A user payment behavior prediction method based on trend, reason and confidence

Abstract

The embodiment of the application provides a user payment behavior prediction method based on a trend, a reason and confidence, and can solve the technical problem of large prediction error of an existing game user payment behavior prediction method. The user payment behavior prediction method based on the trend, the reason and the confidence comprises the following steps: dividing user data into payment behavior data and non-payment behavior data, and respectively training a first prediction model and a second prediction model. The first prediction model predicts future payment behavior of a user by learning a past payment behavior trend. The second prediction model learns a mapping relationship between non-payment behavior and user payment behavior from the perspective of a reason. The two prediction models predict payment behavior from different perspectives, and the result is more accurate. In addition, the application designs and combines model confidence to predict user payment behavior, which considers the influence of the reliability of the prediction model on payment behavior.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for predicting user payment behavior based on trends, causes, and confidence levels. Background Technology

[0002] Game user payment behavior prediction involves understanding user payment behavior to predict whether users will make payments or purchase a particular product or service. This prediction helps game platforms make more informed business decisions and strategic plans, such as improving game pricing strategies, game product features, and enhancing user experience and retention rates.

[0003] Currently, the analysis of game users' behavioral data and other relevant characteristics involves manually screening and extracting payment-related behavioral features.

[0004] However, manual analysis is greatly influenced by the analyst's subjectivity and cannot adaptively control the degree of influence of different features on payment behavior, which can easily lead to unsatisfactory prediction results. Summary of the Invention

[0005] This application provides a method for predicting user payment behavior based on trends, causes, and confidence levels, which can solve the technical problem of large prediction errors in existing game user payment behavior prediction methods.

[0006] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:

[0007] In a first aspect, embodiments of this application provide a method for predicting user payment behavior based on trends, causes, and confidence levels. This method includes: dividing user data into first data and second data; the first data consists of payment behavior data for each user, including daily payment data for each user, and the payment data includes time characteristics; the second data consists of non-paying behavior data for each user, including daily non-paying data for each user, and the non-paying data includes multiple non-paying features; encoding the first data based on time characteristics to obtain first encoded data; periodically traversing the first encoded data to construct a first dataset; training a first prediction model based on a first loss function and the first dataset; encoding the second data based on non-paying features to obtain second encoded data; periodically traversing the second encoded data to construct a second dataset; training a second prediction model based on a second loss function and the second dataset; and identifying a first set of paying users, a second set of paying users, and a third set of paying users based on historical payment rates. User set; Based on the first dataset, establish a first evaluation dataset, a second evaluation dataset, and a third evaluation dataset corresponding to the first, second, and third paid user sets, respectively; Based on the second dataset, establish a fourth evaluation dataset, a fifth evaluation dataset, and a sixth evaluation dataset corresponding to the first, second, and third paid user sets, respectively; According to the first prediction model, obtain a first confidence level based on the first evaluation dataset, a second confidence level based on the second evaluation dataset, and a third confidence level based on the third evaluation dataset; According to the second prediction model, obtain a fourth confidence level based on the fourth evaluation dataset, a fifth confidence level based on the fifth evaluation dataset, and a sixth confidence level based on the sixth evaluation dataset; Obtain the paid prediction result based on the first combination probability, the second combination probability, or the third combination probability; The first combination probability includes the first confidence level and the fourth confidence level; The second combination probability includes the second confidence level and the fifth confidence level; The third combination probability includes the third confidence level and the sixth confidence level.

[0008] Based on the above description of the user payment behavior prediction method based on trend, cause, and confidence provided in the embodiments of this application, it can be seen that this method includes dividing user data into payment behavior data and non-payment behavior data, and training a first prediction model and a second prediction model respectively. The first prediction model predicts future user payment behavior by learning past payment behavior trends. The second prediction model, from the perspective of cause, learns the mapping relationship between non-payment behavior and user payment behavior. The two prediction models predict payment behavior from different perspectives, resulting in more accurate results.

[0009] In addition, this application designs and combines model confidence to predict user payment behavior, which takes into account the impact of the reliability of the prediction model itself on payment behavior.

[0010] In the feasible implementation of the first aspect, the first coded data includes a first coded element, a second coded element, a third coded element, a fourth coded element, a fifth coded element, a sixth coded element, a seventh coded element, an eighth coded element, and a ninth coded element. With a seven-day cycle, when performing the step of periodically traversing the first coded data to construct the first dataset, the user payment behavior prediction method based on trends, causes, and confidence further includes: using the first coded element, the second coded element, the third coded element, the fourth coded element, the fifth coded element, the sixth coded element, and the seventh coded element as the first input data of the first dataset; using the non-coded data corresponding to the eighth coded element as the first data label of the first input data; using the second coded element, the third coded element, the fourth coded element, the fifth coded element, the sixth coded element, the seventh coded element, and the eighth coded element as the second input data of the first dataset; and using the non-coded data corresponding to the ninth coded element as the second data label of the first input data.

[0011] In the feasible implementation of the first aspect, the first loss function is the cross-entropy loss function, and the first prediction model is based on the Transformer model.

