of the following data:

[0044] Target behavior operation information of the account performing the target behavior during the trial period;

[0045] Target application access information of the target application that the account accesses to achieve the target behavior during the trial period;

[0046] The account's network environment information during the trial period;

[0047] The account's network usage during the trial period.

[0048] In a possible implementation, the target application for implementing the target behavior includes an accelerator, the target behavior includes an operation of accelerating a game application through the accelerator, and the behavior characteristic data further includes at least one of the following data:

[0049] Game access information of the account accessing the game application during the trial period;

[0050] The access information of the account to similar applications during the trial period, wherein the similar applications include other accelerators other than the accelerator.

[0051] In a possible implementation, the target behavior operation information includes at least one of the following information:

[0052] Information about changes in the network efficiency of the game application after the account accelerates the game application through the accelerator during the trial period;

[0053] The acceleration duration information of the account using the accelerator to accelerate the game application during the trial period;

[0054] The success rate of the account accelerating the game application through the accelerator during the trial period.

[0055] In a possible implementation, after the account accelerates the game application through the accelerator during the trial period, the network efficiency change information of the game application includes at least one of the following information:

[0056] The average network delay of the game application after the account accelerates the game application through the accelerator during the trial period, wherein the average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application;

[0057] The ratio of the second average network delay of the game application in the first sub-period of the account to the second average network delay of the game application in the second sub-period, wherein the average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application, the first sub-period is a sub-period of the trial period, the second sub-period is a sub-period of the trial period, and the first sub-period is an adjacent period after the end of the second sub-period.

[0058] In a possible implementation, the acceleration duration information of the account using the accelerator to accelerate the game application during the trial period includes at least one of the following information:

[0059] The total duration of acceleration of the game application by the account through the accelerator during the trial period;

[0060] the total duration of acceleration of the account in a third sub-period, where the third sub-period is a sub-period of the trial period;

[0061] The ratio of the total acceleration duration of the account in the fourth sub-period to the total acceleration duration in the fifth sub-period, wherein the fourth sub-period is a sub-period of the trial period, the fifth sub-period is a sub-period of the trial period, and the fourth sub-period is an adjacent period after the end of the fifth sub-period.

[0062] In a possible implementation, the account includes a third-party account for logging into a target application for implementing the target behavior, and the attribute characteristic data includes at least one of the following data:

[0063] Account setting information of the third-party account;

[0064] Creation time information of the third-party account;

[0065] Account level information of the third-party account.

[0066] In a possible implementation, the attribute characteristic data further includes at least one of the following data used to characterize the network preference degree of the third-party account:

[0067] Group information of the third-party account in an application corresponding to the third-party account;

[0068] The account access information of the third-party account accessing the application corresponding to the third-party account during a specified period of time.

[0069] In a possible implementation manner, the service processing unit is specifically configured to:

[0070] If the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period is less than the intention probability threshold, the electronic media setting information is sent to the account to be predicted so that the account to be predicted performs the conversion operation by setting the resource conversion mode.

[0071] In a third aspect of the present application, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in the first aspect and any possible implementation manner is implemented.

[0072] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed on a computer, the computer executes the method described in the first aspect and any possible implementation manner.

[0073] Since the embodiment of the present application adopts the above technical solution, it has at least the following technical effects:

[0074] In the implementation example of the present application, by utilizing the account characteristic data of the historical account and the correlation between the conversion operation results of whether the conversion operation was performed in the non-trial period, it is possible to automatically estimate the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period based on the account characteristic data of the account to be predicted, and then perform business processing on the account to be predicted based on the estimated operation intention probability value. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 An example diagram of an application scenario of business processing provided in an embodiment of the present application;

[0076] Figure 2 An example diagram of a business process flow provided in an embodiment of the present application;

[0077] Figure 3 An example diagram of the principle of estimating the probability value of an operation intention provided in an embodiment of the present application;

[0078] Figure 4 A structural diagram of a service processing device provided in an embodiment of the present application;

[0079] Figure 5 A structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0080] In order to better understand the technical solution provided by the embodiments of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0081] In order to help those skilled in the art better understand the technical solution of the present application, the technical terms involved in the present application are explained below.

[0082] Target behavior: operations that meet the conditions that the electronic resources can be used during the trial period without conversion operations, and can only be used after conversion operations are performed during the non-trial period; target behaviors may include operations that change the network efficiency of a second application through a first application, such as operations that accelerate a game application through an accelerator; target behaviors may also include operations that improve the efficiency of a certain behavior in a specified application, such as game equipment usage operations in game applications that use specified game equipment to improve the combat ability of game characters; target behaviors may also include operations that obtain permissions for a certain service in a specified application, such as operations that obtain permissions to watch specified types of videos in a video application.

[0083] Electronic resources: The electronic resources in the embodiments of the present application include funds used to purchase designated resources, such as funds for purchasing the right to use the target behavior, which may at least include legal tender, electronic currency, etc.; the so-called legal tender refers to a currency that is legally authorized for compulsory circulation and use, such as RMB, US dollars, etc.; the so-called electronic currency refers to the currency stored in electronic form in an electronic wallet held by an account (such as QQ wallet, WeChat wallet, etc.), and electronic currency may include but is not limited to electronic tickets, digital currency (an unregulated, digital currency), game resources (such as game coins, game equipment, etc.), etc.

[0084] Conversion operation: The conversion operation of the embodiment of the present application includes the operation of transferring electronic resources, such as the payment operation of transferring digital currency in an account, or the game payment operation of transferring game coins in a game application, or the game resource conversion operation of transferring game equipment in a game application, etc.

[0085] Ensemble Learning: Sometimes also generally referred to as the boosting method, it is widely used in classification and regression tasks. It uses some different methods to change the distribution of the original training samples to build multiple different classifiers, and then uses the voting method for the classification results of each classifier to obtain the final classification result.

[0086] Extreme Gradient Boosting (xgboost): A decision tree-based ensemble learning algorithm that uses the Gradient Boosting framework; it often performs better than other algorithms or frameworks when predicting problems related to unstructured data (such as images, text, etc.).

[0087] The design concept of this application is described below.

[0088] In the related art, for specific operations that change the network efficiency of a specified application, or specific operations that obtain certain permissions in a specified application, or specific operations that improve the effects of certain operations in a specified application, such specified applications often have corresponding trial periods and non-trial periods. During the trial period, an account can use the above-mentioned specific operations without transferring electronic resources, but during the non-trial period, the account must transfer electronic resources for the above-mentioned specific operations before continuing to use the above-mentioned specified operations; however, when estimating whether an account will transfer electronic resources for a specific operation during the non-trial period, the operator of the application that implements the above-mentioned specific operation often analyzes the account's operation data to determine the estimation result. For example, for an accelerator with a trial period, when estimating whether an account newly connected to the accelerator will transfer electronic resources for accelerated behavior during the non-trial period, the operator often needs to analyze the behavior data of the account after connecting to the accelerator, and estimate whether the account will transfer electronic resources for accelerated behavior during the non-trial period based on the analysis results. This process requires a lot of manpower and material resources, and the estimation result is heavily dependent on the business experience of the operator.

[0089] In view of this, the inventors have designed a business processing method, apparatus, device and computer storage medium. Considering that whether an account will transfer electronic resources in response to a target behavior is highly correlated with the account's electronic resource holdings and the degree of preference for the target behavior, in an embodiment of the present application, based on the estimated value of the electronic resource holdings of the account to be predicted and the account characteristic data of the predicted account's preference for the target behavior during the trial period of the target behavior, it is estimated whether the account to be predicted will transfer electronic resources in response to the target behavior during the non-trial period. On the other hand, considering that accounts using the target behavior usually have similar behavioral characteristics, in an embodiment of the present application, the account characteristic data of the account to be predicted is combined with the account characteristic data of each historical account and the degree of correlation between each historical account and whether the electronic resources were transferred during their respective non-trial periods to estimate the probability of the account to be predicted's operational intention to transfer electronic resources for the target behavior during the non-trial period.

