Account recommendation model training method, account recommendation method, device and apparatus

Abstract

The embodiment of the application provides a kind of account recommendation model training method, account recommendation method, it is related to artificial intelligence technical field.The account recommendation model training method includes: obtaining target sample resource object, and the multiple first recommended sample account associated with target sample resource object, and the target sample resource object is input to the general account recommendation model that is pre-trained, and then, the predicted resource transfer tendency degree of each first recommended sample account is obtained for target sample resource object;To obtain the predicted standard deviation corresponding to target sample resource object according to the predicted resource transfer tendency degree;Finally, based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced by the predicted standard deviation, and the special account recommendation model of target sample resource object is obtained.In the method, the accuracy of account recommendation can be improved by determining the special account recommendation model.

Description

Technical Field

[0001] This application relates to the Internet field, and in particular to a training method for an account recommendation model, an account recommendation method, an apparatus, a system, a computer device, a storage medium, and a computer program product. Background Technology

[0002] With the continuous development of Internet technology, data mining technology is being used more and more widely in various industries. Data mining is the process of extracting hidden, unprecedented, and beneficial relationships, patterns, and trends from datasets or massive amounts of data, and using the extracted knowledge and rules to build models for decision support, providing predictive decision support methods, tools, and processes.

[0003] Currently, in the process of mining relevant data for target accounts (e.g., users) of target resource objects, the general approach is to analyze user behavior and then recommend resource objects to users based on user characteristics. This approach can lead to a problem of fewer resource object recommendations and lower transaction volumes. Therefore, current account recommendation technologies suffer from a significant problem of low transaction volumes for a considerable number of resource objects. Summary of the Invention

[0004] Therefore, it is necessary to provide a training method, account recommendation method, device, system, computer equipment, storage medium, and computer program product for an account recommendation model to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for training an account recommendation model. The method includes:

[0006] Obtain a pre-trained general account recommendation model;

[0007] Obtain a target sample resource object, and obtain multiple first sample accounts to be recommended associated with the target sample resource object; the first sample accounts to be recommended are accounts that have potential preset behaviors for the target sample resource object;

[0008] The target sample resource object and the plurality of first sample accounts to be recommended are input into the general account recommendation model to obtain the predicted resource transfer tendency of each first sample account to be recommended in relation to the target sample resource object; the predicted resource transfer tendency represents the predicted probability that the first sample account to be recommended will perform a preset behavior in relation to the target sample resource object;

[0009] Based on the predicted resource transfer tendency of the target sample resource object, obtain the predicted standard deviation corresponding to the target sample resource object;

[0010] Based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object; the initial standard deviation is a preset fixed value of the standard deviation in the general account recommendation model.

[0011] In one embodiment, obtaining the target sample resource object includes: obtaining multiple candidate sample resource objects and multiple second recommended sample accounts associated with each candidate sample resource object; the candidate sample resource objects are resource objects to be transferred in a financial business system; the second recommended sample accounts are accounts with potential preset behaviors towards the candidate sample resource objects; inputting the candidate sample resource objects and the multiple second recommended sample accounts associated with each candidate sample resource object into the general account recommendation model to obtain the predicted resource transfer tendency of each second recommended sample account towards the candidate sample resource object; and obtaining the target sample resource object from the multiple candidate sample resource objects based on the predicted resource transfer tendency of the candidate sample resource objects.

[0012] In one embodiment, obtaining the target sample resource object from the plurality of candidate sample resource objects based on the predicted resource transfer tendency of the candidate sample resource objects includes: obtaining the target sample account with the highest predicted resource transfer tendency from the plurality of second-to-be-recommended sample accounts based on the predicted resource transfer tendency of the candidate sample resource objects; the target sample account is the sample account with the highest predicted resource transfer tendency among the plurality of second-to-be-recommended sample accounts; obtaining the prediction error of the candidate sample resource object based on the predicted resource transfer tendency of the target sample account for the candidate sample resource object; obtaining the average prediction error of each candidate sample resource object based on the prediction error corresponding to each candidate sample resource object; and determining the candidate sample resource object whose prediction error is less than the average prediction error as the target sample resource object.

[0013] In one embodiment, obtaining the prediction error of the candidate sample resource object based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource object includes: obtaining actual account feature information corresponding to the target sample account, and obtaining predicted account feature information corresponding to the target sample account based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource object; and determining the difference between the actual account feature information and the predicted account feature information as the prediction error.

[0014] In one embodiment, the step of inputting the target sample resource object and the plurality of first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object includes: extracting object feature information of the target sample resource object and account feature information corresponding to each of the first recommended sample accounts; inputting the object feature information and the account feature information into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0015] In one embodiment, obtaining a pre-trained general account recommendation model includes: obtaining multiple training sample resource objects and multiple third recommendation sample accounts associated with each training sample resource object; extracting training object feature information of the training sample resource objects and training account feature information of the third recommendation sample accounts associated with the training sample resource objects, and obtaining the actual resource transfer tendency corresponding to each of the third recommendation sample accounts corresponding to the training object feature information; inputting the training object feature information and the training account feature information into the general account recommendation model to be trained to obtain the predicted resource transfer tendency of each of the third recommendation sample accounts for the training sample resource objects; and training the general account recommendation model based on the difference between the predicted resource transfer tendency of the training sample resource objects and the actual resource transfer tendency.

[0016] Secondly, this application provides an account recommendation method. The method includes:

[0017] Obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended.

[0018] The resource object to be recommended and the plurality of candidate accounts to be recommended are input into the dedicated account recommendation model of the resource object to be recommended, so as to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in the first aspect;

[0019] Accounts with a predicted resource transfer tendency greater than a preset transfer tendency threshold are selected from the candidate recommended accounts and used as the target recommended accounts corresponding to the recommended resource objects.