[0012] In the feasible implementation of the first aspect, the second coded data includes the first non-paid data; the first non-paid data includes a first non-paid feature, a second non-paid feature, a third non-paid feature, a fourth non-paid feature, a fifth non-paid feature, and a sixth non-paid feature; when performing the step of encoding the second data based on the non-paid features to obtain the second coded data, the user payment behavior prediction method based on trends, causes, and confidence further includes: obtaining a first feature value based on the first non-paid feature through a first preset condition; the first feature value includes a first upper limit value; obtaining a second feature value based on the second non-paid feature through a second preset condition; the second feature value includes a second upper limit value; Based on the third non-paid feature, a third feature value is obtained through a third preset condition; the third feature value includes a third upper limit value. Based on the fourth non-paid feature, a fourth feature value is obtained through a fourth preset condition; the fourth feature value includes a fourth upper limit value. Based on the fifth non-paid feature, a fifth feature value is obtained through a fifth preset condition; the fifth feature value includes a fifth upper limit value. Based on the sixth non-paid feature, a sixth feature value is obtained through a sixth preset condition; the sixth feature value includes a sixth upper limit value. Wherein, the first upper limit value is less than the second upper limit value, the second upper limit value is less than the third upper limit value, the third upper limit value is less than the fourth upper limit value, the fourth upper limit value is less than the fifth upper limit value, and the fifth upper limit value is less than the sixth upper limit value.

[0013] In a feasible implementation of the first aspect, when performing the step of obtaining the payment prediction result based on the first combination probability, the second combination probability, or the third combination probability, the user payment behavior prediction method based on trend, cause, and confidence further includes: identifying the first historical payment behavior and historical non-paying behavior of the user to be predicted; inputting the first historical payment behavior into the first prediction model to obtain the first predicted payment rate; inputting the historical non-paying behavior into the second prediction model to obtain the second predicted payment rate; identifying the second historical payment behavior of the user to be predicted; obtaining the payment rate to be tested based on the second historical payment behavior; identifying the target payment level based on the payment rate to be tested; the target payment level is the first set of paying users, the second set of paying users, or the third set of paying users; if the target payment level is the first set of paying users, then obtaining the payment prediction result based on the first combination probability.

[0014] In the feasible implementation of the first aspect, before performing the step of dividing user data into first data and second data, the user payment behavior prediction method based on trends, causes and confidence also includes: obtaining historical behavior log data of users in the game platform, and cleaning and preprocessing the historical behavior log data to obtain user data.

[0015] In the feasible implementation of the first aspect, the second loss function is the cross-entropy loss function, and the second prediction model is based on the Transformer model.

[0016] In the feasible implementation of the first aspect, the encoding formula for calculating the first encoded data includes:

[0017] ;

[0018] in, Represented as the first The day of the week corresponding to each day ranges from 1 to 7; Represented as the first The number of users in the past Is there any payment involved?

[0019] Secondly, embodiments of this application provide a user payment behavior prediction system based on trends, causes, and confidence levels. The user payment behavior prediction system based on trends, causes, and confidence levels includes: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method provided in the first aspect.

[0020] The user payment behavior prediction system based on trends, causes, and confidence levels divides user data into payment behavior data and non-paying behavior data by implementing the methods provided in the first aspect, and then trains a first prediction model and a second prediction model respectively. The first prediction model predicts future user payment behavior by learning past payment behavior trends. The second prediction model, from a causal perspective, learns the mapping relationship between non-paying behavior and user payment behavior. The two prediction models predict payment behavior from different perspectives, resulting in more accurate results.

[0021] Thirdly, embodiments of this application provide a computer-readable medium having computer program instructions stored thereon, which can be executed by a processor to implement the method provided in the first aspect.

[0022] Computer program instructions in a computer-readable medium, by implementing the method provided in the first aspect, divide user data into paid behavior data and non-paid behavior data, and train a first prediction model and a second prediction model respectively. The first prediction model predicts future user paid behavior by learning past paid behavior trends. The second prediction model, from a causal perspective, learns the mapping relationship between non-paid behavior and user paid behavior. The two prediction models predict paid behavior from different perspectives, resulting in more accurate predictions. Attached Figure Description

[0023] Figure 1 A schematic diagram of a user payment behavior prediction system based on trend, cause, and confidence level provided in an embodiment of this application;

[0024] Figure 2 This is a flowchart illustrating a user payment behavior prediction method based on trends, causes, and confidence levels, provided in an embodiment of this application. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings. In the description of the embodiments of the present invention, unless otherwise stated, "multiple" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0026] Furthermore, to facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with substantially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" are not necessarily different. Meanwhile, in the embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is being used as an example, illustration, or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0027] The principles and features of this application are described below. The examples given are only for explaining this application and are not intended to limit the scope of this application.

[0028] This application provides a user payment behavior prediction system based on trends, causes, and confidence levels, which can execute the user payment behavior prediction method based on trends, causes, and confidence levels provided in this application. Figure 1 This is a schematic diagram of the structure of a user payment behavior prediction system based on trend, cause, and confidence level, provided in an embodiment of this application.

[0029] like Figure 1 As shown, the user payment behavior prediction system 001 based on trend, cause, and confidence includes at least one processor 011 and a memory 012 communicatively connected to the at least one processor; wherein, the memory 012 stores instructions that can be executed by the at least one processor 011, and the instructions are executed by the at least one processor 011 to enable the at least one processor 011 to execute the user payment behavior prediction method based on trend, cause, and confidence provided in the embodiments of this application.