[0090] Here, the trial period and the non-trial period in the embodiment of the present application are described:

[0091] The non-trial period is the time period after the end of the trial period. The non-trial period can be an adjacent time period after the end of the trial period, or a non-adjacent time period after the end of the trial period. The trial period can be a time period in which the account has the authority to perform the target behavior without performing a conversion operation. The trial period can be a time period from the moment the account first obtains the authority to perform the target behavior to a set period of time. The trial period can also be a time period from the moment the account first accesses the target application for implementing the target behavior to a set period of time. Technical personnel in this field can set the above-mentioned trial period and non-trial period according to actual needs.

[0092] If the target behavior is the game equipment usage operation of designated game equipment, the trial period can be the time period from the time when the account first obtains the use of the above-mentioned designated game equipment to m1 hours or m2 days later, and the non-trial period is the time period m1 hours after the account first obtains the use of the above-mentioned designated game equipment. Both m1 and m2 are positive numbers and there is no limit on the values ​​of m1 and m2, such as setting m1 to 24 and m2 to 3, etc.; when the target behavior is the operation of accelerating the game application through an accelerator, the trial period can be the time period from the start time of the account accessing the accelerator (i.e., the target application) to m3 hours or m4 days after the start time. Both m3 and m4 are positive numbers and there is no limit on the values ​​of m3 and m4, such as setting m3 to 24 and m4 to 3, etc. At this time, the trial period is within 3 days after the account accesses the accelerator, and the non-trial period is from the 4th to the 10th day after the account accesses the accelerator, or the non-trial period is from the 6th to the 10th day after the account accesses the accelerator.

[0093] In order to more clearly understand the design concept of the present application, a detailed description is given below in conjunction with the accompanying drawings.

[0094] Please refer to Figure 1 , which is a diagram showing an example of an application scenario of a business process. The application scenario includes multiple terminal devices 100 and a server 200. A user can access a first application and a second application on each terminal device 100 through an account, and the terminal device 100 and the server 200 can communicate with each other.

[0095] After the account accesses the first application, the terminal device 100 can collect data related to the account and send the collected data related to the account to the server 200 through the communication network between the terminal device 100 and the server 200 .

[0096] The server 200 receives the account-related data sent by the terminal device 100, and extracts the account feature data of the account therefrom, and uses the correlation between the account feature data of each historical account and the conversion operation results of each historical account in the respective non-trial periods in the historical data, and estimates the probability value of the operation intention of the above-mentioned account to be predicted to perform a conversion operation in the non-trial period based on the account feature data of the above-mentioned account to be predicted, and the above-mentioned conversion operation includes the result of whether the above-mentioned conversion operation is performed.

[0097] based on Figure 1 The following is an example of the business processing method involved in the embodiment of the present application; please refer to Figure 2 , indicating that the service processing method designed in the embodiment of the present application specifically includes the following steps:

[0098] Step S201, obtaining account characteristic data of the account to be predicted, the account characteristic data including attribute characteristic data of the account to be predicted and behavior characteristic data of the account to be predicted in the trial period of the target behavior, the target behavior including operations that can be used after the conversion operation of transferring electronic resources in the non-trial period, the non-trial period being the adjacent period after the end of the trial period.

[0099] Among them, the above-mentioned attribute characteristic data is used to characterize the estimated value of the electronic resource holdings of the above-mentioned account to be predicted, and the behavior characteristic data characterizes the preference degree of the above-mentioned account to be predicted for the above-mentioned target behavior; and the introduction to the target behavior, electronic resources and conversion operations can be found in the above content, and will not be repeated here.

[0100] Considering that the situation of the account performing the above-mentioned target behavior intuitively reflects the account's preference for the target behavior, the higher the preference, the higher the account's willingness to perform the conversion operation; on the other hand, the situation of accessing the target application to achieve the target behavior, the account's network environment information, and the account's network usage level can all reflect the possibility of the account performing the target operation to a certain extent. Therefore, the behavioral characteristic data in the embodiments of the present application may include at least one of the account's target behavior operation information during the trial period, target application access information of the target application to achieve the target behavior, network environment information, and network usage level.

[0101] As an embodiment, the electronic resource holdings of an account will largely affect whether the account performs a conversion operation, and the account's network usage level will also affect whether the account performs a conversion operation. When the account includes a third-party account for logging into the target application, the account information of the third-party account can largely reflect the electronic resource holdings of the account. Therefore, in the embodiment of the present application, at least one of the account setting information, creation time information, and account level information of the third-party account can be used as attribute feature data to characterize the estimated value of the electronic resource holdings of the account to be predicted.

[0102] Step S202, utilizing the account characteristic data of each historical account in the historical data and the correlation between the conversion operation results of each historical account in their respective non-trial periods, based on the account characteristic data of the account to be predicted, estimating the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period, wherein the conversion operation result includes the result of whether the above-mentioned conversion operation is performed, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in their respective trial periods.

[0103] As an embodiment, in the embodiment of the present application, machine learning can be used to learn the correlation between the account feature data of each historical account in the historical data and the conversion operation results of each historical account in the non-trial period, and based on the learned correlation, the probability of the operation intention to perform the conversion operation in the non-trial period represented by the account feature data of the account to be predicted is estimated; such as learning the above correlation through supervised classification learning, unsupervised classification learning, reinforcement learning, deep reinforcement learning and other machine learning methods.

[0104] Step S203 , performing business processing on the account to be predicted according to the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period.

[0105] As an embodiment, after estimating the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period, the account to be predicted can be processed according to the estimated operation intention probability value. Specifically, if the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period is less than the intention probability threshold, the electronic media setting information can be sent to the account to be predicted, so that the account to be predicted performs the above conversion operation by setting the resource conversion method.

[0106] Among them, the electronic resource is the electronic currency in the electronic wallet corresponding to the account, and when the conversion operation is a payment operation for transferring electronic currency for the target behavior, the above-mentioned set electronic media information can be an electronic currency coupon such as an electronic currency coupon that can be used within a specified time period or an electronic currency coupon with no time limit for use. In this case, the above-mentioned set resource conversion method is a discount payment method.

[0107] The following is a detailed description of the account characteristic data of the account to be predicted and the historical account in the embodiment of the present application, where the account can refer to both the account to be predicted and the historical account.

[0108] The behavior characteristic data in the account characteristic data of the embodiment of the present application may include at least one of the following behavior characteristic data A1 to behavior characteristic data A4:

[0109] Behavior characteristic data A1: target behavior operation information of the account performing the above target behavior during the trial period.

[0110] The target behavior operation information may include at least one of the following information: the number of target behaviors during the trial period, the duration of the target behaviors, and the network revenue information brought to the account by the target operation.

[0111] Behavior characteristic data A2: target application access information of the target application accessed by the account to achieve the target behavior during the trial period.

[0112] The longer the account accesses the target application and the more times it accesses, the greater the possibility that the account will perform the target behavior through the target application. Therefore, the above-mentioned target application access information may include at least one of the total duration of the account accessing the target application during the trial period, the total number of times the target application is accessed during the trial period, the specific period of time during which the target application is accessed during the trial period, and the specific time when the target application is accessed during the trial period.

[0113] Behavior characteristic data A3: network environment information of the account during the trial period.

[0114] The above network environment information may include network speed, network stability, etc.

[0115] Behavioral characteristic data A4: the network usage level of the account during the trial period.