[0020] Thirdly, this application provides a training apparatus for an account recommendation model. The apparatus includes:

[0021] The training module is used to obtain a pre-trained general account recommendation model;

[0022] The acquisition module is used to acquire a target sample resource object and acquire multiple first sample accounts to be recommended associated with the target sample resource object; the first sample accounts to be recommended are accounts that have potential preset behaviors for the target sample resource object;

[0023] The first calculation module is used to input the target sample resource object and the plurality of first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0024] The second calculation module is used to obtain the predicted standard deviation corresponding to the target sample resource object based on the predicted resource transfer tendency for the target sample resource object.

[0025] The adjustment module is used to replace the initial standard deviation in the general account recommendation model with the predicted standard deviation based on the predicted standard deviation, so as to obtain the special account recommendation model for the target sample resource object; the initial standard deviation is a preset fixed value of the standard deviation in the general account recommendation model.

[0026] Fourthly, this application provides an account recommendation device. The device includes:

[0027] The acquisition module is used to acquire the resource object to be recommended, and to acquire multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended.

[0028] The calculation module is used to input the resource object to be recommended and the plurality of candidate accounts to be recommended into the dedicated account recommendation model of the resource object to be recommended, and to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in the first aspect;

[0029] The filtering module is used to obtain accounts from the candidate recommended accounts whose predicted resource transfer tendency is greater than a preset transfer tendency threshold, and use them as target recommended accounts corresponding to the recommended resource objects.

[0030] Fifthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0031] Obtain a pre-trained general account recommendation model;

[0032] Obtain a target sample resource object and a plurality of first sample accounts to be recommended associated with the target sample resource object; the target sample resource object is an object in the financial business system that has a resource transfer requirement; the first sample accounts to be recommended are accounts that have potential preset behaviors toward the target sample resource object;

[0033] The target sample resource object and the plurality of first sample accounts to be recommended are input into the general account recommendation model to obtain the predicted resource transfer tendency of each first sample account to be recommended for the target sample resource object;

[0034] Based on the predicted resource transfer tendency, obtain the predicted standard deviation corresponding to the target sample resource object;

[0035] Based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object.

[0036] And / or, obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended;

[0037] The resource object to be recommended and the plurality of candidate accounts to be recommended are input into the dedicated account recommendation model of the resource object to be recommended, so as to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in the first aspect;

[0038] Accounts with a predicted resource transfer tendency greater than a preset transfer tendency threshold are selected from the candidate recommended accounts and used as the target recommended accounts corresponding to the recommended resource objects.

[0039] Sixthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0040] Obtain a pre-trained general account recommendation model;

[0041] Obtain a target sample resource object and a plurality of first sample accounts to be recommended associated with the target sample resource object; the target sample resource object is an object in the financial business system that has a resource transfer requirement; the first sample accounts to be recommended are accounts that have potential preset behaviors toward the target sample resource object;

[0042] The target sample resource object and the plurality of first sample accounts to be recommended are input into the general account recommendation model to obtain the predicted resource transfer tendency of each first sample account to be recommended for the target sample resource object;

[0043] Based on the predicted resource transfer tendency, obtain the predicted standard deviation corresponding to the target sample resource object;

[0044] Based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object.

[0045] And / or, obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended;

[0046] The resource object to be recommended and the plurality of candidate accounts to be recommended are input into the dedicated account recommendation model of the resource object to be recommended, so as to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in the first aspect;

[0047] Accounts with a predicted resource transfer tendency greater than a preset transfer tendency threshold are selected from the candidate recommended accounts and used as the target recommended accounts corresponding to the recommended resource objects.

[0048] Seventhly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0049] Obtain a pre-trained general account recommendation model;

[0050] Obtain a target sample resource object and a plurality of first sample accounts to be recommended associated with the target sample resource object; the target sample resource object is an object in the financial business system that has a resource transfer requirement; the first sample accounts to be recommended are accounts that have potential preset behaviors toward the target sample resource object;

[0051] The target sample resource object and the plurality of first sample accounts to be recommended are input into the general account recommendation model to obtain the predicted resource transfer tendency of each first sample account to be recommended for the target sample resource object;

[0052] Based on the predicted resource transfer tendency, obtain the predicted standard deviation corresponding to the target sample resource object;

[0053] Based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object.

[0054] And / or, obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended;

[0055] The resource object to be recommended and the plurality of candidate accounts to be recommended are input into the dedicated account recommendation model of the resource object to be recommended, so as to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in the first aspect;

[0056] Accounts with a predicted resource transfer tendency greater than a preset transfer tendency threshold are selected from the candidate recommended accounts and used as the target recommended accounts corresponding to the recommended resource objects.

[0057] In the aforementioned training method, account recommendation method, device, system, computer equipment, storage medium, and computer program product for the account recommendation model, a target sample resource object and multiple first-to-be-recommended sample accounts associated with the target sample resource object can be obtained. The target sample resource object and the multiple first-to-be-recommended sample accounts are then input into a pre-trained general account recommendation model. This yields the predicted resource transfer tendency of each first-to-be-recommended sample account for the target sample resource object. Based on the predicted resource transfer tendency, the predicted standard deviation corresponding to the target sample resource object is obtained. Finally, based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain a dedicated account recommendation model for the target sample resource object. The account recommendation model acquisition method provided in this application embodiment can adaptively adjust the general account recommendation model by calculating the predicted standard deviation of the target sample resource object to obtain a dedicated account recommendation model suitable for the target sample resource object. This can improve the accuracy of account recommendations and thus alleviate the problems of uneven recommendations, insufficient recommendations for some resource objects, and low transaction volumes. Attached Figure Description

