Training Method and Device for Bicycle Riding Card Usage Frequency Prediction Model, Resource Acquisition Value Prediction Method and Device, Computer Equipment, Storage Medium and Computer Program Product

Abstract

The present application relates to a training method and device for a riding card usage frequency prediction model, a resource acquisition value prediction method and device, a computer device, a storage medium, and a computer program product. The method includes: dividing at least one riding times category; for any riding times category, determining the category feature of the riding times category, the first target historical riding card usage frequency, and the target current riding card usage frequency according to the account features of each target user account in the riding times category; inputting the category features of each riding times category and the first target historical riding card usage frequency corresponding to each riding times category into an initial riding card usage frequency prediction model to obtain the expected riding card usage frequency corresponding to each riding times category; and training the model according to each expected riding card usage frequency and each target current riding card usage frequency. Using this method can improve the prediction accuracy of the riding card usage frequency.

Description

Technical Field

[0001] The present application relates to the field of big data, and in particular to a method and device for training a prediction model for the usage frequency of a riding card, a method and device for predicting a resource acquisition value, a computer device, a storage medium, and a computer program product. Background Art

[0002] With the popularization of shared electric bicycles, more and more users choose to purchase riding cards to save travel costs. Therefore, reasonably predicting the frequency of users' use of riding cards and providing relevant services based on the frequency of users' use of riding cards are crucial for both user satisfaction and the platform's operation strategy.

[0003] In the prior art, the platform usually predicts the future usage frequency of a riding card by the historical usage frequency of the riding card by the user. However, there are many factors affecting whether a user uses a riding card. If only predicting the future usage frequency of a riding card by the historical usage frequency of the user's riding card will bring a large error, resulting in poor prediction accuracy. Summary of the Invention

[0004] Based on this, it is necessary to provide a method and device for training a prediction model for the usage frequency of a riding card, a method and device for predicting a resource acquisition value, a computer device, a storage medium, and a computer program product for the above technical problems.

[0005] In a first aspect, the present application provides a method for training a prediction model for the usage frequency of a riding card. The method includes:

[0006] Dividing at least one riding frequency category according to the historical riding times corresponding to each user account, and determining the user accounts corresponding to each riding frequency category;

[0007] For any one of the riding frequency categories, determining the category feature of the riding frequency category according to the account features of each target user account corresponding to the riding frequency category, determining a first target historical riding card usage frequency according to the historical riding card usage frequencies of each target user account, and determining a target current riding card usage frequency according to the current riding card usage frequencies of each target user account;

[0008] Inputting the category features of each riding frequency category and the first target historical riding card usage frequencies corresponding to each riding frequency category into an initial riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each riding frequency category;

[0009] Adjusting the initial riding card usage frequency prediction model according to each expected riding card usage frequency and each target current riding card usage frequency to obtain a trained riding card usage frequency prediction model.

[0010] In one embodiment, dividing at least one riding frequency category according to the historical riding frequencies corresponding to each user account includes:

[0011] Obtaining the historical riding frequencies corresponding to each user account, and respectively determining the number of user accounts corresponding to each of the historical riding frequencies;

[0012] Dividing at least one riding frequency category according to each of the historical riding frequencies and the number of user accounts corresponding to each of the historical riding frequencies, so that the number of user accounts corresponding to each of the riding frequency categories respectively meets a preset condition.

[0013] In one embodiment, dividing at least one riding frequency category according to each of the historical riding frequencies and the number of user accounts corresponding to each of the historical riding frequencies includes:

[0014] For any one of the historical riding frequencies, determining a second target historical riding card usage frequency corresponding to the historical riding frequency according to the historical riding card usage frequency of the user account corresponding to the historical riding frequency;

[0015] Determining a change rate of the second target historical riding card usage frequency with respect to the historical riding frequency according to each of the historical riding frequencies and each of the second target historical riding card usage frequencies;

[0016] Dividing at least one riding frequency category according to each of the historical riding frequencies, the change rate, and the number of user accounts corresponding to each of the historical riding frequencies, so that for any one of the riding frequency categories, the change rate is less than a change rate threshold within the range of historical riding frequencies corresponding to the riding frequency category, and the number of user accounts corresponding to each of the riding frequency categories respectively meets a preset condition.

[0017] In one embodiment, determining the category feature of the riding frequency category according to the account features of each target user account corresponding to the riding frequency category includes:

[0018] Determining each target user account corresponding to the riding frequency category;

[0019] For any preset account feature category, determine the degree of dispersion of the target account features of each of the target user accounts belonging to the preset account feature category. When the degree of dispersion is less than the preset degree-of-dispersion threshold, determine the first target account feature based on each of the target account features, and use the first target account feature as the category feature of the riding frequency category corresponding to the preset account feature category; or when the degree of dispersion is greater than the preset degree-of-dispersion threshold, perform clustering processing on each of the target user accounts according to the target account features, determine the second target account feature based on the number of target user accounts in each of the target user account categories obtained after clustering, and use the second target account feature as the category feature of the riding frequency category corresponding to the preset account feature category.

[0020] In one embodiment, the determining the second target account feature based on the number of target user accounts in each of the target user account categories obtained after clustering includes:

[0021] Based on the number of target user accounts in each of the target user account categories obtained after clustering and the total number of each of the target user accounts, respectively determine the account number proportion corresponding to each of the target user account categories, and use each of the account number proportions as the second target account feature.

[0022] In a second aspect, the present application also provides a method for predicting a resource acquisition value. The method includes:

[0023] Divide at least one riding frequency category according to the historical riding frequencies corresponding to each user account, and determine the user accounts corresponding to each riding frequency category;

[0024] For any one of the riding frequency categories, determine the category feature of the riding frequency category according to the account features of each target user account corresponding to the riding frequency category, and determine the first target historical riding card usage frequency according to the historical riding card usage frequencies of each of the target user accounts;

[0025] Input the category features of each of the riding frequency categories and the first target historical riding card usage frequencies corresponding to each of the riding frequency categories into a riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding frequency categories;

[0026] Determine the expected riding card resource acquisition values corresponding to each of the riding frequency categories according to the expected riding card usage frequencies corresponding to each of the riding frequency categories;

[0027] Wherein, the riding card usage frequency prediction model is the trained riding card usage frequency prediction model described in any of the foregoing embodiments.