[0030] Figure 2 This is a flowchart illustrating a user payment behavior prediction method based on trend, cause, and confidence level, provided as an embodiment of this application. Figure 2 As shown, in some embodiments, the user payment behavior prediction method based on trends, causes, and confidence levels includes the following steps:

[0031] S1 divides user data into first data and second data.

[0032] The first data consists of each user's payment behavior data, including daily payment data for each user, which incorporates time-related characteristics. For example, payment behavior data from the user data is used as the first data. , , where subscript Represented as the total number of users. Represented as the first Data on a user's paid behavior over the past month. , where subscript This represents the total number of days in the past month. Represented as the first The number of users in the past Is there a payment transaction on the first day? The number of users in the past If there is a paid behavior, then The value of is 1. If the th The number of users in the past If there is no payment activity on that day, then The value is 0. This shows that the payment behavior is associated with user information (the first...). (number of users) and time information (number of users) (Days), meaning that paid data includes time characteristics.

[0033] The second data consists of non-paid behavior data for each user, including daily non-paid data for each user, which comprises multiple non-paid features. In one implementation, the non-paid behavior data from the user data is used as the second data. , , where subscript Represented as the total number of users. Represented as the first Data on a user's non-paid behavior over the past month. , where subscript This represents the total number of days in the past month. Represented as the first The number of users in the past Non-paid data for days. It can include multiple non-paid features. In one example, It can include six non-paid features. ,in, Represented as the first The user's past number The number of visits to the online store per day Represented as the first The user's past number The number of times a day visits products in the online store. Represented as the first The user's past number Number of operations on the recharge interface Represented as the first The user's past number The time spent in the mall by the day, Represented as the first The user's past number Total number of games played per day Represented as the first The user's past number Number of logins per day.

[0034] In this embodiment, user historical behavior is divided into paid behavior data and non-paid behavior data. Historical paid behavior represents a user's past payment trends. A trend-based first prediction model is constructed, incorporating time-encoded features to facilitate the discovery of periodic correlations between user paid behavior and time. Non-paid behavior data consists of behaviors related to user payments. These behaviors are often the triggers for user payments, such as the effective dwell time on the online store, the number of clicks on products, and the frequency of visits. These can be used to predict the probability of user payments. Therefore, a cause-based second prediction model is constructed to help uncover the correlation between user non-paid and paid behaviors, predicting user payments from multiple dimensions with greater accuracy.

[0035] User data is divided into paid behavior data and non-paid behavior data in preparation for training the first and second prediction models respectively.

[0036] In some embodiments, before performing step S1, the user payment behavior prediction method based on trends, causes, and confidence levels further includes the following steps:

[0037] S100 acquires historical behavior log data of users on the game platform, cleans and preprocesses the historical behavior log data to obtain user data.

[0038] In some embodiments, the data cleaning method includes deleting user data for users who have been registered for less than one month. The cleaning method may also include deleting user data for users who have not logged in since registration or who log in less than three times per month. The cleaning method may further include, for users registered for more than one year, extracting data from the most recent month. For users registered for less than one month, all user data is retained.

[0039] S2, based on time characteristics, encodes the first data to obtain the first encoded data.

[0040] In some embodiments, the encoding formula for calculating the first encoded data includes:

[0041] ;

[0042] in, Represented as the first The day of the week corresponding to each day ranges from 1 to 7. Represented as the first The number of users in the past Is there any payment involved?

[0043] For example, the first If a user made a payment on the first day of the past month, and that first day was a Wednesday, then... The value is 3. The value is 1. The corresponding first encoded data .

[0044] Using the above method for traversal, in one example, the first... Composed of coded data on each user's paid behavior over the past month, based on their introduction time information. ,but .in, The value is 30.

[0045] Continuing to iterate using the above method, we obtain the coded payment behavior data of all users, which is then the first coded data. .

[0046] S3, periodically traverse the first encoded data to construct the first dataset.

[0047] In some embodiments, when multiple users are included, when performing step S3, the users are sequentially selected from... arrive Traversal The Middle User payment behavior coding data The first dataset is obtained. In one implementation, the process iterates through the data and selects the first... Seven consecutive days of user payment behavior coding data This is the input data for the first dataset. The label used as the input data is non-coded payment behavior data, with a value of 0 or 1.

[0048] For example, the first encoded data Including the first encoded element Second encoding element Third encoding element Fourth encoding element Fifth coding element The sixth encoding element The seventh coding element Eighth coding element and the ninth encoded element The user payment behavior prediction method based on trends, causes, and confidence levels also includes the following steps when performing the step of periodically traversing the first encoded data and constructing the first dataset, with a seven-day cycle:

[0049] S31, the first encoding element, the second encoding element, the third encoding element, the fourth encoding element, the fifth encoding element, the sixth encoding element and the seventh encoding element are the first input data of the first dataset.

[0050] First encoding element Second encoding element Third encoding element Fourth encoding element Fifth coding element The sixth encoding element The seventh coding element , as the first input data.

[0051] S32, using the non-encoded data corresponding to the eighth encoded element as the first data label of the first input data.

[0052] The value of the non-coded data corresponding to the eighth coded element is 0 or 1.