[0058] Figure 1 A flowchart illustrating a training method for an account recommendation model provided in one embodiment;

[0059] Figure 2 A schematic diagram of the process for obtaining a target sample resource object is provided for one embodiment;

[0060] Figure 3 A schematic diagram illustrating the process of obtaining a target sample resource object from multiple candidate sample resource objects, provided as an embodiment;

[0061] Figure 4 A flowchart illustrating an account recommendation method provided in one embodiment;

[0062] Figure 5 A flowchart illustrating a training method for an account recommendation model, provided for another embodiment;

[0063] Figure 6 A structural block diagram of a training device for an account recommendation model provided in one embodiment;

[0064] Figure 7 A structural block diagram of an account recommendation device provided in one embodiment;

[0065] Figure 8 An internal structural diagram of a computer device provided for one embodiment. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit the scope of this application. In one embodiment, such as... Figure 1 As shown, a training method for an account recommendation model is provided. This embodiment illustrates the method by applying it to a server. It is understood that this method can also be applied to a terminal, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0067] Step S101: Obtain the pre-trained general account recommendation model.

[0068] The general account recommendation model can be applied to all objects within a financial business system that have resource transfer needs. Taking a target resource object as an example, the general account recommendation model is used to obtain the resource transfer inclination of each recommended account associated with the target resource object for that target resource object based on the label distribution relationship between the target resource object and the associated recommended accounts. Furthermore, it can also obtain the account with the highest resource transfer inclination for that target resource object. The target resource object can be a resource object within the financial business system that has resource transfer needs; the recommended accounts associated with the target resource object can be users who have a resource transfer inclination for the target resource object, or users with potential pre-defined behaviors towards the target resource object, such as users with transaction records related to the target resource object in the financial business system, and users who have browsed or consulted about the target resource object; the resource transfer inclination can be the probability of performing a resource transfer, such as the transaction intention.

[0069] Step S102: Obtain the target sample resource object and the multiple first sample accounts to be recommended associated with the target sample resource object.

[0070] Among them, the target sample resource object is the resource object to be transferred in the financial business system whose prediction error is less than the average prediction error; the first sample account to be recommended is an account with potential preset behavior for the target sample resource object.

[0071] Step S103: Input the target sample resource object and multiple first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0072] First, feature extraction can be performed on the target sample resource object to obtain its object feature information. Feature extraction can then be performed on multiple first-to-be-recommended sample accounts to obtain their respective account feature information. Second, the object feature information and account feature information are input into a general account recommendation model to obtain the predicted resource transfer tendency of each first-to-be-recommended sample account towards the target sample resource object. This predicted resource transfer tendency characterizes the predicted probability that the first-to-be-recommended sample account will perform a preset behavior towards the target sample resource object, such as a transaction probability. This preset behavior could be a resource transfer action towards the target sample resource object, such as trading or purchasing the target sample resource object.

[0073] Step S104: Obtain the predicted standard deviation of the target sample resource object based on the predicted resource transfer tendency of the target sample resource object.

[0074] In this step, the standard deviation of each sample object in the pre-trained general account recommendation model is a preset fixed value. Based on the predicted resource transfer tendency of the target sample resource object, the account with the highest resource tendency for the target sample resource object can be obtained. Then, based on the feature information of the account with the highest resource tendency, the predicted standard deviation corresponding to the target sample resource object can be obtained. This predicted standard deviation is applicable to the target sample resource object and is used to represent the distribution of multiple first-to-be-recommended sample accounts associated with the target sample resource object. The larger the predicted standard deviation, the more dispersed the distribution of multiple first-to-be-recommended sample accounts; the smaller the predicted standard deviation, the more concentrated the distribution of multiple first-to-be-recommended sample accounts is near the account with the highest resource tendency.

[0075] Step S105: Based on the predicted standard deviation, replace the initial standard deviation in the general account recommendation model with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object.

[0076] The initial standard deviation is a preset fixed value for each sample object. This dedicated account recommendation model is an account recommendation model applicable to target sample resource objects. It is used to obtain the resource transfer tendency of each first-to-be-recommended sample account associated with the target sample resource object for the target sample resource object based on the label distribution relationship between the target sample resource object and the associated first-to-be-recommended sample account. In turn, it can also obtain the account with the highest resource transfer tendency for the target sample resource object, which is more targeted and has higher accuracy in account recommendation.

[0077] In this embodiment, the method obtains a target sample resource object and multiple first-to-be-recommended sample accounts associated with the target sample resource object. The target sample resource object and the multiple first-to-be-recommended sample accounts are then input into a pre-trained general account recommendation model. This yields the predicted resource transfer tendency of each first-to-be-recommended sample account for the target sample resource object. Based on the predicted resource transfer tendency, the predicted standard deviation corresponding to the target sample resource object is obtained. Finally, based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain a dedicated account recommendation model for the target sample resource object. The account recommendation model acquisition method provided in this embodiment can adaptively adjust the general account recommendation model by calculating the predicted standard deviation of the target sample resource object to obtain a dedicated account recommendation model suitable for the target sample resource object. This improves the accuracy of account recommendations and addresses the problems of uneven recommendations, insufficient recommendations for some resource objects, and low transaction volumes.

[0078] In some embodiments, such as Figure 2 As shown, obtaining the target sample resource object in step S102 may include:

[0079] Step S201: Obtain multiple candidate sample resource objects, and obtain multiple second sample accounts to be recommended associated with each candidate sample resource object.