[0116] The higher the network usage level of the account during the trial period, the higher the possibility of using the target behavior, and thus the higher the possibility of converting the target behavior. Therefore, the above-mentioned network usage level may include but is not limited to the total time the account is connected to the network during the trial period, the specific time period when the account is connected to the network during the trial period, the total number of times the account is connected to the network during the trial period, etc.

[0117] The following are examples of possible scenarios for behavior feature data for different target behaviors:

[0118] Example 1:

[0119] In this example, the target application is a game application, and the target behavior is a game equipment usage operation using specified game equipment in the game application.

[0120] The target behavior operation information in the behavior characteristic data may be information on the use of game equipment in a game application by an account during the trial period of the specified game equipment, which may include at least one of the following information: the number of times the specified game equipment is used during the trial period, the duration of use of the specified game equipment, the date type of the date to which the specific moment of use of the specified game equipment belongs, and network revenue information brought to the account by the use of the specified game equipment. The network revenue information may be the number of victories in a competition activity in the game application. The date type in the embodiment of the present application may include working days (Monday to Friday) and non-working days (Saturday and Sunday).

[0121] The above-mentioned target application access information may include at least one of the following information: the duration of the account's access to the game application during the trial period, the number of times the game application is accessed during the trial period, the device information of the device used to access the game application during the trial period, the specific time or specific period of access to the game application, etc.

[0122] The network environment information of the above-mentioned account during the trial period may include: at least one of the network speed of the account accessing the game application during the trial period, the network stability of the account during the trial period, and the like.

[0123] The network usage level of the above account during the trial period may include: information about the duration of the account's access to the network during the trial period, information about the number of times the account accesses the network during the trial period, and the like.

[0124] Example 2:

[0125] In this example, the target application is an accelerator, and the target behavior includes an operation of accelerating a game application through the accelerator.

[0126] For ease of explanation, the trial period in this example includes the time of accessing the accelerator and a period of 3 days from the date of accessing the accelerator. The trial period in this example is expressed as a 3-day trial period below; however, the 3-day trial period is only for ease of understanding and is not a limitation on the trial period in this example. Technical personnel in this field can limit the period in this example according to actual needs.

[0127] Considering that when an account accelerates a game application through an accelerator, the acceleration effect (such as the acceleration success rate and the network efficiency of the game application after acceleration, etc.) will have a greater impact on the account's acceleration experience, and the acceleration experience will largely affect the account's preference for the target behavior. The better the account's acceleration experience, the greater the possibility that the account will perform a conversion operation; therefore, in the example, the target behavior operation information in the behavior feature data of the account during the trial period may include, but is not limited to, at least one of the following acceleration information B1 to acceleration information B3.

[0128] Acceleration information B1: information about changes in the network efficiency of the game application after the account uses an accelerator to accelerate the game application during the trial period.

[0129] The network efficiency change information of the game application may include the average network delay of the game application after the account accelerates the game application through an accelerator during the trial period; the average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application; for example, the average value of the network delay of the network transmission channel created by the accelerator in the game application during the 3-day trial period is determined as the above-mentioned network efficiency change information.

[0130] Furthermore, the stability of the acceleration effect of the account using the accelerator to accelerate the game application in different sub-periods of the trial period will affect the account's acceleration experience, that is, the higher the stability of the acceleration effect, the higher the account's preference for the target behavior, and the stronger the account's willingness to perform conversion operations. Therefore, the stability of the acceleration effect of the account using the accelerator to accelerate the game application during the trial period can be determined as the network efficiency change information of the above-mentioned game application.

[0131] Therefore, the network efficiency change information can be the ratio of the second average network delay of the game application of the account in the first sub-period and the second average network delay of the game application in the second sub-period, wherein the above average network delay includes the average value of the network delay of the network transmission channel created by the accelerator in the game application, the above first sub-period is a sub-period of the trial period, the second sub-period is a sub-period of the trial period, and the first sub-period is the adjacent period after the end of the second sub-period. For example, the first sub-period is the second day of the 3-day trial period, and the second sub-period is the third day of the 3-day trial period. The ratio of the average network delay of the game application on the above second day to the average network delay of the game application on the above first day can be determined as the network efficiency change information; the first sub-period is the third day of the 3-day trial period, and the second sub-period is the second day of the 3-day trial period. The ratio of the average network delay of the game application on the above third day to the average network delay of the game application on the above second day can be determined as the network efficiency change information.

[0132] Acceleration information B2: acceleration duration information of the account using the accelerator to accelerate the game application during the trial period.

[0133] Furthermore, the total acceleration duration of the account during the trial period, the specific period of acceleration, and the stability of the acceleration duration can reflect the degree of dependence of the account on the operation of accelerating the game application through the accelerator. The higher the degree of dependence, the higher the preference of the account for the target behavior and the greater its willingness to perform the conversion operation. Therefore, the above acceleration duration information includes at least one of the following acceleration duration information C1 to acceleration duration information C3:

[0134] Acceleration duration information C1: the total duration of acceleration used by the account to accelerate the game application through the accelerator during the trial period; for example, the total duration of acceleration used by the account to accelerate the game application during the 3-day trial period.

[0135] Acceleration duration information C2: the total acceleration duration of the account in the third sub-period, where the third sub-period is a sub-period of the trial period;

[0136] The third sub-period may include but is not limited to at least one of the working period, entertainment period, and sleeping period; the above-mentioned working period may include the time period from 7:00 in the morning to 6:00 in the evening every day, the above-mentioned entertainment period may include the time period from 18:00 to 23:00 in the evening every day, and the above-mentioned sleeping period may include the time period from 23:00 to 24:00 and 0:00 to 7:00 in the evening every day, where 24:00 of one day and 0:00 of the next day are the same expression. Technical personnel in this field can also divide the third sub-period of the trial period according to actual needs.

[0137] Moreover, the total acceleration duration of the third sub-period can reflect the life status of the user corresponding to the account to a certain extent. The user's life status can reflect to a certain extent whether the user has sufficient time to play games. When the user has sufficient time to play games, the greater the possibility that the account corresponding to the user will accelerate the game application through the accelerator, the stronger the willingness to perform the conversion operation.

[0138] Acceleration duration information C3: the ratio of the total acceleration duration of the account in the fourth sub-period to the total acceleration duration in the fifth sub-period, wherein the fourth sub-period is a sub-period of the trial period, the fifth sub-period is a sub-period of the trial period, and the fourth sub-period is the adjacent period after the end of the fifth sub-period.

[0139] For example, if the fourth sub-period is the second day of the three-day trial period, and the fifth sub-period is the first day of the three-day trial period, the ratio of the total acceleration time of the game application on the second day to the total acceleration time of the game application on the first day can be determined as the acceleration time information; if the fourth sub-period is the third day of the three-day trial period, and the fifth sub-period is the second day of the three-day trial period, the ratio of the total acceleration time of the game application on the third day to the total acceleration time of the game application on the second day can be determined as the network efficiency change information.

[0140] Acceleration information B3: the success rate of the account accelerating the game application through the accelerator during the trial period.

[0141] Due to factors such as the network environment, when an account uses an accelerator to accelerate a gaming application, acceleration failure may occur. Acceleration failure will reduce the account's preference for the target behavior, thereby affecting the account's willingness to perform conversion operations.

[0142] As an embodiment, considering that when an account accelerates a game application through an accelerator, some information of the account accessing the accelerator will, to a certain extent, affect the account's willingness to accelerate the game application through the accelerator, and further affect the account's intention to perform a conversion operation. Specifically, the target application access information in the behavioral data characteristics of the above account during the trial period may include at least one of the following information:

[0143] The duration of the account's access to the accelerator during the trial period, the number of times the account accesses the accelerator during the trial period, and the first access information of the account when it first accesses the accelerator during the trial period.