[0053] S33, using the second, third, fourth, fifth, sixth, seventh, and eighth encoded elements as the second input data of the first dataset.

[0054] Second encoding element Third encoding element Fourth encoding element Fifth coding element The sixth encoding element The seventh coding element Eighth coding element , as the second input data.

[0055] S34, using the non-encoded data corresponding to the ninth encoded element as the second data label of the first input data.

[0056] The value of the non-coded data corresponding to the ninth coded element is 0 or 1.

[0057] It is understood that steps S31 to S34 only provide two illustrative examples of periodic traversal operations and do not constitute a limitation on the number of periodic traversal operations. This application does not limit the number of periodic traversal operations. Furthermore, the periodic traversal operations can be found in steps S31 to S34.

[0058] S4, based on the first loss function and the first dataset, train the first prediction model.

[0059] In some embodiments, the first loss function is the cross-entropy loss function, and the first prediction model is... Based on the Transformer model.

[0060] In one implementation, the first loss function The calculation formulas include:

[0061] ;

[0062] in, Represented as from the first dataset The number of data samples in the middle; Represented as sampling The first of the data The data, if the first The data is the first Individual user's payment behavior data; then ; Indicated as Input the first prediction model The output results obtained; yes The corresponding real tags.

[0063] Based on the above dataset and loss function Train the first prediction model until the loss function value no longer decreases, thus obtaining the optimized first prediction model. .

[0064] In this way, the first prediction model can predict users' future payment behavior by learning from past payment behavior trends.

[0065] S5, based on the non-paid feature, encodes the second data to obtain the second encoded data.

[0066] Non-paying characteristics include behavioral data that can influence user payment. For example, non-paying characteristics may include the number of times a user visits the online store, the number of times they access products within the store, the number of times they interact with the recharge interface, and the effective dwell time in the store (effective dwell time refers to the time spent engaging in other user actions during that dwell time). In one example, non-paying data... It can include six non-paid features. ,in, Represented as the first non-paid feature, the second The user's past number The number of visits to the online store per day; This is represented as the second non-paid feature, the... The user's past number Number of times a day visits products in the online store; This is represented as the third non-paid feature. The user's past number Number of operations allowed on the daily recharge interface; Represented as the fourth non-paid feature, the... The user's past number The time spent in the mall by the day; This is represented as the fifth non-paid feature, the... The user's past number Total number of games played per day; Represented as the sixth non-paid feature, the... The user's past number Number of logins per day.

[0067] When performing step S5, the user payment behavior prediction method based on trends, causes, and confidence levels also includes the following steps:

[0068] S51, based on the first non-paid feature, the first feature value is obtained through the first preset conditions.

[0069] The first eigenvalue includes the first upper limit value.

[0070] In some embodiments, the first preset condition includes a calculation formula. The first eigenvalue is obtained. The calculated value, and, based on the first upper limit value and the first eigenvalue. The calculated value confirms the first eigenvalue. The final result.

[0071] In one implementation, the first eigenvalue The value range is 0-100, with a first upper limit of 100. That is to say, at the first eigenvalue... When the calculated value is greater than 100, the first eigenvalue The final result is 100. For example, the first non-paid feature. The first eigenvalue is 110. The calculated value is 110, then the first eigenvalue The final result is 100.

[0072] S52, based on the second non-paid feature, obtain the second feature value through the second preset conditions.

[0073] The second eigenvalue includes the second upper limit value.

[0074] In some embodiments, the second preset condition includes a calculation formula. The second eigenvalue is obtained. The calculated value, and, based on the second upper limit value and the second eigenvalue. The calculated value confirms the second eigenvalue. The final result.

[0075] In one implementation, the second eigenvalue The value range is 100-200, with a second upper limit of 200. That is to say, the second eigenvalue... When the calculated value is greater than 200, the second eigenvalue The final result is 200. For example, the second non-paid feature. The value is 110, the second eigenvalue. The calculated value is 210, then the second eigenvalue The final result is 200.

[0076] S53, based on the third non-paid feature, obtains the third feature value through the third preset conditions.

[0077] The third eigenvalue includes the third upper limit value.

[0078] In some embodiments, the third preset condition includes a calculation formula. The third eigenvalue is obtained. The calculated value, and, based on the third upper limit and the third eigenvalue. The calculated value confirms the third eigenvalue. The final result.

[0079] In one implementation, the third eigenvalue The value range is 200-400, with the third upper limit being 400. That is to say, within the third eigenvalue... When the calculated value is greater than 400, the third eigenvalue The final result is 400. For example, the third non-paid feature. The value is 210, the third eigenvalue. The calculated value is 410, then the third eigenvalue The final result is 400.

[0080] S54, based on the fourth non-paid feature, obtains the fourth feature value through the fourth preset condition.

[0081] The fourth eigenvalue includes the fourth upper limit value.

[0082] In some embodiments, the fourth preset condition includes a calculation formula. The fourth eigenvalue is obtained. The calculated value, and, based on the fourth upper limit and the fourth eigenvalue. The calculated value confirms the fourth eigenvalue. The final result.

[0083] In one implementation, the fourth eigenvalue The value range is 400-500, with the fourth upper limit being 500. That is to say, within the fourth eigenvalue... When the calculated value is greater than 500, the fourth eigenvalue The final result is 500. For example, the fourth non-paid feature. The fourth eigenvalue is 210. The calculated value is 610, then the fourth eigenvalue The final result is 500.