[0028] In one embodiment, determining the expected cycling card resource acquisition value corresponding to each cycling times category according to the expected cycling card usage frequency corresponding to each cycling times category includes:

[0029] For any one of the cycling times categories, determine the expected cycling card resource acquisition value corresponding to the cycling times category according to the cycling card resource compensation rate, the expected cycling card usage frequency, and the single cycling card resource acquisition value.

[0030] In a third aspect, the present application also provides a training device for a cycling card usage frequency prediction model. The device includes:

[0031] A division module, configured to divide at least one cycling times category according to the historical cycling times corresponding to each user account, and determine the user accounts corresponding to each cycling times category respectively;

[0032] A determination module, configured to, for any one of the cycling times categories, determine the category characteristics of the cycling times category according to the account characteristics of each target user account corresponding to the cycling times category, determine the first target historical cycling card usage frequency according to the historical cycling card usage frequencies of each target user account, and determine the target current cycling card usage frequency according to the current cycling card usage frequencies of each target user account;

[0033] An input module, configured to input the category characteristics of each cycling times category and the first target historical cycling card usage frequency corresponding to each cycling times category into an initial cycling card usage frequency prediction model, and obtain the expected cycling card usage frequency corresponding to each cycling times category respectively;

[0034] An adjustment module, configured to adjust the initial cycling card usage frequency prediction model according to each expected cycling card usage frequency and each target current cycling card usage frequency, and obtain a trained cycling card usage frequency prediction model.

[0035] In one embodiment, the division module is further configured to:

[0036] Obtain the historical cycling times corresponding to each user account, and respectively determine the number of user accounts corresponding to each historical cycling times;

[0037] Divide at least one cycling times category according to each historical cycling times and the number of user accounts corresponding to each historical cycling times, so that the number of user accounts corresponding to each cycling times category respectively meets a preset condition.

[0038] In one embodiment, the division module is further configured to:

[0039] For any one of the historical riding times, according to the historical riding card usage frequency of the user account corresponding to the historical riding time, determine the second target historical riding card usage frequency corresponding to the historical riding time;

[0040] According to each of the historical riding times and each of the second target historical riding card usage frequencies, determine the change rate of the second target historical riding card usage frequency with respect to the historical riding time;

[0041] According to each of the historical riding times, the change rate, and the number of the user accounts corresponding to each of the historical riding times, divide at least one riding time category, so that for any one of the riding time categories, the change rate is less than a change rate threshold within the historical riding time range corresponding to the riding time category, and the number of each of the user accounts corresponding to each of the riding time categories meets a preset condition.

[0042] In one embodiment, the determining module is further configured to:

[0043] Determine each target user account corresponding to the riding time category;

[0044] For any one of the preset account feature categories, determine the dispersion degree of the target account features of each of the target user accounts belonging to the preset account feature category, and when the dispersion degree is less than a preset dispersion degree threshold, determine a first target account feature according to each of the target account features, and use the first target account feature as the category feature of the riding time category corresponding to the preset account feature category; or when the dispersion degree is greater than the preset dispersion degree threshold, perform clustering processing on each of the target user accounts according to the target account features, determine a second target account feature according to the number of the target user accounts in each of the target user account categories obtained after clustering, and use the second target account feature as the category feature of the riding time category corresponding to the preset account feature category.

[0045] In one embodiment, the determining module is further configured to:

[0046] According to the number of the target user accounts in each of the target user account categories obtained after clustering and the total number of each of the target user accounts, respectively determine the proportion of the account number corresponding to each of the target user account categories, and use each of the account number proportions as a second target account feature.

[0047] Fourthly, the present application further provides a resource acquisition value prediction device. The device includes:

[0048] A division module, configured to divide at least one riding frequency category according to the historical riding frequencies corresponding to each user account, and determine the user accounts corresponding to each riding frequency category respectively;

[0049] A first determination module, configured to, for any one of the riding frequency categories, determine the category characteristics of the riding frequency category according to the account characteristics of each target user account corresponding to the riding frequency category, and determine the first target historical riding card usage frequency according to the historical riding card usage frequencies of each of the target user accounts;

[0050] An input module, configured to input the category characteristics of each of the riding frequency categories and the first target historical riding card usage frequencies corresponding to each of the riding frequency categories into a riding card usage frequency prediction model, to obtain the expected riding card usage frequencies corresponding to each of the riding frequency categories respectively;

[0051] A second determination module, configured to determine the expected riding card resource acquisition values corresponding to each of the riding frequency categories according to the expected riding card usage frequencies corresponding to each of the riding frequency categories;

[0052] Wherein, the riding card usage frequency prediction model is the trained riding card usage frequency prediction model described in any one of the foregoing embodiments.

[0053] In one embodiment, the second determination module is further configured to:

[0054] For any one of the riding frequency categories, determine the expected riding card resource acquisition value corresponding to the riding frequency category according to the riding card resource compensation rate, the expected riding card usage frequency, and the single - time riding card resource acquisition value.

[0055] In a fifth aspect, the present application further provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the above - mentioned method in any one of the above items is implemented.

[0056] In a sixth aspect, the present application further provides a computer - readable storage medium. On the computer - readable storage medium, a computer program is stored, and when the computer program is executed by a processor, the above - mentioned method in any one of the above items is implemented.

[0057] In a seventh aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above - mentioned method in any one of the above items is implemented.