[0144] The above-mentioned first access information may include information such as the accelerator promotion link corresponding to the account's first access to the accelerator during the trial period, the time information of the first access to the above-mentioned accelerator, etc.; the accelerator promotion link may be an advertising link created by the accelerator operator through various advertising website platforms to promote the accelerator; the above-mentioned time information of the first access to the accelerator may be the specific time of the first access to the accelerator, or the date type of the date to which the specific time of the first access to the accelerator belongs, etc.

[0145] As an embodiment, the network environment in which the account accesses the network will affect the account's preference for the target behavior, such as when the network speed is poor, the account prefers to use the accelerator; and the device information of the device in which the account accesses the accelerator can reflect the network environment to a certain extent, the network environment when the account accesses the accelerator; therefore, the network environment information of the account during the trial period includes the device information of the device used by the account to access the accelerator during the trial period, and the network environment in which the account accesses the accelerator during the trial period;

[0146] The device information may be the number of devices used by the account to access the accelerator during the trial period, which can reflect to a certain extent the stability of the account's access environment to the network. The stability of the account's access environment to the network will, to a certain extent, affect the account's preference for accelerating gaming applications through the accelerator.

[0147] The device information may also be the proportion of abnormal devices that access the accelerator during the trial period of the account, which is the ratio of the number of abnormal devices used by the account to access the accelerator to the total number of devices used by the account to access the accelerator. If the number of accounts that access the accelerator during a specified time period is greater than a threshold number of accounts, the device is the above-mentioned abnormal device. The above-mentioned specified time period may be a period of one month, one week, or the like.

[0148] The device information may also be the device environment corresponding to the device used by the account to access the accelerator during the trial period, and the device environment may include whether the device is a device in an Internet cafe, or whether the device is a device of a certain enterprise.

[0149] Furthermore, when the electronic resource is the electronic currency in the electronic wallet corresponding to the account, and the conversion operation is a payment operation for transferring electronic currency for the target behavior, the device type of the device can reflect the account's holding of electronic currency to a certain extent. Therefore, the above-mentioned device information can also be the device type of the device used by the account to access the accelerator during the trial period. If the device type is a device with a higher price, the account corresponding to the device type may hold more electronic currency, and thus the possibility of performing the conversion operation is also higher.

[0150] The network environment for the above-mentioned account to access the accelerator during the trial period may be the geographical location information when the account accesses the accelerator during the trial period; the geographical location information when the above-mentioned account accesses the accelerator may be determined by the network protocol address (IP address) of the device used when the account accesses the accelerator, and the geographical location corresponding to the above-mentioned network protocol address is determined as the geographical location information when the account accesses the accelerator, and the above-mentioned geographical location may be but is not limited to the corresponding longitude and latitude or administrative area, etc.

[0151] The above-mentioned geographic location information can reflect the geographic location of the device used by the account, and the geographic location of the device can reflect the geographic location of the user corresponding to the account to a certain extent; the electronic resource is the electronic currency in the electronic wallet corresponding to the account. When the conversion operation is a payment operation for transferring electronic currency for the target behavior, accounts in different geographic locations have different holdings of electronic currency. The holdings of electronic currency by the account will largely affect whether the account performs payment operations. For example, if the economic situation of the geographic location corresponding to the geographic location information when the account accesses the accelerator is good, the user corresponding to the account may be more empowered, and the account will have a stronger intention to perform payment operations for the target behavior.

[0152] The network environment in which the above account accesses the accelerator during the trial period may also be information about the network operator of the traffic used by the account, the network signal strength and signal stability of different network operators in different areas, and the network signal strength and signal stability will affect the account's network environment.

[0153] As an embodiment, the gaming applications accessed by an account will, to a certain extent, reflect the account's preference for gaming applications. Generally, the higher the account's preference for gaming applications, the stronger the account's willingness to accelerate gaming applications through an accelerator. Therefore, the above-mentioned behavioral characteristic data may also include game access information of the account accessing gaming applications during the trial period.

[0154] Since the number of times the account accesses game applications during the trial period, the number of game applications accessed, the duration of the game applications accessed, and the network speed requirements of the game applications accessed, etc., can to a certain extent reflect the account's willingness to accelerate the game applications through an accelerator. For example, if the account accesses game applications a large number of times, or the account accesses game applications for a long time, or the account accesses game applications with high network speed requirements, the account is willing to accelerate the game applications through an accelerator. Therefore, the above-mentioned game access information may be at least one of the number of times the game applications are accessed, the duration of the game applications accessed, the game types of the game applications accessed, and the number of the game applications accessed. For example, the game access information includes the total number of game applications accessed by the account during the trial period, the total duration of the game applications accessed, whether the game applications are accessed to foreign server services, and the total number of game applications accessed; the above-mentioned game types may include competitive games, virtual reality games, domestic games, foreign games, first-person shooter games, etc.

[0155] As an embodiment, there are currently multiple accelerators. The experience of an account using accelerators other than the above accelerators will also have a certain degree of impact on the account's willingness to use the above accelerators. Therefore, the above behavioral characteristic data may also include access information of the account to similar applications during the trial period, and the similar applications may be accelerators other than the above accelerators.

[0156] The following is a detailed description of the attribute feature data in the account feature data of the embodiment of the present application.

[0157] The third-party account may be an account of an instant messaging application, such as a QQ account of a QQ application. When the account is a third-party account, the attribute characteristic data may include at least one of the following data:

[0158] Account setting information of the third-party account, creation time information of the above third-party account, and account level information of the above third-party account.

[0159] Taking into account that the accounts corresponding to people of different ages hold different amounts of electronic resources, the accounts corresponding to males and females hold different amounts of electronic resources, and the accounts corresponding to users in different regions also hold different amounts of electronic resources. Therefore, in the embodiments of the present application, the estimated value of the electronic resource holdings of the third-party account is characterized by the age, gender, native place, etc. set by the user, that is, the above-mentioned account setting information may include the age, gender, native place, etc. set by the user through the third-party account; if the electronic resource is electronic currency, when the conversion operation is a payment operation, older people are relatively wealthy, and their corresponding accounts hold relatively more electronic currency, then the accounts corresponding to older people are more willing to perform payment operations in the form of transferring electronic currency for the target behavior.

[0160] In the creation time information of the above-mentioned third-party account, the longer the time when the third-party account was created, the higher the level of Internet access of the third-party account is generally and the more electronic resources it holds, and the willingness value of the third-party account to perform the above-mentioned conversion operation is higher. Therefore, in the embodiment of the present application, the creation time information of the third-party account is used to represent the estimated value of the electronic resources held by the third-party account.

[0161] The higher the account level in the account level information of the third-party account, the higher the third-party account's Internet access level is generally and the more electronic resources it holds, and the third-party account has a higher willingness value to perform conversion operations. Therefore, in the embodiment of the present application, the account level information of the third-party account is used to represent the estimated value of the electronic resources held by the third-party account.

[0162] Furthermore, the above-mentioned attribute characteristic data also includes at least one data for characterizing the network preference degree of the third-party account: group information of the above-mentioned third-party account in the application corresponding to the above-mentioned third-party account, account access information of the above-mentioned third-party account accessing the application corresponding to the above-mentioned third-party account during a specified time period; wherein the above-mentioned specified time period is not limited, for example, it can be set to one month or one week before the current moment, etc.

[0163] The following content of the embodiment of the present application provides a process of estimating the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period through the ensemble learning method in the supervised classification learning in the above step S202, which is specifically as follows:

[0164] First, before estimating the probability value of the operation intention of the account to be predicted to perform conversion operations in the non-trial period, an operation intention prediction model can be trained based on an ensemble learning method, using the account feature data of each historical account and the operation data of whether each historical account performed conversion operations in its respective non-trial period as training samples.

