Information processing method and device, electronic equipment and storage medium

Abstract

The invention relates to an information processing method and device, electronic equipment and a storage medium. The method comprises the steps that multiple pieces of candidate recommendation information are acquired; each piece of candidate recommendation information is recommendation information for one target object; based on a first predicted resource quantity corresponding to a target object in each piece of candidate recommendation information and a second predicted resource quantity corresponding to the target object, sorting the multiple pieces of candidate recommendation information to obtain an information sorting result; the first predicted resource quantity represents an operation resource quantity generated when a target account performs operation on associated recommendation information in the target object, and the second predicted resource quantity represents a resource consumption quantity corresponding to resource consumption operation performed by the target account in the target object; determining target recommendation information based on the information sorting result; the target recommendation information is used for being pushed to the target account. The reasonability of determining the to-be-recommended resources can be improved, and then the utilization rate of the recommended information is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of information processing technology, and in particular to an information processing method, device, electronic device and storage medium. Background Art

[0002] In the related art, when recommending information, target recommended information is generally determined from candidate recommended information and the target recommended information is pushed; in the process of information recommendation, there may be unreasonable distribution of recommended information, resulting in low information utilization. Summary of the invention

[0003] The present disclosure provides an information processing method, device, electronic device and storage medium to at least solve the problem in the related art that in the information recommendation process, there may be unreasonable distribution of recommended information, resulting in low information utilization. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided an information processing method, including:

[0005] Acquire multiple candidate recommendation information; each candidate recommendation information is recommendation information for a target object;

[0006] Based on a first predicted resource amount corresponding to a target object in each candidate recommendation information and a second predicted resource amount corresponding to the target object, the plurality of candidate recommendation information are sorted to obtain an information sorting result; the first predicted resource amount represents an operation resource amount generated by the target account performing an operation on the associated recommendation information in the target object, and the second predicted resource amount represents a consumption resource amount corresponding to the resource consumption operation performed by the target account on the target object;

[0007] Based on the information sorting result, target recommendation information is determined; the target recommendation information is used to push to the target account.

[0008] In an exemplary embodiment, the first predicted resource amount includes a first coarse predicted resource amount and a first fine predicted resource amount, and the second predicted resource amount includes a second coarse predicted resource amount and a second fine predicted resource amount;

[0009] The sorting of the plurality of candidate recommendation information based on the first predicted resource amount corresponding to the target object in each candidate recommendation information and the second predicted resource amount corresponding to the target object to obtain the information sorting result includes:

[0010] Performing a first resource prediction based on the account information of the target account and the characteristic information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount;

[0011] Based on the first rough ranking predicted resource amount and the second rough ranking predicted resource amount, the plurality of candidate recommendation information are roughly ranked to obtain a rough ranking result;

[0012] Determining a plurality of rough-ranked recommendation information from the plurality of candidate recommendation information based on the rough-ranking result;

[0013] Performing a second resource prediction based on the account information of the target account and the characteristic information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount;

[0014] Based on the first fine ranking predicted resource quantity and the second fine ranking predicted resource quantity, the plurality of coarse ranking recommendation information are finely ranked to obtain the information ranking result.

[0015] In an exemplary embodiment, performing the first resource prediction based on the account information of the target account and the feature information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount includes:

[0016] Inputting the account information of the target account and the characteristic information of the target object into a first rough ranking prediction model for resource prediction to obtain the first rough ranking predicted resource amount; the first rough ranking prediction model is obtained by training a first preset rough ranking model based on the account information of the sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0017] The account information of the target account and the characteristic information of the target object are input into a second coarse-order prediction model for resource prediction to obtain the second coarse-order predicted resource amount; the second coarse-order prediction model is obtained by training a second preset coarse-order model based on the account information of the sample account, the characteristic information of the sample object, and the consumed resource amount corresponding to the resource consumption operation performed by the sample account in the sample object.

[0018] In an exemplary embodiment, the method further comprises:

[0019] Acquire a training sample; the training sample includes account information of the sample account, feature information of the sample object, and sample resource consumption amount corresponding to resource consumption operation performed by the sample account in the sample object;

[0020] Inputting the account information of the sample account and the feature information of the sample object into the second preset coarse sorting model to perform resource prediction, and obtaining the training prediction resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0021] Perform loss processing based on the sample consumption resource amount and the training predicted resource amount to determine resource amount loss information;

[0022] The model parameters of the second preset rough-order model are updated based on the resource loss information to obtain the second rough-order prediction model.