[0058] The training method and device for the above-mentioned cycling card usage frequency prediction model, the resource acquisition value prediction method and device, the computer device, the storage medium, and the computer program product divide multiple cycling times categories according to the historical cycling times, comprehensively obtain the category features for each user account in the cycling times category, the first target historical cycling card usage frequency and the target current cycling card usage frequency of the cycling times category, and train the model through the category features, the first target historical cycling card usage frequency, and the target current cycling card usage frequency. Furthermore, in the application process, the model is used to predict the cycling card usage frequency of each cycling times category. Since the cycling card usage frequency of a single user account fluctuates greatly, using the data of a single user account as a sample to train the model may lead to poor model accuracy. Therefore, in the embodiments of the present application, user accounts are aggregated according to the historical cycling times of the user accounts, and the problem of predicting the relationship between the historical cycling card usage frequency, the cycling times, and the future cycling card usage frequency of a user account is transformed into the problem of predicting the relationship between the historical cycling card usage frequency and the future cycling card usage frequency of a user group with a certain range of historical cycling times. That is to say, when predicting the future cycling card usage frequency of a user, the characteristics of other users with the same historical cycling times as the user are referred to. Therefore, the trained model performs better in predicting the overall future cycling card usage frequency of a certain user group, and can improve the prediction accuracy of the cycling card usage frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic flowchart of the training method for the cycling card usage frequency prediction model in an embodiment;

[0060] Figure 2 It is a schematic flowchart of step 102 in an embodiment;

[0061] Figure 3 It is a schematic flowchart of step 104 in an embodiment;

[0062] Figure 4 It is a schematic flowchart of the resource acquisition value prediction method in an embodiment;

[0063] Figure 5 It is a block diagram of the structure of the training device for the cycling card usage frequency prediction model in an embodiment;

[0064] Figure 6 It is a block diagram of the structure of the training device for the resource acquisition value prediction model in an embodiment;

[0065] Figure 7 It is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] To make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0067] In one embodiment, as Figure 1 shown, a method for training a prediction model of the usage frequency of a cycling card is provided. In this embodiment, taking the application of this method to a server as an example, it can be understood that this method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0068] Step 102, divide at least one cycling times category according to the historical cycling times corresponding to each user account, and determine the user accounts corresponding to each cycling times category respectively.

[0069] In the embodiment of the present application, the historical cycling times refer to the cycling times of a user account within a certain preset time period before. The preset time period can be set by those skilled in the art, such as the previous 7 days, the previous 30 days, the previous 60 days, etc. The cycling times category is divided according to the historical cycling times corresponding to each user account, and one cycling times category corresponds to one historical cycling times interval. In one example, the range covered by the historical cycling times can be determined, and the range can be equally spaced according to a preset interval to obtain each cycling times category. For example, when the preset time period is the previous 30 days, the minimum historical cycling times is 1 time, and the maximum is 30 times. If the preset interval is 10, the divided cycling times categories are 1 time - 10 times, 11 times - 20 times, and 21 times - 30 times respectively.

[0070] In one embodiment, dividing at least one cycling times category according to the historical cycling times corresponding to each user account includes:

[0071] Dividing at least one cycling times category according to each historical cycling times and the number of user accounts corresponding to each historical cycling times, so that the number of user accounts corresponding to each cycling times category meets a preset condition.

[0072] In the embodiment of the present application, for each historical cycling times, the number of user accounts corresponding to this historical cycling times can be determined, and the cycling times category can be divided according to the number of user accounts.

[0073] In one example, to make the extracted category features more stable, the preset condition can be set such that the number of user accounts corresponding to each riding frequency category is greater than a preset threshold. When dividing the riding frequency categories according to this preset condition, the historical riding frequencies can be traversed sequentially starting from the highest or lowest historical riding frequency, and a temporary riding frequency category can be established. The first traversed historical riding frequency is added to the temporary riding frequency category, and the range of historical riding frequencies covered by the current temporary riding frequency category is the first traversed historical riding frequency. If the number of user accounts corresponding to the temporary riding frequency category is greater than the preset threshold, the temporary riding frequency category is used as a riding frequency category. Then, the temporary riding frequency category is reset, and the next traversed historical riding frequency is added to the temporary riding frequency category. If the number of user accounts corresponding to the temporary riding frequency category is less than or equal to the preset threshold, the next traversed historical riding frequency is added to the temporary riding frequency category, so that the range of historical riding frequencies covered by the temporary riding frequency category becomes the historical riding frequencies already covered by the temporary riding frequency category and the next traversed historical riding frequency. The above process is repeated until all historical riding frequencies are traversed.

[0074] In another example, to make the quality of the category features extracted for each riding frequency category approximate, the preset condition can be set such that the difference in the number of user accounts corresponding to each riding frequency category is as small as possible. The number of riding frequency categories to be divided can be preset, and according to the number of riding frequency categories and the total number of user accounts, the target number of user accounts corresponding to each riding frequency category can be determined, and the historical riding frequencies are traversed in the manner of the foregoing example to divide the historical riding frequencies into each riding frequency category. In this example, if the current number of user accounts corresponding to the temporary riding frequency category is less than or equal to the target number, it is judged which is smaller between the difference between the target number and the current number and the difference between the target number and the sum of the current number and the number of user accounts corresponding to the next historical riding frequency to be traversed. If the difference between the target number and the current number is smaller, the temporary riding frequency category is used as a riding frequency category, and the temporary riding frequency is reset. If the difference is smaller after adding the number of user accounts corresponding to the next historical riding frequency to be traversed, the next historical riding frequency is traversed and added to the temporary riding frequency category. The above process is repeated until all historical riding frequencies are traversed.

[0075] In one embodiment, as Figure 2 shown, in step 102, dividing at least one riding frequency category according to each historical riding frequency and the number of user accounts corresponding to each historical riding frequency includes:

[0076] Step 202: For any historical number of rides, determine the second target historical usage frequency of the historical ride card corresponding to the historical number of rides according to the historical usage frequency of the historical ride card of the user account corresponding to the historical number of rides.

[0077] Step 204: Determine the change rate of the second target historical usage frequency of the historical ride card with respect to the historical number of rides according to each historical number of rides and each second target historical usage frequency of the historical ride card.

[0078] Step 206: Divide at least one category of the number of rides according to each historical number of rides and the change rate, so that for any one category of the number of rides, the change rate is less than the change rate threshold within the range of the historical number of rides corresponding to the category of the number of rides.