[0080] The candidate sample resource objects are resource objects to be transferred in the financial business system; the second recommended sample accounts are accounts with potential pre-defined behaviors towards the candidate sample resource objects. The candidate sample resource objects can be resource objects to be transferred obtained from the financial business system without precise screening. The pre-defined behavior can be a resource transfer action towards the candidate sample resource objects, such as trading or purchasing the candidate sample resource objects.

[0081] Step S202: Input the candidate sample resource object and the multiple second recommended sample accounts associated with the candidate sample resource object into the general account recommendation model to obtain the predicted resource transfer tendency of each second recommended sample account for the candidate sample resource object.

[0082] First, feature extraction can be performed on candidate sample resource objects to obtain their object feature information. Feature extraction can then be performed on multiple second-to-be-recommended sample accounts to obtain their account feature information. Second, this object feature information and the account feature information are input into a general account recommendation model to obtain the predicted resource transfer tendency of each second-to-be-recommended sample account towards the candidate sample resource object. This predicted resource transfer tendency is the predicted probability that each second-to-be-recommended sample account will perform a preset behavior towards the candidate sample resource object, such as a transaction probability. This preset behavior can be a resource transfer behavior towards the candidate sample resource object, such as trading or purchasing the candidate sample resource object. The object feature information can be the product characteristics of the candidate sample resource object, such as time, amount, monthly rate of return, and risk level. The account feature information can be the characteristics of the potential users associated with the candidate sample resource object, i.e., the second-to-be-recommended sample accounts, related to the candidate sample resource object, such as age, occupation, and income.

[0083] Step S203: Based on the predicted resource transfer tendency of the candidate sample resource objects, obtain the target sample resource object from multiple candidate sample resource objects.

[0084] Based on the predicted resource transfer tendency of candidate sample resource objects, the account with the highest resource transfer tendency for the candidate sample resource objects can be obtained. Furthermore, based on the error of the account with the highest resource transfer tendency among the candidate sample resource objects, multiple preliminary sample resource objects with more accurate account recommendations can be selected from multiple candidate sample resource objects. These preliminary sample resource objects are precisely selected resource objects to be transferred. The target sample resource object is any one of the preliminary sample resource objects.

[0085] In the method of this embodiment, candidate sample resource objects can be screened based on the accuracy of the prediction of the general account recommendation model, i.e. the prediction error, to obtain multiple preliminary sample resource objects with more accurate account recommendations. This can obtain more accurate sample data, which is beneficial for obtaining a dedicated account recommendation model for the target sample resource objects in the future, and helps to improve efficiency and accuracy.

[0086] In some embodiments, such as Figure 3 As shown, step S203 may include:

[0087] Step S301: Based on the predicted resource transfer tendency of the candidate sample resource objects, obtain the target sample account with the highest second resource transfer tendency from multiple second recommended sample accounts.

[0088] The target sample account is the one with the highest predicted resource transfer tendency among multiple second-tier recommended sample accounts. The target sample account can be calculated based on the following formula (1):

[0089]

[0090] Where, x i For the i-th sample object in the candidate sample resource objects, w j For the j-th sample account among multiple second sample accounts to be recommended; α i The mean value corresponding to the i-th sample object corresponds to the sample account with the highest predicted resource transfer tendency, i.e., the target sample account. Let σ be the predicted resource transfer propensity of the j-th sample account to be recommended, associated with the i-th sample object; σ is the initial standard deviation; and Z is the normalization factor. Formula (1) above can be used to determine sample accounts with a predicted resource transfer propensity greater than the propensity threshold, and also to determine target sample accounts.

[0091] Step S302: Based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource object, obtain the prediction error of the candidate sample resource object.

[0092] According to formula (1) in step S301, the mean of the candidate sample resource object can be determined, which is the sample account with the highest predicted resource transfer tendency, i.e., the target sample account. Furthermore, the actual account feature information corresponding to this mean can also be determined in the second sample account to be recommended. Then, the difference between the predicted account feature information and the actual account feature information of the target sample account is determined as the prediction error of the candidate sample resource object. This prediction error is used to characterize the accuracy of predicting the resource transfer tendency of the candidate sample resource object based on the general account recommendation model. The larger the prediction error, the lower the accuracy of the predicted resource transfer tendency of the candidate sample resource object predicted by the general account recommendation model; the smaller the prediction error, the higher the accuracy of the predicted resource transfer tendency of the candidate sample resource object predicted by the general account recommendation model. Therefore, based on the prediction error, it can be determined whether the candidate sample resource object is suitable for the general account recommendation model.

[0093] Step S303: Based on the prediction errors corresponding to each candidate sample resource object, obtain the average prediction error of each candidate sample resource object.

[0094] The average prediction error can be the average of the prediction errors corresponding to each candidate sample resource object.

[0095] Step S304: Candidate sample resource objects whose prediction error is less than the average prediction error are identified as target sample resource objects.

[0096] Among them, candidate sample resource objects with prediction errors less than the average prediction error can be candidate sample resource objects with high accuracy in predicting resource transfer propensity obtained by the general account recommendation model, that is, candidate sample resource objects applicable to the general account recommendation model. These candidate sample resource objects can be resource objects to be transferred that have not undergone precise screening and are obtained from the financial business system. The initial screening sample resource objects are resource objects to be transferred that have undergone precise screening based on prediction errors. The target sample resource object is any one of the initial screening sample resource objects.

[0097] In the method of this embodiment, sample objects can be screened based on prediction error to obtain sample objects suitable for general account recommendation models, which is more conducive to subsequent adaptive adjustment to obtain dedicated account recommendation models and improve the accuracy of dedicated account recommendation models.

[0098] In some embodiments, step S302 may include:

[0099] Obtain the actual account feature information corresponding to the target sample account, and obtain the predicted account feature information corresponding to the target sample account; the difference between the actual account feature information and the predicted account feature information is determined as the prediction error.