[0084] Understandably, the effective time spent in the mall is calculated in minutes, with any fraction of a minute counted as a full minute.

[0085] In some embodiments, the fourth non-paid feature The value is 210. If the time exceeds 100 minutes, it will be calculated as 100 minutes.

[0086] S55, based on the fifth non-paid feature, obtains the fifth feature value through the fifth preset condition.

[0087] The fifth eigenvalue includes the fifth upper limit value.

[0088] In some embodiments, the fifth preset condition includes a calculation formula. The fifth eigenvalue is obtained. The calculated value, and, based on the fifth upper limit value and the fifth eigenvalue. The calculated value confirms the fifth eigenvalue. The final result.

[0089] In one implementation, the fifth eigenvalue The value range is 500-1500, with the fifth eigenvalue having an upper limit of 1500. In other words, within the fifth eigenvalue... When the calculated value is greater than 1500, the fifth eigenvalue The final result is 1500. For example, the fifth non-paid feature. The value is 1110, the fifth eigenvalue. The calculated value is 1610, then the fifth eigenvalue The final result is 1500.

[0090] S56, based on the sixth non-paid feature, obtains the sixth feature value through the sixth preset condition.

[0091] The sixth eigenvalue includes the sixth upper limit value.

[0092] In some embodiments, the sixth preset condition includes a calculation formula. The sixth eigenvalue is obtained. The calculated value, and, based on the sixth upper limit and the sixth eigenvalue. The calculated value confirms the sixth eigenvalue. The final result.

[0093] In one implementation, the sixth eigenvalue The value range is 1500-1600, with the sixth upper limit being 1600. That is to say, within the sixth eigenvalue... When the calculated value is greater than 1600, the sixth eigenvalue The final result is 1600. For example, the sixth non-paid feature. The value is 1110, the sixth eigenvalue. The calculated value is 2610, then the sixth eigenvalue The final result is 1600.

[0094] It should be noted that the first upper limit is less than the second upper limit, the second upper limit is less than the third upper limit, the third upper limit is less than the fourth upper limit, the fourth upper limit is less than the fifth upper limit, and the fifth upper limit is less than the sixth upper limit.

[0095] By executing steps S51 to S56, the operation of encoding the second data based on the non-paying feature is completed, resulting in the second encoded data. The obtained second encoded data... .

[0096] The purpose of encoding is to transform user data into dictionary data that the Transformer model can recognize. Because many values ​​in the original data are repeated, directly inputting them into the model cannot distinguish their meanings. For example, the number of product visits is 2, and the number of recharge operations is also 2. These two values ​​are identical (and there may even be other values ​​of 2), and the model cannot distinguish which data has the value 2. However, after encoding, the value for the number of product visits is 100 + 2 = 102, and the value for the number of recharge operations is 200 + 2 = 202. Although the actual values ​​are the same, the encoded results are different, and the model will learn the meaning of these different values. Therefore, encoding requires that all values ​​are unique and that each value has a different meaning.

[0097] S6: Periodically traverse the second encoded data to construct the second dataset.

[0098] In some embodiments, when multiple users are included, when performing step S6, the users are sequentially selected from... arrive Traversal The Middle Non-paid behavior data of individual users Construct a second dataset .

[0099] In one implementation, the first step is to iterate through the sequence and select the... Seven consecutive days of non-paid behavior data of individual users This is the input data for the second dataset. The label for this input data is non-coded, non-paid behavior data, with a value of 0 or 1.

[0100] It is understandable that the method for periodic traversal can be found in step S3, and will not be elaborated here.

[0101] S7, based on the second loss function and the second dataset, train the second prediction model.

[0102] In some embodiments, the second loss function is the cross-entropy loss function, and the second prediction model is... Based on the Transformer model.

[0103] In one implementation, the formula for calculating the second loss function includes:

[0104] ;

[0105] in, Represented as from the second dataset The number of data samples in the middle; Represented as sampling The first of the data The data, if the first The data is the first Non-paid behavior data of each user; then ; Indicated as Input the second prediction model The output results obtained; yes The corresponding real tags.

[0106] Based on the second dataset mentioned above Second loss function The second prediction model is trained until the loss function value no longer decreases, resulting in the optimized second prediction model.

[0107] In this way, the second predictive model learns the mapping relationship between non-paid behaviors and user paid behaviors from the perspective of causes.

[0108] S8, based on historical payment rates, identifies the first set of paying users, the second set of paying users, and the third set of paying users.

[0109] In some embodiments, the historical payment rate for each user is calculated based on historical payment behavior data in historical behavior log data. It is understood that the historical behavior log data includes data from multiple users.

[0110] In one implementation, the formula for calculating the historical payment rate includes:

[0111] ;

[0112] in, Represented as the first One user; This represents the total number of days in the past month; m represents the m-th day of the past month. Represented as the first Historical payment rate of each user; Represented as the first The number of users in the past Is there any payment involved?

[0113] Traverse the first data using the above method To obtain the paying rate of all users :

[0114] .

[0115] Based on any user's historical payment rate, the user is divided into different sets of paying users. These different sets represent different payment frequencies. For example, payment frequency can include three categories: high-frequency payment, normal-frequency payment, and low-frequency payment. High-frequency payment corresponds to the first set of paying users. Normal-frequency payment corresponds to the second set of paying users. Low-frequency payment corresponds to the third set of paying users.