[0079] In the embodiments of the present application, the historical usage frequencies of the historical ride cards of the user accounts corresponding to each historical number of rides can be synthesized to obtain the second target historical usage frequency of the historical ride card. For example, the average value of each historical usage frequency of the historical ride card can be taken as the second target historical usage frequency of the historical ride card, or each historical usage frequency of the historical ride card can be weighted and averaged (the weights can be positively correlated with the credibility of the user account, for example, the longer the usage duration of the user account, the higher the credibility, or the more the total number of rides of the user account, the higher the credibility, etc.) to obtain the second target historical usage frequency of the historical ride card. The embodiments of the present application do not make specific limitations on this.

[0080] After obtaining the second target historical usage frequencies of the historical ride cards corresponding to each historical number of rides, the change rate of the second target historical usage frequency of the historical ride card with respect to the historical number of rides can be determined. This change rate includes the change rate of the second target historical usage frequency of the historical ride card at each historical number of rides. For example, the second target historical usage frequency of the historical ride card and the historical number of rides can be fitted by a function, and the derivative of the function at each historical number of rides can be used as the change rate of the second target historical usage frequency of the historical ride card at each historical number of rides.

[0081] To ensure the prediction accuracy of the model, it is possible to make the change rate of the second target historical riding card usage frequency less than the change rate threshold within the historical riding times range corresponding to each riding times category, so that the riding card usage frequencies of each user account in the riding times category are not very different, and to avoid the situation where, for an individual user, the predicted riding card usage frequency is too different from the actual riding card usage frequency of the user. The change rate being less than the change rate threshold can mean that the change rate of the second target historical riding card usage frequency at each historical riding time within the historical riding times range is less than the change rate threshold, or it can mean that the average value of the change rates at each historical riding time within the historical riding times range is less than the change rate threshold. The embodiments of the present application do not make specific limitations on this, and it can be selected by those skilled in the art according to actual needs. Each historical riding time can be traversed in the manner of the foregoing example, and each historical riding time is divided into various riding times categories. In this example, it is determined whether the change rate corresponding to the temporary riding times category will be greater than or equal to the change rate threshold after adding the next historical riding time to be traversed to the temporary riding times category. If so, the next historical riding time to be traversed is not added to the temporary riding times category, the temporary riding times category is used as the riding times category, and the temporary riding times are reset. If the change rate is still less than the change rate threshold after adding the number of user accounts corresponding to the next historical riding time to be traversed, the next historical riding time is added to the temporary riding times category. The above process is repeated until all historical riding times are traversed.

[0082] After determining each riding times category, the user accounts whose historical riding times are within the historical riding times interval covered by the riding times category are used as the user accounts corresponding to the riding times category.

[0083] Step 104, for any riding times category, according to the account characteristics of each target user account corresponding to the riding times category, determine the category characteristics of the riding times category, determine the first target historical riding card usage frequency according to the historical riding card usage frequencies of each target user account, and determine the target current riding card usage frequency according to the current riding card usage frequencies of each target user account.

[0084] In the embodiments of the present application, a user account corresponding to a certain category of riding times is referred to as the target user account of this category of riding times. The first target historical riding card usage frequency and the target current riding card usage frequency corresponding to the category of riding times are determined according to the historical riding card usage frequency and the current riding card usage frequency of each target user account in the category of riding times respectively. The historical riding card usage frequency is the proportion of the number of times a user rides using a riding card within a preset time period in all the user's riding times. The current riding card usage frequency is the proportion of the number of times a user rides using a riding card within the time period to be predicted (such as the next 1 day, the next 7 days) in all the user's riding times. The average value, weighted average value, etc. of the historical riding card usage frequencies of each target user account can be obtained to get the first target historical riding card usage frequency (the first target historical riding card usage frequency and the second target historical riding card usage frequency can be the same or different), and the average value, weighted average value, etc. of the current riding card usage frequencies can be obtained to get the target current riding card usage frequency.

[0085] The category characteristics of the category of riding times can be comprehensively obtained according to the account characteristics of each target user account corresponding to the category of riding times. In one example, for account characteristics in the format of numerical values (such as the frequency of a user account using a riding card, the time point when a user account rides, etc.), the category characteristics can be obtained by comprehensively processing each account characteristic in any way such as taking the average value, variance, peak value, or calculating the proportion of the number of user accounts in each interval after dividing the interval. For example, the average value of the number of times a user account obtains a riding card within a preset time period can be calculated to obtain the category characteristics; the age interval can be divided (for example, divided into 9 intervals: [<18], [18, 25], [26, 30], [31, 35], [36, 40], [41, 45], [46,50], [51, 55], [>56]), and then the proportion of the number of user accounts corresponding to each interval in the total number of all user accounts can be calculated to obtain the category characteristics.

[0086] For account features in other formats (such as what day of the week it is when a user account rides a bike, whether it is a holiday, the area where the user account rides a bike, etc.), the account features can be converted into numerical values through one-hot encoding, and then category features can be obtained through the processing method for account features in numerical format. It is also possible to classify each account feature by obtaining all possible values of the account feature, and then calculate the number of target user accounts corresponding to each category of account features and the proportion in the total number of target user accounts to obtain the category feature. For example, for the account feature of what day of the week it is when a user account rides a bike, all possible values of this account feature are seven days from Monday to Sunday. This account feature can be divided into seven categories from Monday to Sunday, or divided into two categories such as weekdays and rest days, etc. Determine the number of target user accounts belonging to each category, calculate the proportion of the number of target user accounts in each category in the total number of target user accounts, and use the set of ratios of each category as the category feature.