[0100] The prediction error can be calculated based on formula (2):

[0101]

[0102] in, The feature information of the target sample account, i.e., the predicted account feature information; α i The feature information of the initial sample accounts, i.e., the feature information of the actual accounts; e i This represents the prediction error for the candidate sample resource object.

[0103] In some embodiments, step S103 may include:

[0104] Extract the object feature information of the target sample resource object and the account feature information corresponding to each first recommended sample account; input the object feature information and account feature information into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0105] First, features can be extracted from the target sample resource object to obtain its object feature information. Then, features can be extracted from multiple first-to-be-recommended sample accounts to obtain their respective account feature information. Second, the object feature information and account feature information are input into a general account recommendation model to obtain the predicted resource transfer tendency of each first-to-be-recommended sample account towards the target sample resource object. This predicted resource transfer tendency characterizes the predicted probability, such as the transaction probability, of the first-to-be-recommended sample account performing a preset behavior towards the target sample resource object.

[0106] In this embodiment, the method can obtain the predicted resource transfer tendency of the target sample resource object based on a trained general account recommendation model, which facilitates the subsequent calculation of the prediction standard deviation. This allows for the acquisition of a dedicated account recommendation model, improving the accuracy of account recommendations and mitigating issues such as uneven recommendations, insufficient recommendations for some resource objects, and low transaction volumes. Furthermore, it facilitates the identification of potential users and the largest potential users for the target sample resource object.

[0107] In some embodiments, step S101 may include:

[0108] Multiple training sample resource objects and multiple third-party recommendation sample accounts associated with each training sample resource object are obtained. The training object feature information of the training sample resource objects and the training account feature information of the associated third-party recommendation sample accounts are extracted. The actual resource transfer tendency corresponding to each third-party recommendation sample account is also obtained based on the training object feature information. The training object feature information and the training account feature information are input into the general account recommendation model to be trained, obtaining the predicted resource transfer tendency of each third-party recommendation sample account for the training sample resource object. Based on the difference between the predicted resource transfer tendency and the actual resource transfer tendency of the training sample resource object, the general account recommendation model to be trained is trained.

[0109] In the method of this embodiment, a well-trained general account recommendation model can be accurately obtained. Then, the general account recommendation model can be adaptively adjusted to obtain a special account recommendation model.

[0110] In one embodiment, such as Figure 4 As shown, an account recommendation method is provided, which may include:

[0111] Step S401: Obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended.

[0112] Among them, the resource objects to be recommended are the resource objects to be transferred in the financial business system; the candidate accounts to be recommended are accounts that have been obtained in advance and have potential preset behaviors for the resource objects to be recommended.

[0113] Step S402: Input the resource object to be recommended and multiple candidate accounts to be recommended into the dedicated account recommendation model of the resource object to be recommended, and obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended.

[0114] Among them, the predicted resource transfer tendency represents the predicted probability that each candidate account to be recommended will perform a preset behavior for the resource object to be recommended.

[0115] Step S403: Obtain accounts from the candidate recommended accounts whose predicted resource transfer tendency is greater than the preset transfer tendency threshold, and use them as the target recommended accounts corresponding to the recommended resource objects.

[0116] In this embodiment, the method obtains a target sample resource object and multiple first-to-be-recommended sample accounts associated with the target sample resource object. The target sample resource object and the multiple first-to-be-recommended sample accounts are then input into a pre-trained general account recommendation model. This yields the predicted resource transfer tendency of each first-to-be-recommended sample account for the target sample resource object. Based on the predicted resource transfer tendency, the predicted standard deviation corresponding to the target sample resource object is obtained. Finally, based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain a dedicated account recommendation model for the target sample resource object. The account recommendation model acquisition method provided in this embodiment can adaptively adjust the general account recommendation model by calculating the predicted standard deviation of the target sample resource object to obtain a dedicated account recommendation model suitable for the target sample resource object. This improves the accuracy of account recommendations and addresses the problems of uneven recommendations, insufficient recommendations for some resource objects, and low transaction volumes.

[0117] In another embodiment, such as Figure 5 As shown, a training method for an account recommendation model is provided, which may include:

[0118] Step S501: Obtain multiple training sample resource objects and multiple third-party recommendation sample accounts associated with each training sample resource object.

[0119] Step S502: Extract the training object feature information of the training sample resource object and the training account feature information of the third recommendation sample account associated with the training sample resource object, and obtain the actual resource transfer tendency corresponding to each third recommendation sample account corresponding to the training object feature information.

[0120] Step S503: Input the feature information of the object to be trained and the feature information of the account to be trained into the general account recommendation model to be trained, and obtain the predicted resource transfer tendency of each third sample account to be recommended for the resource object to be trained.

[0121] Step S504: Based on the difference between the predicted resource transfer tendency and the actual resource transfer tendency of the resource objects in the training sample, train the general account recommendation model to be trained.

[0122] Step S505: Obtain multiple candidate sample resource objects, and obtain multiple second recommended sample accounts associated with each candidate sample resource object.

[0123] The candidate sample resource objects are resource objects to be transferred in the financial business system; the second recommended sample accounts are accounts with potential pre-defined behaviors towards the candidate sample resource objects. The candidate sample resource objects can be resource objects to be transferred obtained from the financial business system without precise screening. The pre-defined behavior can be a resource transfer action towards the candidate sample resource objects, such as trading or purchasing the candidate sample resource objects.

[0124] Step S506: Input the candidate sample resource object and the multiple second recommended sample accounts associated with the candidate sample resource object into the general account recommendation model to obtain the predicted resource transfer tendency of each second recommended sample account for the candidate sample resource object.