[0165] Furthermore, when estimating the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period, the trained operation intention prediction model is used, the extracted account feature data of the account to be predicted is input, and the operation intention probability value of the account to be predicted to perform the above-mentioned conversion operation in the non-trial period output by the above-mentioned operation intention prediction model is obtained.

[0166] Specifically, the trained operation intention prediction model includes multiple classifiers and a decision maker. When estimating the operation intention probability value of the account to be predicted to perform a conversion operation in the non-trial period, each classifier makes a judgment based on the account feature data of the account to be predicted and outputs a first classification result. The first classification result is used to indicate the correlation between the account feature data and the conversion operation result in the non-trial period. The decision maker obtains the operation intention probability value of the account to be predicted to perform the above-mentioned conversion operation in the non-trial period based on the first classification results output by each classifier.

[0167] Furthermore, the trained operation intention prediction model may be a machine learning model that uses an xgboost regression tree as a classifier, the first classification result may be a classification result of each xgboost regression tree, and the prediction principle of the operation intention prediction model that uses xgboost as a classifier is shown in Formula 1 below.

[0168] Formula 1:

[0169] In Formula 1, K represents the number of classifiers in the operation intention prediction model (i.e., the number of regression trees in the operation intention prediction model), wherein each classifier is automatically constructed by the operation intention prediction model based on at least one feature data in the account feature data; is the estimated probability value of the account's intention to convert during the non-trial period; k () is the prediction function of the kth regression tree (i.e., the kth classifier), which is used to map the training samples to the leaf nodes of the regression tree; x i is a feature vector composed of the account feature data in the historical data used as training samples.

[0170] The loss function representing the above operation intention prediction model using xgboost as a classifier is shown in Formula 2 below.

[0171] Formula 2:

[0172] In Formula 2, Obj is the optimization target of the operation intention prediction model, which is the error of the conversion operation result of the training sample predicted by the operation intention prediction model and the complexity penalty of the operation intention prediction model; n is the number of training samples; y i is the conversion operation result of each i-th training sample in the non-trial period; represents the output value of the i-th training sample of the operation intention prediction model; K represents the number of classifiers in the operation intention prediction model; f k() is the prediction function of the kth regression tree (i.e., the kth classifier), which is used to map the training samples to the leaf nodes of the regression tree; xi is the feature vector composed of the account feature data in the historical data used as the training samples, is the error of the operation intention prediction model in predicting the conversion operation result of the i-th training sample, Ω(f k ) is the regular term of the k-th regression tree (i.e., the k-th classifier), that is, the complexity penalty of the k-th regression tree in the operation intention prediction model.

[0173] The following content of the embodiment of the present application provides a training process of an operation intention prediction model, which mainly includes a sample processing process and a model training process, which are as follows:

[0174] (I) Sample processing

[0175] First, historical data is obtained from the target application that implements the target behavior as training samples.

[0176] Considering that the account feature data in the training samples include multiple discrete data, and the data types of the multiple discrete data may also be different, it is not convenient for the operation intention prediction model to learn the correlation between the account feature data and the conversion operation results in the non-trial period. Therefore, the account feature data can be set and processed before training the operation intention prediction model to convert the multiple discrete data in the account feature data into data with certain distribution characteristics and easy to process.

[0177] For example, the account feature data extracted from the training samples can be subjected to feature normalization processing and onehot encoding processing, and then the data processing rules used in the above normalization processing and the encoding rules used in the onehot encoding processing can be saved, so that the prediction server can obtain the final estimation result according to the output of the operation intention prediction model and the data processing rules and encoding rules; wherein the feature normalization processing can include, according to the data processing rules, converting each feature-representing data in the account feature data into a numerical value between 0 and 1; the onehot encoding processing is also called one-bit valid encoding, which uses an N-bit state register to encode N states, each state has its own independent register bit, and only one bit is valid at any time. In the embodiment of the present application, the onehot encoding processing can digitize the features of each discrete data in the account feature data according to the set encoding rules.

[0178] As an embodiment, taking into account the specific application scenarios of the operation intention prediction model, since some abnormal historical data or historical data containing specified features will affect the prediction accuracy of the operation intention prediction model, the acquired historical data can be subjected to at least one of the following data cleaning processes, and the historical data after the data cleaning process can be determined as training samples for training.

[0179] The first data cleaning process is to discard the historical data with conversion operations during the trial period.

[0180] Since the operation intention prediction model needs to learn the correlation between the account characteristic data of each historical account in the historical data and the conversion operation results of each historical account in the non-trial period, the conversion operations of the historical accounts in the trial period will affect the learning process of the operation intention prediction model. It is possible that the correlation between the account characteristic data of the historical accounts and the conversion operation results of the historical accounts in the trial period will affect the estimation results of the operation intention prediction model. Therefore, the historical data with conversion operations in the trial period can be discarded to improve the accuracy of the correlation between the account characteristic data of the historical accounts and the conversion operation results of the historical accounts in the non-trial period learned by the operation intention prediction model.

[0181] The second data cleaning process: During the non-trial period, there is historical data that performs the above conversion operations by setting resource conversion methods.

[0182] Taking into account the specific application scenarios of the operation intention prediction model, in some application scenarios, it may be necessary to predict the summary value of the operation intention of the account to perform conversion operations in the non-trial period without interference conditions. The above interference condition may be that the account has the authority to perform conversion operations by setting the resource conversion method. However, the historical data of the conversion operation by setting the resource conversion method, due to the correlation between the account feature data of the historical account learned by the operation intention prediction model and the conversion operation results of the historical account in the non-trial period, there are conversion operation results of the conversion operation by setting the resource conversion method in the non-trial period, which affects the estimation accuracy of the operation intention prediction model. ; If the conversion operation is a payment operation, and the resource conversion method is set to include the account using a discount coupon to perform the payment operation, the operation intention prediction model may learn the account feature data of the historical account and the correlation between the historical account's payment operations in the non-trial period through the coupon method. When the operation intention prediction model is applied to the application scenario in which the account is predicted to not use the preferential method to perform the payment operation in the non-trial period, the accuracy of the estimation will decrease. Therefore, in the embodiment of the present application, the historical data of the above-mentioned conversion operations performed by setting the resource conversion method in the non-trial period can be deleted to improve the estimation accuracy of the operation intention prediction model in specific application scenarios.

[0183] The third data cleaning process: discard historical data whose total duration of the target behavior is less than the duration threshold.

[0184] Specifically, considering that the proportion of historical accounts that performed conversion operations during the non-trial period in the training samples is relatively small, if the conversion operation was performed when the total duration of the target behavior in some historical data was too short, it may, to a certain extent, indicate that the historical accounts performed the conversion operation when the preference for the target behavior was low. If the other data in the account feature data of these historical accounts is not much different from the account feature data corresponding to other historical accounts, the operation intention prediction model may learn that the account performed the conversion operation when the preference for the target behavior was low. The estimation accuracy of the trained operation intention prediction model may decrease. Therefore, the historical data with a total duration of the target behavior that is less than the duration threshold can be discarded when obtaining the training samples.

[0185] The historical data obtained may have missing data problems, such as missing one or more data in the attribute feature data. In this case, the missing data in the historical data can be processed by setting a string, such as filling the missing numerical data with 0, filling the missing character data with an empty string, etc.

[0186] (II) Model training process

[0187] The task of training the operation intention prediction model in the embodiment of the present application is to find a set of parameter values ​​of the regression tree that minimizes the optimization target (Obj in the above formula 2).