[0087] In one embodiment, as Figure 3 shown, in step 104, according to the account features of each target user account corresponding to the riding frequency category, determine the category feature of the riding frequency category, including:

[0088] Step 302, determine each target user account corresponding to the riding frequency category;

[0089] Step 304, for any preset account feature category, determine the dispersion degree of the target account features of each target user account belonging to the preset account feature category, and when the dispersion degree is less than the preset dispersion degree threshold, determine the first target account feature according to each target account feature, and use the first target account feature as the category feature of the riding frequency category corresponding to the preset account feature category; or,

[0090] Step 306, when the dispersion degree is greater than the preset dispersion degree threshold, perform clustering processing on each target user account according to the target account features, determine the second target account feature according to the number of target user accounts in each target user account category obtained after clustering, and use the second target account feature as the category feature of the riding frequency category corresponding to the preset account feature category.

[0091] In the embodiments of the present application, for account features with little difference among each target user account, a feature can be obtained by synthesizing these account features as the category feature. For account features with large differences among each target user account, clustering can be performed on these account features, and the category feature can be determined according to the number of user accounts belonging to each clustering category.

[0092] The preset account feature category refers to the pre-determined category of account features that need to be obtained. According to each preset account feature category, a category feature can be synthesized based on the target account features of each target user account belonging to the preset account feature category. The degree of dispersion of the target account features of the target user account belonging to the preset account feature category can be obtained according to statistical methods. For example, the degree of dispersion can be obtained by calculating the variance, standard deviation, etc. of each target account feature. The embodiments of the present application do not make specific limitations on this.

[0093] When the degree of dispersion is small (less than the preset dispersion threshold), a category feature can be obtained based on each target account feature. For example, the category feature can be obtained by summing, averaging, etc. of each target account feature. When the degree of dispersion is large, the category feature obtained by synthesizing each target account feature does not accurately represent each target account feature. Therefore, the target user accounts can be clustered according to the target account features, and the second target account feature can be determined according to the number of target user accounts belonging to each clustering category. In one example, each clustering category can be sorted in descending order according to the number of target user accounts, and the obtained sequence can be used as the second target account feature. Taking the preset account feature category as the week, the target account features including seven days from Monday to Sunday, and the clustering categories obtained during clustering being working days and rest days as an example, when obtaining the second target account feature according to this method, the second target account feature of a certain riding frequency category may be {working days, rest days}, or it may be {rest days, working days}. When training the model, the second target account feature that can be input into the model is obtained by converting the working days and rest days into different encodings (for example, converting the working days to 1 and the rest days to 0).

[0094] In another example, according to the number of target user accounts in each target user account category obtained after clustering and the total number of target user accounts, the proportion of the number of accounts corresponding to each target user account category can be determined respectively, and each proportion of the number of accounts can be used as the second target account feature. Still taking the preset account feature category as the week and the clustering categories obtained during clustering being working days and rest days as an example, the second target account feature of a certain riding frequency category may be {30%, 70%}, {50%, 50%}, etc.

[0095] Step 106: Input the category features of each riding frequency category and the first target historical riding card usage frequency corresponding to each riding frequency category into the initial riding card usage frequency prediction model to obtain the expected riding card usage frequency corresponding to each riding frequency category.

[0096] In the embodiments of the present application, the initial riding card usage frequency prediction model is used to predict the expected riding card usage frequencies of each user account in a riding frequency category covering a certain historical riding frequency range based on the first target historical riding card usage frequency, that is, the riding card usage frequencies that each user account may have within the time period to be predicted. The embodiments of the present application do not specifically limit the structure of the initial riding card usage frequency prediction model, and any model is applicable to the embodiments of the present application. Through the initial riding card usage frequency prediction model, the expected riding card usage frequency corresponding to each riding frequency category can be predicted.

[0097] Step 108, adjust the initial riding card usage frequency prediction model according to each expected riding card usage frequency and each target current riding card usage frequency to obtain a trained riding card usage frequency prediction model.

[0098] In the embodiments of the present application, based on the expected riding card usage frequencies corresponding to each riding frequency category predicted by the initial riding card usage frequency prediction model and the target current riding card usage frequencies actually corresponding to each riding frequency category within the time period to be predicted, the loss of the initial riding card usage frequency prediction model can be calculated. The embodiments of the present application do not specifically limit the loss function used for calculating the loss, and any loss function is applicable to the embodiments of the present application. Adjust the parameters of the initial riding card usage frequency prediction model according to the loss, and re - perform prediction and loss calculation based on the adjusted initial riding card usage frequency prediction model until the loss is less than the loss threshold, then a trained riding card usage frequency prediction model can be obtained.

[0099] The training method of the riding card usage frequency prediction model provided by the embodiment of the present application divides multiple riding times categories according to the historical riding times, comprehensively obtains the category features of each user account in the riding times category, the first target historical riding card usage frequency and the target current riding card usage frequency of the riding times category, and trains the model through the category features, the first target historical riding card usage frequency and the target current riding card usage frequency. Furthermore, in the application process, the model is used to predict the riding card usage frequency of each riding times category. Since the riding card usage frequency of a single user account fluctuates greatly, using the data of a single user account as a sample to train the model may lead to poor model accuracy. Therefore, the embodiment of the present application aggregates user accounts according to the historical riding times of the user accounts, and transforms the problem of predicting the relationship between the historical riding card usage frequency, the riding times and the future riding card usage frequency of a user account into the problem of predicting the relationship between the historical riding card usage frequency and the future riding card usage frequency of a user group with a historical riding times within a certain range. That is to say, when predicting the future riding card usage frequency of a user, the features of other users with the same historical riding times as the user are referred to. Therefore, the trained model performs better in predicting the overall future riding card usage frequency of a certain user group, and can improve the prediction accuracy of the riding card usage frequency.

[0100] In one embodiment, as Figure 4 shown, a resource acquisition value prediction method is provided. The embodiment of the present application takes the application of this method to a server as an example for description, and includes the following steps:

[0101] Step 402: Divide at least one riding times category according to the historical riding times corresponding to each user account, and determine the user accounts corresponding to each riding times category;

[0102] Step 404: For any riding times category, determine the category features of the riding times category according to the account features of each target user account corresponding to the riding times category, and determine the first target historical riding card usage frequency according to the historical riding card usage frequencies of each target user account;

[0103] Step 406: Input the category features of each riding times category and the first target historical riding card usage frequency corresponding to each riding times category into the riding card usage frequency prediction model to obtain the expected riding card usage frequency corresponding to each riding times category;

[0104] Step 408: Determine the expected riding card resource acquisition value corresponding to each riding times category according to the expected riding card usage frequency corresponding to each riding times category;

[0105] Among them, the riding card usage frequency prediction model is the trained riding card usage frequency prediction model in any of the foregoing embodiments.