[0023] In an exemplary embodiment, the step of roughly ranking the candidate recommendation information based on the first roughly ranked predicted resource amount and the second roughly ranked predicted resource amount to obtain a roughly ranked result includes:

[0024] Performing resource quantity fusion processing on the first coarsely predicted resource quantity and the second coarsely predicted resource quantity to obtain a first fused resource quantity;

[0025] Based on the account information of the target account and the feature information of each candidate recommendation information, an index prediction is performed on each candidate recommendation information to obtain a plurality of rough ranking prediction indexes corresponding to each candidate recommendation information;

[0026] Based on the first fusion resource amount, data fusion processing is performed on the multiple rough prediction indicators respectively to obtain multiple fusion indicators corresponding to each candidate recommendation information;

[0027] Based on the multiple fusion indicators corresponding to each candidate recommendation information, determine the rough ranking comprehensive indicator corresponding to each recommendation information;

[0028] The rough ranking result is obtained by sorting the plurality of candidate recommendation information based on the rough ranking comprehensive indicators corresponding to each of the plurality of candidate recommendation information.

[0029] In an exemplary embodiment, performing resource amount fusion processing on the first coarse-sequence predicted resource amount and the second coarse-sequence predicted resource amount to obtain a first fused resource amount includes:

[0030] The ratio of the sum of the first rough predicted resource amount and the second rough predicted resource amount to the first recommendation coefficient is determined as the first fused resource amount; the first recommendation coefficient represents the correlation information between the expected resource amount obtained for each candidate recommended information and the actually consumed resource amount for each candidate recommended information.

[0031] In an exemplary embodiment, performing the second resource prediction based on the account information of the target account and the characteristic information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount includes:

[0032] Inputting the account information of the target account and the characteristic information of the target object into a first precise ranking prediction model for resource prediction to obtain a first precise ranking predicted resource amount; the first precise ranking prediction model is obtained by training a first preset precise ranking model based on the account information of a sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommended information in the sample object;

[0033] The account information of the target account and the characteristic information of the target object are input into a second precise ranking prediction model for resource prediction to obtain the second precise ranking predicted resource quantity; the second precise ranking prediction model is obtained by training a second preset precise ranking model based on the account information of the sample account, the characteristic information of the sample object, and the resource consumption quantity corresponding to the resource consumption operation performed by the sample account in the sample object.

[0034] In an exemplary embodiment, the step of finely sorting the plurality of coarsely ranked recommendation information based on the first finely ranked predicted resource quantity and the second finely ranked predicted resource quantity to obtain the information sorting result includes:

[0035] Performing resource quantity fusion processing on the first refined predicted resource quantity and the second refined predicted resource quantity to obtain a second fused resource quantity;

[0036] Based on the account information of the target account and the feature information of the candidate recommendation information, an index prediction is performed on each rough-ranked recommendation information to obtain a fine-ranking prediction index corresponding to each rough-ranked recommendation information;

[0037] Performing data fusion processing on the refined ranking prediction index based on the second fused resource quantity to obtain a refined ranking comprehensive index corresponding to each rough ranking recommendation information;

[0038] The information sorting result is obtained by sorting the information based on the respective corresponding comprehensive indicators of the rough ranking of the plurality of recommended information.

[0039] In an exemplary embodiment, performing resource amount fusion processing on the first refined predicted resource amount and the second refined predicted resource amount to obtain a second fused resource amount includes:

[0040] The ratio of the sum of the first fine-rank predicted resource quantity and the second fine-rank predicted resource quantity to the second recommendation coefficient is determined as the second fused resource quantity; the second recommendation coefficient represents the correlation information between the expected resource quantity of each rough-rank recommended information and the actually consumed resource quantity of each rough-rank recommended information.

[0041] In an exemplary embodiment, the method further comprises:

[0042] In the case where the target recommendation information has been pushed to the target account, a first return event and a second return event are received within a preset time period; the first return event includes the amount of operation resources generated by the target account performing an operation on the associated recommendation information in the target object of the target recommendation information, and the second return event includes the amount of consumed resources corresponding to the resource consumption operation performed by the target account in the target object;

[0043] Determine a target consumed resource amount based on a ratio of the sum of the operating resource amount and the consumed resource amount to a third recommendation coefficient; the third recommendation coefficient represents correlation information between the expected acquired resource amount of the target recommended information and the actual consumed resource amount of the target recommended information;

[0044] The target recommendation information is recommended based on the target resource consumption amount.