[0125] First, features can be extracted from candidate resource objects to obtain their object feature information. Then, features can be extracted from multiple second-to-be-recommended sample accounts to obtain their account feature information. Second, the object feature information and account feature information are input into a general account recommendation model to obtain the predicted resource transfer tendency of each second-to-be-recommended sample account towards the candidate resource objects. This predicted resource transfer tendency represents the resource transfer willingness of each second-to-be-recommended sample account towards the candidate resource objects, for example, the transaction probability.

[0126] Step S507: Based on the predicted resource transfer tendency of the candidate sample resource objects, obtain the target sample account with the highest second resource transfer tendency from multiple second recommended sample accounts.

[0127] The target sample account is the one with the highest predicted resource transfer tendency among multiple second-tier recommended sample accounts. The target sample account can be calculated based on the following formula (1):

[0128]

[0129] Where, xi For the i-th sample object in the candidate sample resource objects, w j For the j-th sample account among multiple second sample accounts to be recommended; α i The mean value corresponding to the i-th sample object corresponds to the sample account with the highest predicted resource transfer tendency, i.e., the target sample account. Let σ be the predicted resource transfer propensity of the j-th sample account associated with the j-th sample object; σ be the initial standard deviation; and Z be the normalization factor. Formula (1) above can be used to determine sample accounts with a predicted resource transfer propensity greater than the propensity threshold, and also to determine target sample accounts.

[0130] Step S508: Based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource object, obtain the prediction error of the candidate sample resource object.

[0131] According to formula (1) in step S507, the mean of the candidate sample resource object can be determined, which is the sample account with the highest predicted resource transfer tendency, i.e., the target sample account. Furthermore, the actual account feature information corresponding to this mean can also be determined in the second sample account to be recommended. Then, the difference between the predicted account feature information and the actual account feature information of the target sample account is determined as the prediction error of the candidate sample resource object. This prediction error is used to characterize the accuracy of predicting the resource transfer tendency of the candidate sample resource object based on the general account recommendation model. The larger the prediction error, the lower the accuracy of the predicted resource transfer tendency of the candidate sample resource object predicted by the general account recommendation model; the smaller the prediction error, the higher the accuracy of the predicted resource transfer tendency of the candidate sample resource object predicted by the general account recommendation model. Therefore, based on the prediction error, it can be determined whether the candidate sample resource object is suitable for the general account recommendation model.

[0132] In some possible implementations, step S508 may include:

[0133] Obtain the actual account feature information corresponding to the target sample account, and obtain the predicted account feature information corresponding to the target sample account; the difference between the actual account feature information and the predicted account feature information is determined as the prediction error.

[0134] The prediction error can be calculated based on formula (2):

[0135]

[0136] in, The feature information of the target sample account, i.e., the predicted account feature information; α i The feature information of the initial sample accounts, i.e., the feature information of the actual accounts; e iThis represents the prediction error for the candidate sample resource object.

[0137] Step S509: Based on the prediction errors corresponding to each candidate sample resource object, obtain the average prediction error of each candidate sample resource object.

[0138] The average prediction error can be the average of the prediction errors corresponding to each candidate sample resource object.

[0139] Step S510: Candidate sample resource objects whose prediction error is less than the average prediction error are identified as target sample resource objects.

[0140] Among them, candidate sample resource objects with prediction errors less than the average prediction error can be candidate sample resource objects with high accuracy in predicting resource transfer propensity obtained by the general account recommendation model, that is, candidate sample resource objects applicable to the general account recommendation model. These candidate sample resource objects can be resource objects to be transferred that have not undergone precise screening and are obtained from the financial business system. The initial screening sample resource objects are resource objects to be transferred that have undergone precise screening based on prediction errors. The target sample resource object is any one of the initial screening sample resource objects.

[0141] Step S511: Obtain multiple first sample accounts to be recommended associated with the target sample resource object.

[0142] Among them, the target sample resource object is the resource object to be transferred in the financial business system whose prediction error is less than the average prediction error; the first sample account to be recommended is an account with potential preset behavior for the target sample resource object.

[0143] Step S512: Input the target sample resource object and multiple first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0144] Step S513: Obtain the predicted standard deviation of the target sample resource object based on the predicted resource transfer tendency of the target sample resource object.

[0145] Step S514: Based on the predicted standard deviation, replace the initial standard deviation in the general account recommendation model with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object.

[0146] The initial standard deviation is a preset fixed value for each sample object. This dedicated account recommendation model is an account recommendation model applicable to target sample resource objects. It is used to obtain the resource transfer tendency of each first-to-be-recommended sample account associated with the target sample resource object for the target sample resource object based on the label distribution relationship between the target sample resource object and the associated first-to-be-recommended sample account. In turn, it can also obtain the account with the highest resource transfer tendency for the target sample resource object, which is more targeted and has higher accuracy in account recommendation.

[0147] In this embodiment, the method obtains a target sample resource object and multiple first-to-be-recommended sample accounts associated with the target sample resource object. The target sample resource object and the multiple first-to-be-recommended sample accounts are then input into a pre-trained general account recommendation model. This yields the predicted resource transfer tendency of each first-to-be-recommended sample account for the target sample resource object. Based on the predicted resource transfer tendency, the predicted standard deviation corresponding to the target sample resource object is obtained. Finally, based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain a dedicated account recommendation model for the target sample resource object. The account recommendation model acquisition method provided in this embodiment can adaptively adjust the general account recommendation model by calculating the predicted standard deviation of the target sample resource object to obtain a dedicated account recommendation model suitable for the target sample resource object. This improves the accuracy of account recommendations and addresses the problems of uneven recommendations, insufficient recommendations for some resource objects, and low transaction volumes.