[0106] In the embodiments of the present application, the trained cycling card usage frequency prediction model in the foregoing embodiments can be used in actual applications to predict the expected cycling card usage frequency. First, the user account for which the expected cycling card usage frequency needs to be predicted can be obtained, the cycling times categories can be divided according to the method in the foregoing embodiments, the user account can be assigned to each cycling times category, and then the category features and the first target historical cycling card usage frequency can be extracted for each cycling times category. For each cycling times category, the historical cycling times range covered by the cycling times category, the category features, and the first target historical cycling card usage frequency are input into the cycling card usage frequency prediction model to obtain the expected cycling card usage frequency corresponding to the cycling times category.

[0107] The expected cycling card resource acquisition value refers to the resources that the platform can obtain from the user's single use of the cycling card (the resources can refer to currency, virtual currency, and all other valuable items). This resource value is related to the frequency of the user using the cycling card. For example, for each cycling times category, the total resource value that the platform can obtain from the user's single cycling can be taken as the resource value that the user needs to pay for a single cycling without using the card, and then the product of the resource value that the platform needs to pay when the user uses the card for a single cycling and the frequency of the user using the cycling card can be taken as the resource value that the platform needs to pay for the user's single cycling with the card. The difference between the total resource value and the resource value that the platform needs to pay for the user's cycling with the card is taken as the resource acquisition value of the platform for the user's single cycling corresponding to this cycling times category.

[0108] In one example, since the platform generally attaches a resource compensation rate to the cycling card when providing it to the user (that is, the proportion of the resource value provided by the platform to the user after the user holds the cycling card to the resource value that the user needs to pay for holding the cycling card), therefore, considering the resource compensation rate of the cycling card, for any cycling times category, the expected cycling card resource acquisition value corresponding to the cycling times category can be determined according to the resource compensation rate of the cycling card, the expected cycling card usage frequency, and the single - time resource acquisition value of the cycling card (see formula (1)):

[0109]

[0110] Formula (1)

[0111] Among them, gain is the expected value of obtaining cycling card resources corresponding to the category of cycling times. card_sell_amount is the value of resources that need to be paid when the user holds a cycling card, which is equal to the product of the number of times the cycling card can be used (card_count), the value of resources obtained per use of the cycling card (unit_price), and the resource compensation rate of the cycling card (discount). card_counsume_amount is the value of resources lost by the platform when the user uses the cycling card, which is equal to the product of the value of resources obtained per use of the cycling card (unit_price) and the number of times the user uses the cycling card (card_counsume_count). Therefore, according to the derivation of formula (1), the expected value of obtaining cycling card resources is equal to the difference between the value of resources that need to be paid when the user holds a cycling card and the value of resources lost by the platform when the user uses the cycling card, divided by the number of times the user uses the cycling card. It is also equal to the product of the resource compensation rate of the cycling card divided by the expected cycling card usage frequency minus 1, and the value of resources obtained per use of the cycling card. After obtaining the expected cycling card usage frequency, the expected value of obtaining cycling card resources can be calculated according to the product of the resource compensation rate of the cycling card divided by the expected cycling card usage frequency minus 1, and the value of resources obtained per use of the cycling card.

[0112] The resource acquisition value prediction method provided by the embodiment of the present application divides multiple cycling times categories according to historical cycling times, comprehensively obtains the category characteristics and the first target historical cycling card usage frequency of this cycling times category for each user account in this cycling times category, and predicts the cycling card usage frequency of each cycling times category by inputting the category characteristics and the first target historical cycling card usage frequency into the model. Furthermore, the expected value of obtaining cycling card resources can be calculated according to the cycling card usage frequency. The embodiment of the present application predicts the cycling card usage frequency through a model that can predict the relationship between the historical cycling card usage frequency of a user group with a certain range of historical cycling times and the future cycling card usage frequency. It is equivalent to referring to the characteristics of other users with the same historical cycling times as the user when predicting the future cycling card usage frequency of the user. Therefore, the model used has a high prediction accuracy for the cycling card usage frequency and can improve the prediction accuracy of the resource acquisition value.

[0113] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0114] Based on the same inventive concept, an embodiment of the present application further provides a training device for a riding card usage frequency prediction model for implementing the training method of the riding card usage frequency prediction model involved above. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the training device for the riding card usage frequency prediction model provided below can refer to the limitations on the training method of the riding card usage frequency prediction model in the above text, and will not be repeated here.

[0115] In one embodiment, as Figure 5 shown, a training device 500 for a riding card usage frequency prediction model is provided, including: a division module 502, a determination module 504, an input module 506, and an adjustment module 508, where:

[0116] The division module 502 is configured to divide at least one riding times category according to the historical riding times corresponding to each user account, and determine the user accounts corresponding to each of the riding times categories;

[0117] The determination module 504 is configured to, for any one of the riding times categories, determine the category feature of the riding times category according to the account features of each target user account corresponding to the riding times category, determine the first target historical riding card usage frequency according to the historical riding card usage frequencies of each target user account, and determine the target current riding card usage frequency according to the current riding card usage frequencies of each target user account;

[0118] The input module 506 is configured to input the category features of each of the riding times categories and the first target historical riding card usage frequencies corresponding to each of the riding times categories into the initial riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding times categories;

[0119] An adjustment module 508, configured to adjust the initial riding card usage frequency prediction model according to each of the expected riding card usage frequencies and each of the target current riding card usage frequencies, so as to obtain a trained riding card usage frequency prediction model.