[0116] In one implementation, all users are ranked according to their payment rate. Sort from highest to lowest, and then sort all of them. Each user uses , ,..., This indicates that the sorted... Users are divided into three groups, corresponding to high-frequency (high) payments, normal payments, and low-frequency (low) payments. Thus, high-frequency (high) paying users... Normal paying users Low-frequency (low-paying) users .

[0117] S9. Based on the first dataset, establish the first evaluation dataset, the second evaluation dataset, and the third evaluation dataset corresponding to the first set of paying users, the second set of paying users, and the third set of paying users, respectively.

[0118] First dataset This includes the first encoded data and its corresponding labels. In the first dataset... Find all high-frequency paying users The corresponding data is used to construct the first evaluation dataset. In the first dataset Find all normal paying users. The corresponding data will be used to construct a second evaluation dataset. In the first dataset Find all low-frequency (low) paying users The corresponding data will be used to construct a third evaluation dataset. .

[0119] S10. Based on the second dataset, establish the fourth, fifth, and sixth evaluation datasets corresponding to the first, second, and third paid user sets, respectively.

[0120] Second dataset This includes the second encoded data and its corresponding labels. In the second dataset... Find all high-frequency paying users The corresponding data will be used to construct the fourth evaluation dataset. In the second dataset Find all normal paying users. The corresponding data will be used to construct the fifth evaluation dataset. In the second dataset Find all low-frequency (low) paying users The corresponding data will be used to construct the sixth evaluation dataset. .

[0121] S11, based on the first prediction model, a first confidence level is obtained based on the first evaluation dataset, a second confidence level is obtained based on the second evaluation dataset, and a third confidence level is obtained based on the third evaluation dataset.

[0122] According to the first prediction model ,enter The first confidence level was obtained. .

[0123] In one implementation, the first confidence level The calculation formulas include:

[0124] .

[0125] According to the first prediction model ,enter The second confidence level was obtained. .

[0126] In one implementation, the second confidence level The calculation formulas include:

[0127] .

[0128] According to the first prediction model ,enter The third confidence level was obtained. .

[0129] In one implementation, the third confidence level The calculation formulas include:

[0130] .

[0131] in, Represented as a dataset The total number of data; This indicates that the first prediction model is for the input data. The output prediction results; Indicates a pointer function, for example, It means that when the conditions are met hour, ,otherwise .

[0132] S12, based on the second prediction model, the fourth confidence level is obtained based on the fourth evaluation dataset, the fifth confidence level is obtained based on the fifth evaluation dataset, and the sixth confidence level is obtained based on the sixth evaluation dataset.

[0133] According to the second prediction model ,enter The fourth confidence level was obtained. .

[0134] In one implementation, the fourth confidence level The calculation formulas include:

[0135] .

[0136] According to the second prediction model ,enter The fifth confidence level was obtained. .

[0137] In one implementation, the fifth confidence level The calculation formulas include:

[0138] .

[0139] According to the second prediction model ,enter The sixth confidence level was obtained. .

[0140] In one implementation, the sixth confidence level The calculation formulas include:

[0141] .

[0142] S13. Based on the first combination probability, the second combination probability, or the third combination probability, obtain the paid prediction result.

[0143] The first combination probability includes the first confidence level and the fourth confidence level.

[0144] The second combination probability includes the second confidence level and the fifth confidence level.

[0145] The third combination probability includes the third confidence level and the sixth confidence level.

[0146] In some embodiments, when performing step S13, the user payment behavior prediction method based on trends, causes, and confidence levels further includes the following steps:

[0147] S131, confirm the first historical paid behavior and historical non-paid behavior of the user to be predicted.

[0148] In one implementation, the user to be predicted can be the first... Individual users. The first historical payment behavior can be the coded data of the user to be predicted over the past seven days, including... Historical non-paying behavior can be coded data of the user to be predicted over the past seven days, including... .

[0149] S132, input the first historical payment behavior into the first prediction model to obtain the first predicted payment rate.

[0150] In one implementation, the first historical payment behavior will be... Inputting the data into the first prediction model yields the first predicted payment rate. .

[0151] S133, input historical non-paying behaviors into the second prediction model to obtain the second predicted payment rate.

[0152] In one implementation, historical non-paid behavior will be... Inputting the data into the second prediction model yields the second predicted payment rate. .

[0153] S134, confirm the second historical payment behavior of the user to be predicted.

[0154] In one implementation, the second historical payment behavior can be the first... A user's historical payment behavior over the past month (or 30 days), including .

[0155] S135, based on the second historical payment behavior, the payment rate to be tested is obtained.

[0156] In one implementation, it is based on the second historical payment behavior. To obtain the pay rate to be tested .

[0157] S136, Based on the test payment rate, confirm the target payment level.

[0158] The target payment tiers are the first set of paying users, the second set of paying users, or the third set of paying users.

[0159] In one implementation, based on The value of the first... Each user belongs to a high-frequency paying user (i.e., the first set of paying users), a normal paying user (i.e., the second set of paying users), or a low-frequency paying user (i.e., the third set of paying users). It is understandable that when classifying a user into different paying frequency user sets, the classification result of step S8 can be used as the basis, which will not be elaborated here.