[0045] According to a second aspect of an embodiment of the present disclosure, there is provided an information processing device, including:

[0046] A candidate recommendation information acquisition unit is configured to acquire a plurality of candidate recommendation information; each candidate recommendation information is recommendation information for a target object;

[0047] The information sorting unit is configured to sort the plurality of candidate recommendation information based on a first predicted resource amount corresponding to a target object in each candidate recommendation information and a second predicted resource amount corresponding to the target object, to obtain an information sorting result; the first predicted resource amount represents an operation resource amount generated by the target account performing an operation on the associated recommendation information in the target object, and the second predicted resource amount represents a consumption resource amount corresponding to the resource consumption operation performed by the target account on the target object;

[0048] The target recommendation information determination unit is configured to determine the target recommendation information based on the information sorting result; the target recommendation information is used to push to the target account.

[0049] In an exemplary embodiment, the first predicted resource amount includes a first coarse predicted resource amount and a first fine predicted resource amount, and the second predicted resource amount includes a second coarse predicted resource amount and a second fine predicted resource amount;

[0050] The information sorting unit comprises:

[0051] A first prediction unit is configured to perform a first resource prediction based on the account information of the target account and the feature information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount;

[0052] A rough sorting unit is configured to perform rough sorting on the plurality of candidate recommendation information based on the first rough sorting predicted resource amount and the second rough sorting predicted resource amount to obtain a rough sorting result;

[0053] A first determining unit is configured to determine a plurality of coarse-ranked recommendation information from the plurality of candidate recommendation information based on the coarse-ranking result;

[0054] A second prediction unit is configured to perform a second resource prediction based on the account information of the target account and the feature information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount;

[0055] The fine ranking unit is configured to perform fine ranking on the plurality of coarse ranking recommendation information based on the first fine ranking predicted resource quantity and the second fine ranking predicted resource quantity to obtain the information ranking result.

[0056] In an exemplary embodiment, the first prediction unit is configured to perform:

[0057] Inputting the account information of the target account and the characteristic information of the target object into a first rough ranking prediction model for resource prediction to obtain the first rough ranking predicted resource amount; the first rough ranking prediction model is obtained by training a first preset rough ranking model based on the account information of the sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0058] The account information of the target account and the characteristic information of the target object are input into a second coarse-order prediction model for resource prediction to obtain the second coarse-order predicted resource amount; the second coarse-order prediction model is obtained by training a second preset coarse-order model based on the account information of the sample account, the characteristic information of the sample object, and the consumed resource amount corresponding to the resource consumption operation performed by the sample account in the sample object.

[0059] In an exemplary embodiment, the device model training unit, the model training unit, is configured to perform:

[0060] Acquire a training sample; the training sample includes account information of the sample account, feature information of the sample object, and sample resource consumption amount corresponding to resource consumption operation performed by the sample account in the sample object;

[0061] Inputting the account information of the sample account and the feature information of the sample object into the second preset coarse sorting model to perform resource prediction, and obtaining the training prediction resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0062] Perform loss processing based on the sample consumption resource amount and the training predicted resource amount to determine resource amount loss information;

[0063] The model parameters of the second preset rough-order model are updated based on the resource loss information to obtain the second rough-order prediction model.

[0064] In an exemplary embodiment, the coarse arrangement unit comprises:

[0065] A first resource quantity fusion unit is configured to perform resource quantity fusion processing on the first coarse-order predicted resource quantity and the second coarse-order predicted resource quantity to obtain a first fused resource quantity;

[0066] A first indicator prediction unit is configured to perform indicator prediction on each candidate recommendation information based on the account information of the target account and the feature information of each candidate recommendation information, so as to obtain a plurality of rough ranking prediction indicators corresponding to each candidate recommendation information;

[0067] A first data fusion unit is configured to perform data fusion processing on the plurality of rough ranking prediction indicators based on the first fusion resource amount to obtain a plurality of fusion indicators corresponding to each candidate recommendation information;

[0068] A second determining unit is configured to determine a rough ranking comprehensive index corresponding to each recommendation information based on a plurality of fusion indexes corresponding to each candidate recommendation information;

[0069] The first sorting unit is configured to perform sorting based on the rough sorting comprehensive indicators corresponding to each of the plurality of candidate recommendation information to obtain the rough sorting result.