[0148] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed 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 performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0149] Based on the same inventive concept, embodiments of this application also provide a training apparatus for an account recommendation model to implement the training method for the account recommendation model described above, and an account recommendation apparatus to implement the account recommendation method described above. The solution provided by this apparatus is similar to the solution described in the above method. Therefore, the specific limitations of the training apparatus embodiments for the account recommendation model provided below can be found in the limitations of the training method for the account recommendation model above, and the specific limitations of the one or more account recommendation apparatus embodiments can be found in the limitations of the account recommendation method above, and will not be repeated here.

[0150] In one embodiment, such as Figure 6 As shown, a training apparatus for an account recommendation model is provided, comprising: a training module 601, an acquisition module 602, a first calculation module 603, a second calculation module 604, and an adjustment module 605, wherein:

[0151] Training module 601 is used to obtain a pre-trained general account recommendation model;

[0152] The acquisition module 602 is used to acquire a target sample resource object and acquire a plurality of first sample accounts to be recommended associated with the target sample resource object; the first sample accounts to be recommended are accounts that have potential preset behaviors for the target sample resource object;

[0153] The first calculation module 603 is used to input the target sample resource object and the plurality of first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object.

[0154] The second calculation module 604 is used to obtain the predicted standard deviation of the target sample resource object based on the predicted resource transfer tendency for the target sample resource object.

[0155] The adjustment module 605 is used to replace the initial standard deviation in the general account recommendation model with the predicted standard deviation based on the predicted standard deviation, so as to obtain the special account recommendation model for the target sample resource object; the initial standard deviation is a preset fixed value of the standard deviation in the general account recommendation model.

[0156] Additionally, the acquisition module 602 is further configured to: acquire multiple candidate sample resource objects, and acquire multiple second recommended sample accounts associated with each candidate sample resource object; the candidate sample resource objects are resource objects to be transferred in a financial business system; the second recommended sample accounts are accounts with potential preset behaviors toward the candidate sample resource objects; input the candidate sample resource objects and the multiple second recommended sample accounts associated with each candidate sample resource object into the general account recommendation model to obtain the predicted resource transfer tendency of each second recommended sample account toward the candidate sample resource object; and acquire the target sample resource object from the multiple candidate sample resource objects based on the predicted resource transfer tendency of the candidate sample resource objects.

[0157] The acquisition module 602 is further configured to: obtain the target sample account with the highest predicted resource transfer tendency from the plurality of second-to-be-recommended sample accounts based on the predicted resource transfer tendency of the candidate sample resource object; the target sample account is the sample account with the highest predicted resource transfer tendency among the plurality of second-to-be-recommended sample accounts; obtain the prediction error of the candidate sample resource object based on the predicted resource transfer tendency of the target sample account for the candidate sample resource object; obtain the average prediction error of each candidate sample resource object based on the prediction error corresponding to each candidate sample resource object; and determine the candidate sample resource object whose prediction error is less than the average prediction error as the target sample resource object.

[0158] Furthermore, the acquisition module 602 is also used to: acquire the actual account feature information corresponding to the target sample account, and acquire the predicted account feature information corresponding to the target sample account based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource object; and determine the difference between the actual account feature information and the predicted account feature information as the prediction error.

[0159] The first calculation module 603 is further configured to: extract object feature information of the target sample resource object and account feature information corresponding to each of the first recommended sample accounts; input the object feature information and the account feature information into the general account recommendation model to obtain the predicted resource transfer tendency of each of the first recommended sample accounts for the target sample resource object.

[0160] The training module 601 is further configured to: acquire multiple training sample resource objects and multiple third recommendation sample accounts associated with each training sample resource object; extract training object feature information of the training sample resource objects and training account feature information of the third recommendation sample accounts associated with the training sample resource objects, and acquire the actual resource transfer tendency corresponding to each of the third recommendation sample accounts corresponding to the training object feature information; input the training object feature information and the training account feature information into the general account recommendation model to be trained, and obtain the predicted resource transfer tendency of each of the third recommendation sample accounts for the training sample resource objects; and train the general account recommendation model to be trained based on the difference between the predicted resource transfer tendency of the training sample resource objects and the actual resource transfer tendency.

[0161] In one embodiment, such as Figure 7 As shown, an account recommendation device is provided, including: an acquisition module 701, a calculation module 702, and a filtering module 703, wherein:

[0162] The acquisition module 701 is used to acquire the resource object to be recommended, and to acquire a plurality of candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended.

[0163] The calculation module 702 is used to input the resource object to be recommended and the plurality of candidate accounts to be recommended into the dedicated account recommendation model of the resource object to be recommended, and to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended.

[0164] The filtering module 703 is used to obtain accounts from the candidate recommended accounts whose predicted resource transfer tendency is greater than a preset transfer tendency threshold, and use them as target recommended accounts corresponding to the recommended resource objects.

[0165] The training device and various modules of the aforementioned account recommendation model can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0166] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores account recommendation data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a training method for an account recommendation model and / or an account recommendation method.