[0160] S137, if the target payment level is the first set of paying users, then the payment prediction result is obtained according to the first combination probability.

[0161] If the user to be predicted belongs to the first set of paying users, then based on the first confidence level... Fourth confidence level First predicted paid conversion rate Second predicted payment rate Get the paid prediction results.

[0162] In one implementation, the payment prediction result can be a probability value, and the formula for calculating the payment prediction result includes:

[0163] ;

[0164] in, This is represented as the first predicted payment rate; This is represented as the second predicted payment rate; This is represented as the first confidence level; This is represented as the fourth confidence level.

[0165] For example, assuming the model confidence level is =0, If the model confidence level is 0, then the final R = 0.5. This means the probability of a user paying is 0.5, making it difficult to predict whether a user will pay in the future. This is reasonable because a model confidence level of 0 means the model is unreliable, and we cannot rely on the model's results to determine whether a user will pay in the future. Conversely, if the model confidence level is 1, meaning the model's predictions are highly reliable, then the final result R will directly depend on the model's predictions.

[0166] In other embodiments, if the user to be predicted belongs to the second set of paying users, then based on the second confidence level... Fifth confidence level First predicted paid conversion rate Second predicted payment rate Get the paid prediction results.

[0167] The formula for calculating the paid prediction result includes:

[0168] ;

[0169] In some other embodiments, if the user to be predicted belongs to the third set of paying users, then based on the third confidence level... Sixth confidence level First predicted paid conversion rate Second predicted payment rate Get the paid prediction results.

[0170] The formula for calculating the paid prediction result includes:

[0171] ;

[0172] The advantage of this formula for calculating the probability of predicted payment behavior is that it improves the accuracy of model predictions by combining the predicted payment rate and confidence level.

[0173] This application identifies the trend and causal characteristics influencing user payment behavior and constructs a first and a second prediction model for user payment behavior based on these characteristics. The first prediction model considers the impact of past user payment behavior on future payments, its core being to predict future payment behavior trends through trends in past payment behavior. The second prediction model considers the impact of user non-paying behavior on future payment behavior from a causal perspective. The two prediction models predict payment behavior from different perspectives, resulting in higher accuracy and a lower false positive rate. Furthermore, this application designs and incorporates model confidence levels to predict user payment behavior, taking into account the impact of the reliability of the prediction model itself on payment behavior.

[0174] Furthermore, the calculation method for model prediction confidence considers the reliability of the first and second prediction models in predicting different user payment behaviors from the perspective of model predictive ability. Model prediction confidence is grouped according to the similarity of user payment behaviors, and the model prediction confidence for each group of users paying versus not paying is calculated based on the first and second prediction models. Based on an evaluation dataset constructed from three categories, predictions are made using the trend-based first prediction model and the cause-based second prediction model for user payment behavior, respectively. The model prediction confidence of the first and second prediction models is calculated for different categories based on the prediction results and historical actual payment behavior results. Finally, combining the results of the first and second prediction models and the model prediction confidence to predict future user payment behavior can effectively improve the accuracy of payment behavior prediction.

[0175] In summary, this application combines machine learning, fully considers user payment behavior trends, payment reasons, and the impact of model prediction confidence on user payment behavior, which can effectively improve the accuracy of payment behavior prediction.

[0176] Based on the same application concept, this application also provides a user payment behavior prediction system based on trends, causes, and confidence levels. The method corresponding to the user payment behavior prediction system based on trends, causes, and confidence levels can be the user payment behavior prediction method based on trends, causes, and confidence levels in the aforementioned embodiments, and its problem-solving principle is similar to that method. The user payment behavior prediction system based on trends, causes, and confidence levels provided in this application includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to execute the methods and / or technical solutions of the various embodiments of this application.

[0177] Another embodiment of this application provides a computer-readable storage medium having computer program instructions stored thereon, which can be executed by a processor to implement the methods and / or technical solutions of any one or more embodiments of this application described above.

[0178] Specifically, this embodiment may employ any combination of one or more computer-readable media. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0179] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0180] The program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0181] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0182] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0183] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0184] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or page components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.

[0185] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0186] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.