[0188] When training the operation intention prediction model, the account feature data of the historical accounts in each training sample and the operation data of whether the historical accounts performed conversion operations in the non-trial period can be extracted; then, based on the operation data of whether the historical accounts in each training sample performed conversion operations in the non-trial period, the second classification result of each training sample is determined, and the above second classification result is used to indicate the operation result of whether each historical account performed the above conversion operation in its respective non-trial period; using the mapping relationship between the account feature data of the historical accounts in each training sample and the second classification result, the operation intention prediction model obtained by historical training or the initial operation intention prediction model is adjusted to obtain the trained operation intention prediction model.

[0189] Specifically, when using training samples to train the operation intention prediction model, the initial operation intention prediction model created by the ensemble learning method can be trained based on the initially obtained training samples; or after the trained operation intention prediction model is initially obtained, the operation intention prediction model can be updated according to the newly generated historical data in the target application, such as by updating the training samples according to the newly generated historical data in the target application, and using the account feature data of the historical accounts in the updated training samples and the conversion operation results of the historical accounts during the non-trial period to adjust the operation intention prediction model obtained by historical training to obtain a new trained operation prediction model.

[0190] In the process of training the operation intention prediction model, the embodiment of the present application determines the second classification result of the training sample (i.e., the label of the training sample) according to the conversion operation results of the historical accounts in the training sample during the non-trial period, sets the label of the training sample that has undergone the conversion operation in the conversion operation results as the positive sample label, and sets the label of the training sample that has not undergone the conversion operation in the conversion operation results as the negative sample label, thereby learning the mapping relationship between the account feature data in the training sample and the label of the training sample, adjusting the model parameters of the operation intention prediction model obtained by historical training or the initial operation intention prediction model, and obtaining the trained operation intention prediction model; wherein the above-mentioned adjusted model parameters may include but are not limited to the number of xgboost iterations, the weight of the positive sample label, and the maximum depth of the sub-regression tree in xgboost, etc., and when adjusting the model parameters, it may be possible but not limited to automatically searching for the optimal values ​​of the above-mentioned model parameters using a grid search method.

[0191] For ease of understanding, the following content in the embodiments of the present application provides a specific example of business processing.

[0192] In this example, the target application is an accelerator, the target behavior includes operations to accelerate a gaming application through the accelerator; the electronic resource is the electronic currency in the account's electronic wallet, and the conversion operation includes the account's payment operations for the trial accelerator in the form of transferring electronic currency so that the account can continue to accelerate the gaming application through the accelerator during the non-trial period; the account is a QQ account.

[0193] For the sake of convenience, this example sets the trial period to a period of 3 days from the date the account is connected to the accelerator, and the non-trial period is a period of 4-10 days from the date the account is connected to the accelerator. The trial period in this example is described as a 3-day trial period below; however, the 3-day trial period is only for the convenience of understanding and is not a limitation on the trial period in this example. Those skilled in the art may limit the period in this example according to actual needs.

[0194] See also Figure 3The schematic diagram of the estimated operation intention probability value first collects historical data triggered by historical accounts connected to the accelerator, and discards historical data that affects the estimation accuracy according to the first to third data cleaning processes mentioned above, and determines the historical data after the data cleaning process as training samples for training. The above historical data that affects the estimation accuracy include: historical data of payment operations for transferring electronic currency performed by historical accounts during the 3-day trial period, historical data of payment operations for transferring electronic currency performed in a preferential manner on the 4th to 10th days after the historical accounts were connected to the accelerator, and historical data of a total acceleration time of less than 30 minutes for game applications accelerated by the accelerator.

[0195] After obtaining the training samples, the feature engine is used to perform data processing on the training samples by normalization and onehot encoding. Then, the mapping relationship between the account feature data of each historical account in the training samples after data processing and the conversion operation results of each historical account in their respective non-trial periods is used to train the operation intention prediction model obtained by historical training or the initial operation intention prediction model to obtain a trained operation intention prediction model. The operation intention prediction model is a model that uses xgboost as a classifier. The specific details can be found in the above content and will not be repeated here.

[0196] After obtaining the trained operation intention prediction model, the data triggered after the account to be tested is connected to the accelerator is obtained, and the account feature data of the account to be predicted is extracted from the application data. The account feature data extracted by the feature engine includes the attribute feature data of the account to be predicted and the behavior feature data of the account to be predicted during the 3-day trial period; then, the account feature data of the account to be predicted is input into the trained operation intention prediction model to obtain the operation intention probability value of the account to be predicted performing the above-mentioned payment operation during the non-trial period output by the trained operation intention prediction model. The operation intention probability value can be a value between 0 and 1.

[0197] When performing business operations based on the estimation results, if the intention probability threshold of the account to be predicted is 0.6 and the estimated operation intention probability value of the account to be predicted is 0.45, an electronic currency coupon is sent to the account to be predicted so that the account to be predicted can perform payment operations for the accelerator in a preferential payment manner using the electronic currency coupon.

[0198] The account characteristic data in this example may specifically include the following multiple data items.

[0199] (1) Behavioral characteristic data in account characteristic data includes the following data:

[0200] 1. The target behavior operation information of the account using the accelerator to accelerate the game application during the trial period, including the following data:

[0201] (1) Information about changes in the network efficiency of game applications after an account uses an accelerator to accelerate game applications during the trial period, including:

[0202] Average network delay during the 3-day trial period: the average network delay of the network transmission channel created by the accelerator in the game application during the 3-day trial period is determined as the above-mentioned network efficiency change information.

[0203] Network latency ratio of day 2 to day 1: The ratio of the average network latency of the game app on the second day of the 3-day trial period to the average network latency of the game app on the first day of the 3-day trial period.

[0204] Network latency ratio of day 3 to day 2: The ratio of the average network latency of the game app on the third day of the 3-day trial period to the average network latency of the game app on the second day of the 3-day trial period.

[0205] (2) Information about the duration of time that the account uses an accelerator to accelerate the gaming application during the trial period, including:

[0206] Total acceleration duration during the 3-day trial period: The total duration of the account's acceleration operations for gaming applications through the accelerator during the 3-day trial period.

[0207] Total acceleration time during working hours: the total time the accelerator is used to accelerate the game application from 7:00 am to 6:00 pm during the 3-day trial period.

[0208] Total acceleration time during entertainment hours: the total time the account uses the accelerator to accelerate game applications from 18:00 to 23:00 during the 3-day trial period.

[0209] Total acceleration time during sleep period: The total time that the accelerator is used to accelerate the gaming application during the period from 23:00 to 24:00 and from 0:00 to 7:00 during the 3-day trial period.

[0210] Ratio of acceleration time on the second day to the first day: the ratio of the total acceleration time of the account on the second day of the 3-day trial period to the total acceleration time on the second day of the 3-day trial period.

[0211] Ratio of acceleration time on the 3rd day to the 2nd day: the ratio of the total acceleration time of the account on the 3rd day of the 3-day trial period to the total acceleration time on the 3rd day of the 3-day trial period.

[0212] (3) Acceleration success rate of an account during the trial period: the ratio of the number of successful accelerations of an account during the three-day trial period to the total number of accelerations.

[0213] 2. The account's first access information to the accelerator during the trial period, including:

[0214] (1) New entry date type: the date type of the specific time when the account first accesses the accelerator (whether it is a working day).

[0215] (2) New entry channel: The accelerator promotion link corresponding to the account’s first access to the accelerator, that is, the accelerator promotion link created by the platform through which the account accesses the accelerator.

[0216] (3) First access environment: Whether the device used by the account to access the accelerator for the first time is a device in an Internet cafe.

[0217] 3. Information about the network environment of the account during the trial period, including:

[0218] (1) Device information of the device used by the account to access the accelerator during the trial period, including:

[0219] Total number of devices: The total number of devices that the account uses to access the accelerator during the 3-day trial period.