[0120] The training device for the riding card usage frequency prediction model provided by the embodiments of the present application divides a plurality of riding times categories according to historical riding times, comprehensively obtains category features, the first target historical riding card usage frequency of the riding times category, and the target current riding card usage frequency for each user account in the riding times category, and trains a model through the category features, the first target historical riding card usage frequency, and the target current riding card usage frequency. Furthermore, during the application process, the riding card usage frequency of each riding times category is predicted through this model. Since the riding card usage frequency of a single user account fluctuates greatly, using the data of a single user account as a sample to train the model may result in poor model accuracy. Therefore, the embodiments of the present application aggregate user accounts according to the historical riding times of the user accounts, and convert the problem of predicting the relationship between the historical riding card usage frequency, the riding times, and the future riding card usage frequency of a user account into the problem of predicting the relationship between the historical riding card usage frequency and the future riding card usage frequency of a group of users with historical riding times within a certain range. That is to say, when predicting the future riding card usage frequency of a user, the characteristics of other users with the same historical riding times as the user are referred to. Therefore, the trained model performs better in predicting the overall future riding card usage frequency of a certain user group, and can improve the prediction accuracy of the riding card usage frequency.

[0121] In one of the embodiments, the dividing module 502 is further configured to:

[0122] Obtain the historical riding times corresponding to each user account, and respectively determine the number of user accounts corresponding to each of the historical riding times;

[0123] Divide at least one riding times category according to each of the historical riding times and the number of user accounts corresponding to each of the historical riding times, so that the number of user accounts corresponding to each of the riding times categories respectively meets a preset condition.

[0124] In one of the embodiments, the dividing module 502 is further configured to:

[0125] For any one of the historical riding times, determine the second target historical riding card usage frequency corresponding to the historical riding times according to the historical riding card usage frequency of the user account corresponding to the historical riding times;

[0126] Determine the change rate of the usage frequency of the second target historical cycling card with respect to the historical cycling times according to each of the historical cycling times and each of the usage frequencies of the second target historical cycling card;

[0127] Divide at least one cycling times category according to each of the historical cycling times, the change rate, and the number of the user accounts corresponding to each of the historical cycling times, so that for any one of the cycling times categories, the change rate is less than a change rate threshold within the range of the historical cycling times corresponding to the cycling times category, and the number of each of the user accounts corresponding to each of the cycling times categories meets a preset condition.

[0128] In one embodiment, the determining module 504 is further configured to:

[0129] Determine each target user account corresponding to the cycling times category;

[0130] For any preset account feature category, determine the dispersion degree of the target account features of each of the target user accounts belonging to the preset account feature category, and when the dispersion degree is less than a preset dispersion degree threshold, determine a first target account feature according to each of the target account features, and use the first target account feature as the category feature of the cycling times category corresponding to the preset account feature category; or when the dispersion degree is greater than the preset dispersion degree threshold, perform clustering processing on each of the target user accounts according to the target account features, determine a second target account feature according to the number of the target user accounts in each of the target user account categories obtained after clustering, and use the second target account feature as the category feature of the cycling times category corresponding to the preset account feature category.

[0131] In one embodiment, the determining module 504 is further configured to:

[0132] According to the number of the target user accounts in each of the target user account categories obtained after clustering and the total number of each of the target user accounts, respectively determine the proportion of the account number corresponding to each of the target user account categories, and use each of the account number proportions as the second target account feature.

[0133] In one embodiment, as Figure 6 shown, a resource acquisition value prediction device 600 is provided, including: a dividing module 602, a first determining module 604, an input module 606, and a second determining module 608, wherein:

[0134] The dividing module 602 is configured to divide at least one cycling times category according to the historical cycling times corresponding to each user account, and determine the user accounts corresponding to each cycling times category;

[0135] A first determination module 604, configured to, for any one of the riding frequency categories, determine the category feature of the riding frequency category according to the account features of each target user account corresponding to the riding frequency category, and determine a first target historical riding card usage frequency according to the historical riding card usage frequencies of each target user account;

[0136] An input module 606, configured to input the category features of each riding frequency category and the first target historical riding card usage frequencies respectively corresponding to each riding frequency category into a riding card usage frequency prediction model, to obtain the expected riding card usage frequencies respectively corresponding to each riding frequency category;

[0137] A second determination module 608, configured to determine the expected riding card resource acquisition values corresponding to each riding frequency category according to the expected riding card usage frequencies corresponding to each riding frequency category;

[0138] Wherein, the riding card usage frequency prediction model is the trained riding card usage frequency prediction model described in any one of the foregoing embodiments.

[0139] The resource acquisition value prediction device provided by the embodiment of the present application divides multiple riding frequency categories according to historical riding frequencies, comprehensively obtains the category features and the first target historical riding card usage frequency of the riding frequency category for the account features of each user account in the riding frequency category, and predicts the riding card usage frequency of each riding frequency category by inputting the category features and the first target historical riding card usage frequency into the model, and then can calculate the expected riding card resource acquisition value according to the riding card usage frequency. The embodiment of the present application predicts the riding card usage frequency through a model that can predict the relationship between the historical riding card usage frequency and the future riding card usage frequency of a user group with a certain range of historical riding frequencies. It is equivalent to referring to the features of other users with the same historical riding frequency as the user when predicting the future riding card usage frequency of the user. Therefore, the model used has a higher prediction accuracy for the riding card usage frequency, and can improve the prediction accuracy of the resource acquisition value.

[0140] In one embodiment, the second determination module 608 is further configured to:

[0141] For any one of the riding frequency categories, determine the expected riding card resource acquisition value corresponding to the riding frequency category according to the riding card resource compensation rate, the expected riding card usage frequency, and the single - time resource acquisition value of the riding card.

[0142] Each module in the above device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0143] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for training a riding card usage frequency prediction model.

[0144] Those skilled in the art can understand that Figure 7 the structure shown in

[0145] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0146] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, it implements the steps in each of the above method embodiments.

[0147] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the steps in each of the above method embodiments.

[0148] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties.