[0187] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0189] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A method for predicting user payment behavior based on trends, causes, and confidence levels, characterized in that, include: User data is divided into first data and second data. The first data consists of each user's paid behavior data, including daily paid data for each user, which includes time characteristics. These time characteristics characterize the relationship between user information and time information. The second data consists of each user's non-paid behavior data, including daily non-paid data for each user, which includes multiple non-paid features. These non-paid features include... The user's past number Number of visits to the mall on the day, number of visits The user's past number Number of times the user visits products in the online store The user's past number Number of operations on the recharge interface, number of times The user's past number The time spent in the mall each day, the number The user's past number Total number of games played on the day and the The user's past number Number of logins per day; Based on the aforementioned time characteristics, the first data is encoded to obtain first encoded data; Periodically traverse the first encoded data to construct the first dataset; A first prediction model is trained based on the first loss function and the first dataset. Based on the non-payment feature, the second data is encoded to obtain the second encoded data; Periodically traverse the second encoded data to construct the second dataset; A second prediction model is trained based on the second loss function and the second dataset. Based on historical payment rates, the first set of paying users, the second set of paying users, and the third set of paying users are identified. Based on the first dataset, a first evaluation dataset, a second evaluation dataset, and a third evaluation dataset are respectively established corresponding to the first set of paying users, the second set of paying users, and the third set of paying users. Based on the second dataset, a fourth evaluation dataset, a fifth evaluation dataset, and a sixth evaluation dataset are respectively established corresponding to the first set of paying users, the second set of paying users, and the third set of paying users. Based on the first prediction model, a first confidence level is obtained based on the first evaluation dataset, a second confidence level is obtained based on the second evaluation dataset, and a third confidence level is obtained based on the third evaluation dataset; According to the second prediction model, a fourth confidence level is obtained based on the fourth evaluation dataset, a fifth confidence level is obtained based on the fifth evaluation dataset, and a sixth confidence level is obtained based on the sixth evaluation dataset; The payment prediction result is obtained based on the first combination probability, the second combination probability, or the third combination probability; the first combination probability includes the first confidence level and the fourth confidence level; the second combination probability includes the second confidence level and the fifth confidence level; the third combination probability includes the third confidence level and the sixth confidence level; the payment prediction result is a probability value used to characterize the user's willingness to pay.

2. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1, characterized in that, The first encoded data includes a first encoded element, a second encoded element, a third encoded element, a fourth encoded element, a fifth encoded element, a sixth encoded element, a seventh encoded element, an eighth encoded element, and a ninth encoded element, with a seven-day cycle. When performing the step of periodically traversing the first encoded data to construct the first dataset, the user payment behavior prediction method based on trends, causes, and confidence levels further includes: The first encoding element, the second encoding element, the third encoding element, the fourth encoding element, the fifth encoding element, the sixth encoding element, and the seventh encoding element are used as the first input data of the first dataset; The non-encoded data corresponding to the eighth encoded element is used as the first data label of the first input data; The second encoding element, the third encoding element, the fourth encoding element, the fifth encoding element, the sixth encoding element, the seventh encoding element, and the eighth encoding element are used as the second input data of the first dataset; The non-encoded data corresponding to the ninth encoded element is used as the second data label of the first input data.

3. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, The first loss function is the cross-entropy loss function, and the first prediction model is based on the Transformer model.

4. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, The second coded data includes first non-paid data; the first non-paid data includes first non-paid features, second non-paid features, third non-paid features, fourth non-paid features, fifth non-paid features, and sixth non-paid features; when performing the step of encoding the second data based on the non-paid features to obtain the second coded data, the user payment behavior prediction method based on trends, causes, and confidence levels further includes: Based on the first non-paid feature, a first feature value is obtained through a first preset condition; the first feature value includes a first upper limit value. Based on the second non-paid feature, a second feature value is obtained through a second preset condition; the second feature value includes a second upper limit value. Based on the third non-paid feature, a third feature value is obtained through a third preset condition; the third feature value includes a third upper limit value. Based on the fourth non-paid feature, a fourth feature value is obtained through a fourth preset condition; the fourth feature value includes a fourth upper limit value. Based on the fifth non-paid feature, a fifth feature value is obtained through the fifth preset condition; the fifth feature value includes a fifth upper limit value. Based on the sixth non-paid feature, a sixth feature value is obtained through a sixth preset condition; the sixth feature value includes a sixth upper limit value. Wherein, the first upper limit value is less than the second upper limit value, the second upper limit value is less than the third upper limit value, the third upper limit value is less than the fourth upper limit value, the fourth upper limit value is less than the fifth upper limit value, and the fifth upper limit value is less than the sixth upper limit value.

5. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, When performing the step of obtaining the payment prediction result based on the first combination probability, the second combination probability, or the third combination probability, the user payment behavior prediction method based on trend, cause, and confidence further includes: Identify the first historical paid behavior and historical non-paid behavior of the user to be predicted; Input the first historical payment behavior into the first prediction model to obtain the first predicted payment rate; The historical non-paying behaviors are input into the second prediction model to obtain the second predicted payment rate; Confirm the second historical payment behavior of the user to be predicted; Based on the second historical payment behavior, the payment rate to be tested is obtained; Based on the measured payment rate, a target payment level is determined; the target payment level is the first set of paying users, the second set of paying users, or the third set of paying users. If the target payment level is the first set of paying users, then the payment prediction result is obtained based on the first combination probability.

6. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, Before performing the step of dividing user data into first data and second data, the user payment behavior prediction method based on trend, cause, and confidence level further includes: The user's historical behavior log data on the game platform is obtained, and the historical behavior log data is cleaned and preprocessed to obtain the user data.

7. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, The second loss function is the cross-entropy loss function, and the second prediction model is based on the Transformer model.

8. The user payment behavior prediction method based on trend, cause, and confidence level according to claim 1 or 2, characterized in that, The encoding formula for calculating the first encoded data includes: ； in, Represented as the first The day of the week corresponding to each day ranges from 1 to 7; Represented as the first The number of users in the past Is there any payment involved? 9. A user payment behavior prediction system based on trend, cause, and confidence level, characterized in that, include: At least one processor; A memory that is communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 8.

10. A computer-readable medium having stored thereon computer program instructions that can be executed by a processor to implement the method as described in any one of claims 1 to 8.

Images