[0220] Abnormal device ratio: the ratio of the number of abnormal devices used by an account to access the accelerator during the three-day trial period to the total number of devices used by an account to access the accelerator during the three-day trial period.

[0221] (2) The geographic region where the account accessed the accelerator the most during the three-day trial period: the province with the most access times among the provinces corresponding to the IP addresses of the devices used by the account to access the accelerator during the three-day free period.

[0222] (3) Information about the network operator that used the most traffic during the three-day trial period: Information about the network operator that used the most traffic during the three-day trial period.

[0223] 4. Game access information of game applications accessed by the account during the trial period, including:

[0224] (1) Total number of game applications connected: the total number of game applications connected to the account during the 3-day trial period.

[0225] (2) Total number of game applications accessed through foreign servers: the total number of game applications accessed by the account through foreign servers during the 3-day trial period.

[0226] 5. Access information of other accelerators: the total number or total duration of the account's access to other accelerators other than the above accelerators during the 3-day trial period.

[0227] (II) The attribute characteristic data in the account characteristic data includes the following data:

[0228] 1. Account setting information of QQ account, including:

[0229] Gender: The gender reported by the user to the QQ application through the QQ account, which can include male, female, unknown, etc.

[0230] Age: The age reported by the user to the QQ application through the QQ account.

[0231] 2. QQ account creation time information: The time when the QQ account was created, which can be expressed in days, months or years.

[0232] 3. Account level information of QQ account: a continuous indicator of QQ account at the data level.

[0233] 4. Group information of QQ account, including:

[0234] Number of participating groups: The total number of groups that the QQ account has joined in the corresponding application.

[0235] Number of self-created groups: The total number of groups historically created by a QQ account in the corresponding application.

[0236] 5. The degree of Internet usage of the QQ account, including:

[0237] Total monthly login time: the length of time the QQ account remains online in the month to which the current time belongs, in seconds.

[0238] Total number of days logged in per month: The total number of days the QQ account remains online in the current month.

[0239] In the implementation example of the present application, by utilizing the account characteristic data of the historical account and the correlation between the conversion operation results of whether the conversion operation was performed in the non-trial period, the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period can be automatically estimated based on the account characteristic data of the account to be predicted, and business processing can be performed on the account to be predicted based on the estimated operation intention probability value. In this process, the efficiency of estimating the account's conversion operation for the target behavior in the non-trial period is improved, and the estimation process does not need to rely on the experience of personnel, which reduces the situation where the estimation results are inaccurate due to insufficient experience of personnel, thereby improving the accuracy of the estimated operation intention probability value.

[0240] Please refer to Figure 4 Based on the same inventive concept, the embodiment of the present application provides a service processing device 400, including:

[0241] The feature extraction unit 401 is used to obtain account feature data of the account to be predicted, wherein the account feature data includes attribute feature data of the account to be predicted and behavior feature data of the account to be predicted in the trial period of the target behavior, wherein the attribute feature data is used to characterize the estimated value of the electronic resource holding amount of the account to be predicted, and the behavior feature data characterizes the preference of the account to be predicted for the target behavior, wherein the target behavior includes an operation that can be used after the conversion operation of transferring the electronic resource is performed in the non-trial period, and the non-trial period is the period after the end of the trial period;

[0242] The operation intention prediction unit 402 is used to estimate the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period based on the account characteristic data of the account to be predicted, using the account characteristic data of each historical account and the correlation degree of the conversion operation result of each historical account in the non-trial period in the historical data, wherein the conversion operation result includes the result of whether to perform the conversion operation, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in the respective trial period;

[0243] The business processing unit 403 is used to perform business processing on the account to be predicted according to the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period.

[0244] As an embodiment, the operation intention prediction unit 402 is specifically used for:

[0245] The trained operation intention prediction model is used, the account feature data of the account to be predicted is input, and the operation intention probability value of the account to be predicted performing the conversion operation in the non-trial period output by the operation intention prediction model is obtained. The operation intention prediction model is based on an ensemble learning method, and is trained by using the account feature data of each historical account and the operation data of whether each historical account performs the conversion operation in its respective non-trial period as training samples. The trained operation intention prediction model includes multiple classifiers and a decision maker, wherein:

[0246] Each classifier makes a judgment based on the account feature data of the account to be predicted and outputs a first classification result, wherein the first classification result is used to indicate the correlation between the account feature data and the conversion operation result in the non-trial period;

[0247] The decision maker obtains the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period according to the first classification results output by each classifier.

[0248] As an embodiment, the operation intention prediction unit 402 is further configured to obtain the above-mentioned trained operation intention prediction model in the following manner:

[0249] Extract the account feature data of the historical accounts in each training sample and the operation data of whether the historical accounts have been converted during the non-trial period;

[0250] Determine the second classification result of each training sample according to the operation data of whether the historical account in each training sample performs the above conversion operation in the non-trial period, wherein the second classification result is used to indicate the operation result of whether each historical account performs the above conversion operation in the respective non-trial period;

[0251] By using the mapping relationship between the account feature data of the historical accounts in each training sample and the second classification result, the operation intention prediction model obtained by historical training or the initial operation intention prediction model is adjusted to obtain the above-mentioned trained operation intention prediction model.

[0252] As an embodiment, the operation intention prediction unit 402 is further configured to perform at least one of the following data processing on the historical data, and determine the historical data after the data processing as the training sample:

[0253] Discard the historical data of the above conversion operations during the trial period;

[0254] Discard the historical data of the above conversion operation performed by setting the resource conversion method during the non-trial period;

[0255] Discard historical data whose total duration of the target behavior is less than the duration threshold.

[0256] As an embodiment, the above-mentioned behavior characteristic data includes at least one of the following data:

[0257] Target behavior operation information of the account performing the above-mentioned target behaviors during the trial period;

[0258] Target application access information of the target application that the account accesses to achieve the target behavior during the trial period;

[0259] The account's network environment information during the trial period;

[0260] The account's network usage during the trial period.

[0261] As an embodiment, the target application for realizing the target behavior includes an accelerator, the target behavior includes an operation of accelerating the game application through the accelerator, and the behavior characteristic data further includes at least one of the following data:

[0262] Game access information of the account accessing the above-mentioned game applications during the trial period;

[0263] The account's access information for similar applications during the trial period, where the similar applications include the account's access to other accelerators other than the above accelerator.

[0264] As an embodiment, the target behavior operation information includes at least one of the following information:

[0265] Information about changes in the network efficiency of the game application after the account uses the accelerator to accelerate the game application during the trial period;

[0266] The duration of acceleration of the game application by the account through the accelerator during the trial period;

[0267] The success rate of the account accelerating the above-mentioned game application through the above-mentioned accelerator during the trial period.

[0268] As an embodiment, after the account accelerates the game application through the accelerator during the trial period, the network efficiency change information of the game application includes at least one of the following information:

[0269] The average network delay of the game application after the account accelerates the game application through the accelerator during the trial period, wherein the average network delay includes the average network delay of the network transmission channel created by the accelerator in the game application;

[0270] The ratio of the second average network delay of the game application in the first sub-period of the account to the second average network delay of the game application in the second sub-period, wherein the above average network delay includes the average value of the network delay of the network transmission channel created by the above accelerator in the above game application, the above first sub-period is a sub-period of the above trial period, the above second sub-period is a sub-period of the above trial period, and the above first sub-period is the adjacent period after the end of the above second sub-period.

[0271] As an embodiment, the acceleration duration information of the account using the accelerator to accelerate the game application during the trial period includes at least the following information:

[0272] The total duration of acceleration of the above-mentioned game application by the account through the above-mentioned accelerator during the trial period;

[0273] The total duration of the account's acceleration in the third sub-period, where the third sub-period is a sub-period of the trial period;

[0274] The ratio of the total acceleration duration of the account in the fourth sub-period to the total acceleration duration in the fifth sub-period, wherein the fourth sub-period is a sub-period of the trial period, the fifth sub-period is a sub-period of the trial period, and the fourth sub-period is the adjacent period after the end of the fifth sub-period.