[0149] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0150] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0151] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A training method for a cycling card usage frequency prediction model, characterized in that: The method comprises: According to the historical riding times corresponding to each user account, at least one riding times category is divided, and the user account corresponding to each riding times category is determined; For any of the riding times categories, the category characteristics of the riding times category are determined according to the account characteristics of each target user account corresponding to the riding times category, a first target historical riding card usage frequency is determined according to the historical riding card usage frequency of each target user account, and a target current riding card usage frequency is determined according to the current riding card usage frequency of each target user account; Input the category features of each of the riding times categories and the first target historical riding card usage frequencies corresponding to each of the riding times categories into an initial riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding times categories; According to each of the expected riding card usage frequencies and each of the target current riding card usage frequencies, the initial riding card usage frequency prediction model is adjusted to obtain a trained riding card usage frequency prediction model.

2. The method according to claim 1, characterized in that The method of dividing at least one riding times category according to the historical riding times corresponding to each user account includes: Obtain the historical riding times corresponding to each user account, and respectively determine the number of user accounts corresponding to each of the historical riding times; At least one riding number category is divided according to each of the historical riding numbers and the number of the user accounts corresponding to each of the historical riding numbers, so that the number of the user accounts corresponding to each of the riding number categories meets a preset condition.

3. The method according to claim 2, characterized in that The dividing at least one riding times category according to each of the historical riding times and the number of the user accounts corresponding to each of the historical riding times includes: For any of the historical riding times, determining a second target historical riding card usage frequency corresponding to the historical riding times according to the historical riding card usage frequency of the user account corresponding to the historical riding times; Determine, according to each of the historical riding times and each of the second target historical riding card usage frequencies, a rate of change of the second target historical riding card usage frequency with respect to the historical riding times; At least one riding number category is divided according to each of the historical riding numbers, the change rate and the number of user accounts corresponding to each of the historical riding numbers, so that for any of the riding number categories, the change rate in the historical riding number range corresponding to the riding number category is less than a change rate threshold, and the number of user accounts corresponding to each of the riding number categories respectively meets a preset condition.

4. The method according to claim 1, characterized in that The determining the category characteristics of the riding times category according to the account characteristics of each target user account corresponding to the riding times category includes: Determine each target user account corresponding to the riding frequency category; For any preset account feature category, determine the degree of dispersion of the target account features of each target user account belonging to the preset account feature category, and when the degree of dispersion is less than the preset discrete degree threshold, determine the first target account feature according to each target account feature, and use the first target account feature as the category feature of the preset account feature category corresponding to the number of rides category; or when the degree of dispersion is greater than the preset discrete degree threshold, cluster each target user account according to the target account feature, determine the second target account feature according to the number of target user accounts in each target user account category obtained after clustering, and use the second target account feature as the category feature of the preset account feature category corresponding to the number of rides category.

5. The method according to claim 4, characterized in that The determining the second target account feature according to the number of target user accounts in each target user account category obtained after clustering includes: According to the number of target user accounts in each target user account category obtained after clustering and the total number of target user accounts, the proportion of the number of accounts corresponding to each target user account category is determined respectively, and the proportion of the number of accounts is used as the second target account feature.

6. A resource acquisition value prediction method, characterized in that: The method comprises: According to the historical riding times corresponding to each user account, at least one riding times category is divided, and the user account corresponding to each riding times category is determined; For any of the riding times categories, determine the category characteristics of the riding times category according to the account characteristics of each target user account corresponding to the riding times category, and determine the first target historical riding card usage frequency according to the historical riding card usage frequency of each target user account; Inputting the category features of each of the riding times categories and the first target historical riding card usage frequencies corresponding to each of the riding times categories into a riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding times categories; Determine the expected riding card resource acquisition value corresponding to each riding number category according to the expected riding card usage frequency corresponding to each riding number category; The cycling card usage frequency prediction model is a trained cycling card usage frequency prediction model according to any one of claims 1 to 5.

7. The method according to claim 6, characterized in that The determining, according to the expected riding card usage frequency corresponding to each riding number category, the expected riding card resource acquisition value corresponding to each riding number category, includes: For any of the riding times categories, the expected riding card resource acquisition value corresponding to the riding times category is determined according to the riding card resource compensation rate, the expected riding card usage frequency and the riding card single resource acquisition value.

8. A training device for a cycling card usage frequency prediction model, characterized in that: The device comprises: A classification module, used for classifying at least one riding frequency category according to the historical riding frequency corresponding to each user account, and determining the user account corresponding to each riding frequency category; a determination module, for determining, for any of the riding number categories, a category characteristic of the riding number category according to the account characteristics of each target user account corresponding to the riding number category, determining a first target historical riding card usage frequency according to the historical riding card usage frequency of each target user account, and determining a target current riding card usage frequency according to the current riding card usage frequency of each target user account; An input module, used for inputting the category characteristics of each of the riding frequency categories and the first target historical riding card usage frequencies corresponding to each of the riding frequency categories into an initial riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding frequency categories; The adjustment module is used to adjust the initial riding card usage frequency prediction model according to each of the expected riding card usage frequencies and each of the target current riding card usage frequencies to obtain a trained riding card usage frequency prediction model.

9. A resource acquisition value prediction device, characterized in that: The device comprises: A classification module, used for classifying at least one riding frequency category according to the historical riding frequency corresponding to each user account, and determining the user account corresponding to each riding frequency category; A first determination module is used to determine, for any of the riding times categories, the category characteristics of the riding times category according to the account characteristics of each target user account corresponding to the riding times category, and determine a first target historical riding card usage frequency according to the historical riding card usage frequency of each target user account; An input module, used for inputting the category characteristics of each of the riding number categories and the first target historical riding card usage frequencies corresponding to each of the riding number categories into a riding card usage frequency prediction model to obtain the expected riding card usage frequencies corresponding to each of the riding number categories; A second determination module is used to determine the expected riding card resource acquisition value corresponding to each riding number category according to the expected riding card usage frequency corresponding to each riding number category; The cycling card usage frequency prediction model is a trained cycling card usage frequency prediction model according to any one of claims 1 to 5.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images