[0070] In an exemplary embodiment, the first resource quantity fusion unit is configured to execute:

[0071] The ratio of the sum of the first rough predicted resource amount and the second rough predicted resource amount to the first recommendation coefficient is determined as the first fused resource amount; the first recommendation coefficient represents the correlation information between the expected resource amount obtained for each candidate recommended information and the actually consumed resource amount of each candidate recommended information.

[0072] In an exemplary embodiment, the second prediction unit is configured to perform:

[0073] Inputting the account information of the target account and the characteristic information of the target object into a first precise ranking prediction model for resource prediction to obtain a first precise ranking predicted resource amount; the first precise ranking prediction model is obtained by training a first preset precise ranking model based on the account information of a sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommended information in the sample object;

[0074] The account information of the target account and the characteristic information of the target object are input into a second precise ranking prediction model for resource prediction to obtain the second precise ranking predicted resource quantity; the second precise ranking prediction model is obtained by training a second preset precise ranking model based on the account information of the sample account, the characteristic information of the sample object, and the resource consumption quantity corresponding to the resource consumption operation performed by the sample account in the sample object.

[0075] In an exemplary embodiment, the fine sorting unit comprises:

[0076] A second resource quantity fusion unit is configured to perform resource quantity fusion processing on the first refined ranking predicted resource quantity and the second refined ranking predicted resource quantity to obtain a second fused resource quantity;

[0077] A second indicator prediction unit is configured to perform indicator prediction on each rough-ranked recommendation information based on the account information of the target account and the feature information of the candidate recommendation information, so as to obtain a fine-ranking prediction indicator corresponding to each rough-ranked recommendation information;

[0078] A second data fusion unit is configured to perform data fusion processing on the refined ranking prediction index based on the second fusion resource amount to obtain a refined ranking comprehensive index corresponding to each rough ranking recommendation information;

[0079] The second sorting unit is configured to perform sorting based on the respective corresponding comprehensive indicators of the rough sorting recommendation information to obtain the information sorting result.

[0080] In an exemplary embodiment, the second resource quantity fusion unit is configured to execute:

[0081] The ratio of the sum of the first fine-rank predicted resource quantity and the second fine-rank predicted resource quantity to the second recommendation coefficient is determined as the second fused resource quantity; the second recommendation coefficient represents the correlation information between the expected resource quantity of each rough-rank recommended information and the actually consumed resource quantity of each rough-rank recommended information.

[0082] In an exemplary embodiment, the apparatus further comprises:

[0083] a feedback event receiving unit, configured to receive a first feedback event and a second feedback event within a preset time period when the target recommendation information has been pushed to the target account; the first feedback event includes an amount of operation resources generated by the target account performing an operation on the associated recommendation information in the target object of the target recommendation information, and the second feedback event includes an amount of consumed resources corresponding to the resource consumption operation performed by the target account in the target object;

[0084] a target resource consumption amount determination unit, configured to determine the target resource consumption amount based on the ratio of the sum of the operation resource amount and the consumption resource amount to a third recommendation coefficient; the third recommendation coefficient represents the correlation information between the expected resource amount obtained by the target recommendation information and the actual resource consumption amount of the target recommendation information;

[0085] The information recommendation unit is configured to recommend the target recommendation information based on the target resource consumption amount.

[0086] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the information processing method as described above.

[0087] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the information processing method as described above.

[0088] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, which includes a computer program, wherein the computer program is stored in a readable storage medium, and at least one processor of a computer device reads and executes the computer program from the readable storage medium, so that the device performs the above-mentioned information processing method.

[0089] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0090] In the present disclosure, predictions are made for the target objects in each candidate recommendation information to obtain the first predicted resource amount and the second predicted resource amount of the target objects in each candidate recommendation information. The first predicted resource amount can represent the operation resource amount generated by the target account for operating the associated recommendation information in the target object, and the second predicted resource amount can represent the consumption resource amount corresponding to the resource consumption operation of the target account in the target object. Thus, the first predicted resource amount and the second predicted resource amount can reflect the information of the target object in multiple dimensions. Furthermore, the multiple candidate recommendation information are sorted by combining the predicted resource amounts of each candidate recommendation information in multiple dimensions. Different from the related art where the multiple candidate recommendation information are sorted only based on the predicted resource amount in a single dimension, since only the predicted resource amount in a single dimension is considered, a comprehensive resource amount evaluation of the target objects in each candidate recommendation information cannot be performed, which affects the sorting result of the recommendation information and further leads to an unreasonable situation of the allocated information to be recommended. Accordingly, the present disclosure comprehensively predicts the resource amounts of the target objects in each candidate recommendation information in multiple dimensions, and then the multiple candidate recommendation information can be sorted based on the predicted resource amounts of the target objects in each candidate recommendation information in multiple dimensions to determine the target recommendation information, improving the rationality of determining the information to be recommended, and further improving the information recommendation effect and the utilization rate of the recommendation information.