[0275] As an embodiment, the account includes a third-party account for logging into a target application for implementing the target behavior, and the attribute characteristic data includes at least one of the following data:

[0276] Account setup information for the third-party accounts listed above;

[0277] Creation time information of the above third-party accounts;

[0278] Account level information for the above third-party accounts.

[0279] As an embodiment, the attribute characteristic data further includes at least one of the following data used to characterize the network preference degree of the third-party account:

[0280] Group information of the third-party account in the application corresponding to the third-party account;

[0281] The account access information of the third-party account accessing the application corresponding to the third-party account during the specified time period.

[0282] As an embodiment, the service processing unit 403 is specifically configured to:

[0283] If the operation intention probability value of the above-mentioned account to be predicted to perform the conversion operation in the non-trial period is less than the intention probability threshold, the setting electronic media information is sent to the above-mentioned account to be predicted, so that the above-mentioned account to be predicted performs the above-mentioned conversion operation by setting the resource conversion method.

[0284] As an example, Figure 4 The device in can be used to implement any of the business processing methods discussed above.

[0285] The service processing device 400 is an example of a hardware entity. Figure 5 The computer device shown includes a processor 501 , a storage medium 502 and at least one external communication interface 503 ; the processor 501 , the storage medium 502 and the external communication interface 503 are all connected via a bus 504 .

[0286] The storage medium 502 stores a computer program;

[0287] When the processor 501 executes the computer program, the navigation method of the mobile device discussed above is implemented.

[0288] Figure 5 In the figure, one processor 501 is taken as an example, but the number of processors 501 is not limited in practice.

[0289] The storage medium 502 may be a volatile memory, such as a random-access memory (RAM); the storage medium 502 may also be a non-volatile memory, such as a read-only storage medium, a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), or the storage medium 502 may be any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The storage medium 502 may be a combination of the above storage media.

[0290] Based on the same technical concept, an embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions. When the above-mentioned computer instructions are executed on a computer, the computer executes the business processing method discussed above.

[0291] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0292] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A business processing method, characterized in that: include: Acquire account characteristic data of the account to be predicted, the account characteristic data including attribute characteristic data of the account to be predicted and behavior characteristic data of the account to be predicted in the trial period of the target behavior, the attribute characteristic data is used to characterize the estimated value of the holding amount of the electronic resources of the account to be predicted, the behavior characteristic data characterizes the preference degree of the account to be predicted for the target behavior, the target behavior includes an operation that can be used after the conversion operation of transferring the electronic resources is performed in the non-trial period, and the non-trial period is the period after the end of the trial period; By using the correlation between the account characteristic data of each historical account and the conversion operation results of each historical account in the respective non-trial period in the historical data, based on the account characteristic data of the account to be predicted, the probability value of the operation intention of the account to be predicted to perform the conversion operation in the non-trial period is estimated, wherein the conversion operation result includes the result of whether the conversion operation is performed, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in the respective trial period; According to the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period, business processing is performed on the account to be predicted; The behavior characteristic data includes at least one of the following data: target behavior operation information of the account performing the target behavior during the trial period; target application access information of the account accessing the target application to achieve the target behavior during the trial period; network environment information of the account during the trial period; network usage level of the account during the trial period; The account includes a third-party account for logging into a target application for implementing the target behavior, and the attribute characteristic data includes at least one of the following data: account setting information of the third-party account; creation time information of the third-party account; account level information of the third-party account; The business processing for the account to be predicted based on the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period includes: if the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period is less than the intention probability threshold, sending the setting electronic media information to the account to be predicted, so that the account to be predicted performs the conversion operation by setting the resource conversion method.

2. The method according to claim 1, characterized in that The use of the historical data includes the account characteristic data of each historical account and the correlation between the conversion operation results of each historical account in the respective non-trial period, and estimating the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period based on the account characteristic data of the account to be predicted, specifically including: The trained operation intention prediction model is used, the account feature data of the account to be predicted is input, and the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period output by the operation intention prediction model is obtained. The operation intention prediction model is based on an ensemble learning method, and is trained by using the account feature data of each historical account and the operation data of whether each historical account performs the conversion operation in its respective non-trial period as training samples. The trained operation intention prediction model includes multiple classifiers and a decision maker, wherein: Each classifier makes a judgment based on the account feature data of the account to be predicted and outputs a first classification result, wherein the first classification result is used to indicate the degree of association between the account feature data and the conversion operation result in the non-trial period; The decision maker obtains the operation intention probability value of the account to be predicted to perform the conversion operation in a non-trial period according to the first classification results output by each classifier.

3. The method according to claim 2, characterized in that The trained operation intention prediction model is obtained by: Extract the account feature data of the historical accounts in each training sample and the operation data of whether the historical accounts have performed conversion operations in the non-trial period; Determine a second classification result of each training sample according to the operation data of whether the historical account in each training sample performs the conversion operation in the non-trial period, wherein the second classification result is used to indicate the operation result of whether each historical account performs the conversion operation in the respective non-trial period; The operation intention prediction model obtained through historical training or the initial operation intention prediction model is adjusted by using the mapping relationship between the account feature data of the historical accounts in each training sample and the second classification result to obtain the trained operation intention prediction model.

4. The method according to claim 2, characterized in that Also includes: Perform at least one of the following data processing on the historical data, and determine the historical data after the data processing as the training sample: Discarding historical data of the conversion operation during the trial period; Discarding historical data of the conversion operation performed by setting the resource conversion mode during the non-trial period; Discard historical data whose total duration of the target behavior is less than the duration threshold.

5. A service processing device, characterized in that: include: A feature extraction unit, used to obtain account feature data of the account to be predicted, wherein the account feature data includes attribute feature data of the account to be predicted and behavior feature data of the account to be predicted in a trial period of a target behavior, wherein the attribute feature data is used to characterize an estimated value of the electronic resource holding amount of the account to be predicted, and the behavior feature data characterizes a preference degree of the account to be predicted for the target behavior, wherein the target behavior includes an operation that can be used after a conversion operation of transferring electronic resources is performed in a non-trial period, and the non-trial period is a period after the end of the trial period; An operation intention prediction unit is used to estimate the operation intention probability value of the account to be predicted to perform a conversion operation in the non-trial period based on the account characteristic data of the account to be predicted, using the account characteristic data of each historical account in the historical data and the correlation degree of the conversion operation results of each historical account in the respective non-trial period, wherein the conversion operation result includes the result of whether the conversion operation is performed, and the account characteristic data of each historical account includes the attribute characteristic data of the historical account and the behavior characteristic data of the historical account in the respective trial period; A business processing unit, configured to perform business processing on the account to be predicted according to the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period; The behavior characteristic data includes at least one of the following data: target behavior operation information of the account performing the target behavior during the trial period; target application access information of the account accessing the target application to achieve the target behavior during the trial period; network environment information of the account during the trial period; network usage level of the account during the trial period; The account includes a third-party account for logging into a target application for implementing the target behavior, and the attribute characteristic data includes at least one of the following data: account setting information of the third-party account; creation time information of the third-party account; account level information of the third-party account; The business processing unit performs business processing on the account to be predicted according to the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period, and is specifically used to: if the operation intention probability value of the account to be predicted to perform the conversion operation in the non-trial period is less than the intention probability threshold, send setting electronic media information to the account to be predicted, so that the account to be predicted performs the conversion operation by setting the resource conversion method.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on a computer, the computer is caused to execute the method according to any one of claims 1 to 4.

Images