[0167] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0168] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0169] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0170] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0171] 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 used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. 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), magnetic 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 take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A training method for an account recommendation model, characterized in that, The method includes: Obtain a pre-trained general account recommendation model; Obtain a target sample resource object, and obtain multiple first sample accounts to be recommended associated with the target sample resource object; the first sample accounts to be recommended are accounts that have potential preset behaviors for the target sample resource object; The target sample resource object and the plurality of first sample accounts to be recommended are input into the general account recommendation model to obtain the predicted resource transfer tendency of each first sample account to be recommended in relation to the target sample resource object; the predicted resource transfer tendency represents the predicted probability that the first sample account to be recommended will perform a preset behavior in relation to the target sample resource object; Based on the predicted resource transfer tendency of the target sample resource object, the account with the highest resource tendency for the target sample resource object is obtained; based on the feature information of the account with the highest resource tendency, the predicted standard deviation corresponding to the target sample resource object is obtained; the predicted standard deviation is the standard deviation applicable to the target sample resource object, used to represent the distribution of multiple first recommended sample accounts associated with the target sample resource object; Based on the predicted standard deviation, the initial standard deviation in the general account recommendation model is replaced with the predicted standard deviation to obtain the dedicated account recommendation model for the target sample resource object; the initial standard deviation is a preset fixed value of the standard deviation in the general account recommendation model. The acquisition of target sample resource objects includes: acquiring multiple candidate sample resource objects, and acquiring multiple second recommended sample accounts associated with each candidate sample resource object; the candidate sample resource objects are resource objects to be transferred in the financial business system; the second recommended sample accounts are accounts with potential preset behaviors for the candidate sample resource objects; The candidate sample resource object and the multiple second recommended sample accounts associated with the candidate sample resource object are input into the general account recommendation model to obtain the predicted resource transfer tendency of each second recommended sample account for the candidate sample resource object. Based on the predicted resource transfer tendency of the candidate sample resource objects, a target sample account with the highest predicted resource transfer tendency is obtained from the plurality of second-to-be-recommended sample accounts; the target sample account is the sample account with the highest predicted resource transfer tendency among the plurality of second-to-be-recommended sample accounts; the actual account feature information corresponding to the target sample account is obtained, and the predicted account feature information corresponding to the target sample account is obtained based on the predicted resource transfer tendency of the target sample account towards the candidate sample resource objects; the difference between the actual account feature information and the predicted account feature information is determined as the prediction error of the candidate sample resource object; based on the prediction error corresponding to each candidate sample resource object, the average prediction error of each candidate sample resource object is obtained; the candidate sample resource objects whose prediction error is less than the average prediction error are determined as target sample resource objects.

2. The method of claim 1, wherein, The step of inputting the target sample resource object and the plurality of first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object includes: Extract the object feature information of the target sample resource object and the account feature information corresponding to each of the first sample accounts to be recommended; The object feature information and the account feature information are input into the general account recommendation model to obtain the predicted resource transfer tendency of each of the first sample accounts to be recommended for the target sample resource object.

3. The method of claim 1, wherein, The process of obtaining a pre-trained general account recommendation model includes: Obtain multiple training sample resource objects and multiple third recommendation sample accounts associated with each of the training sample resource objects; Extract the training object feature information of the training sample resource object and the training account feature information of the third recommendation sample account associated with the training sample resource object, and obtain the actual resource transfer tendency corresponding to each of the third recommendation sample accounts corresponding to the training object feature information; The feature information of the object to be trained and the feature information of the account to be trained are input into the general account recommendation model to be trained to obtain the predicted resource transfer tendency of each of the third sample accounts to be recommended for the resource object to be trained. The general account recommendation model is trained based on the difference between the predicted resource transfer tendency and the actual resource transfer tendency of the resource objects in the training sample.

4. An account recommendation method, characterized in that, The method includes: Obtain the resource object to be recommended, and obtain multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended. The resource object to be recommended and the plurality of candidate accounts to be recommended are input into the dedicated account recommendation model of the resource object to be recommended, so as to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability of each candidate account performing a preset behavior for the resource object to be recommended; the dedicated account recommendation model of the resource object to be recommended is trained by the training method of the account recommendation model as described in any one of claims 1 to 3; Accounts with a predicted resource transfer tendency greater than a preset transfer tendency threshold are selected from the candidate recommended accounts and used as the target recommended accounts corresponding to the recommended resource objects.

5. A training device for an account recommendation model, characterized in that, The training apparatus for the account recommendation model is used to implement the training method for the account recommendation model according to any one of claims 1-3, and the apparatus comprises: The training module is used to obtain a pre-trained general account recommendation model; The acquisition module is used to acquire a target sample resource object and acquire multiple first sample accounts to be recommended associated with the target sample resource object; the first sample accounts to be recommended are accounts that have potential preset behaviors for the target sample resource object; The first calculation module is used to input the target sample resource object and the plurality of first recommended sample accounts into the general account recommendation model to obtain the predicted resource transfer tendency of each first recommended sample account for the target sample resource object. The second calculation module is used to obtain the predicted standard deviation corresponding to the target sample resource object based on the predicted resource transfer tendency for the target sample resource object. The adjustment module is used to replace the initial standard deviation in the general account recommendation model with the predicted standard deviation based on the predicted standard deviation, so as to obtain the special account recommendation model for the target sample resource object; the initial standard deviation is a preset fixed value of the standard deviation in the general account recommendation model.

6. An account recommendation apparatus characterized by comprising: The device includes: The acquisition module is used to acquire the resource object to be recommended, and to acquire multiple candidate accounts to be recommended that are pre-associated with the resource object to be recommended; the candidate accounts to be recommended are accounts that have potential preset behaviors for the resource object to be recommended. A calculation module is used to input the resource object to be recommended and the plurality of candidate accounts to be recommended into a dedicated account recommendation model for the resource object to be recommended, and to obtain the predicted resource transfer tendency of each candidate account for the resource object to be recommended; the predicted resource transfer tendency represents the predicted probability that each candidate account will perform a preset behavior for the resource object to be recommended; the dedicated account recommendation model for the resource object to be recommended is trained by the training method of the account recommendation model as described in any one of claims 1 to 3; The filtering module is used to obtain accounts from the candidate recommended accounts whose predicted resource transfer tendency is greater than a preset transfer tendency threshold, and use them as target recommended accounts corresponding to the recommended resource objects. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-3.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-3.

9. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-3.

Images