[0091] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure, and do not constitute an improper limitation to the present disclosure.

[0093] Figure 1 is a schematic diagram of an implementation environment shown according to an exemplary embodiment.

[0094] Figure 2 is a flowchart of an information processing method shown according to an exemplary embodiment.

[0095] Figure 3 is a flowchart of an information sorting method shown according to an exemplary embodiment.

[0096] Figure 4 is a flowchart of a training method for a second rough ranking prediction model shown according to an exemplary embodiment.

[0097] Figure 5 is a flowchart of a rough ranking method based on predicted resource amounts shown according to an exemplary embodiment.

[0098] Figure 6 The present invention is a flow chart of a method for precise sorting based on predicted resource quantity according to an exemplary embodiment.

[0099] Figure 7 The present invention is a flow chart of an information recommendation method according to an exemplary embodiment.

[0100] Figure 8 The figure is a schematic diagram of a recommendation information processing flow according to an exemplary embodiment.

[0101] Fig. 9 The invention is a block diagram of an information processing device according to an exemplary embodiment.

[0102] Fig.10 The figure is a schematic diagram showing the structure of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0103] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0104] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0105] 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 display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0106] See also Figure 1 , which shows a schematic diagram of an implementation environment provided by an embodiment of the present disclosure, the implementation environment may include: at least one electronic terminal 110 and a recommendation server 120, and the electronic terminal 110 and the recommendation server 120 can communicate data through a network.

[0107] Specifically, the electronic terminal 110 can send an information recommendation request to the recommendation server 120, and the recommendation server 120 can recall multiple candidate recommendation information based on the account information of the target account corresponding to the electronic terminal 110 and the characteristic information of the information to be recommended; further, based on the account information of the target account and the target object in each candidate recommendation information, the resource quantity of the target object can be predicted, and then based on the predicted resource quantity, multiple candidate recommendation information can be sorted and the target recommendation information can be determined; and the target recommendation information can be pushed to the electronic terminal 110.

[0108] The electronic terminal 110 can communicate with the recommendation server 120 based on the browser / server mode (B / S) or the client / server mode (C / S). The electronic terminal 110 may include: physical devices such as smart phones, tablet computers, laptops, digital assistants, smart wearable devices, vehicle terminals, servers, etc., and may also include software running in physical devices, such as applications, etc. The operating system running on the electronic terminal 110 in the embodiment of the present disclosure may include but is not limited to Android system, IOS system, Linux, Windows, etc.

[0109] The recommendation server 120 and the electronic terminal 110 may establish a communication connection via wired or wireless communication. The recommendation server 120 may include an independently operated server, or a distributed server, or a server cluster consisting of multiple servers, wherein the server may be a cloud server.

[0110] In order to solve the problem that in the process of information recommendation in the related art, there may be unreasonable distribution of recommended information, which leads to low information utilization, the embodiment of the present disclosure provides an information processing method, and the execution subject of the method can be the above-mentioned recommendation server, please refer to Figure 2 , the method may include:

[0111] S210. Acquire multiple candidate recommendation information; each candidate recommendation information is recommendation information for a target object.

[0112] In this embodiment, multiple candidate recommendation information may be recommendation information recalled from the recommendation information set and adapted to the account information of the target account. Each candidate recommendation information is recommendation information for a target object, that is, each candidate recommendation information is recommendation information for recommending a target object. The candidate recommendation information may include description information of the target object and object pointing information. The object pointing information may be a download address or download link of the target object, etc. The target object may specifically be a game object, a short video object, an instant messaging object, etc. The target object may be an independent object program, or a small program that depends on other object programs, etc. This embodiment does not make specific limitations. Multiple different candidate recommendation information may correspond to the same target object. In this case, these multiple different candidate recommendation information may be recommendation information for the same target object in different recommendation forms, for example, it may include recommendation information in text form for the target object, recommendation information in picture form for the target object, recommendation information in video form for the target object, etc.

[0113] S220. Based on the first predicted resource amount corresponding to the target object in each candidate recommendation information and the second predicted resource amount corresponding to the target object, the multiple candidate recommendation information are sorted to obtain an information sorting result; the first predicted resource amount represents the operation resource amount generated by the target account performing operations on the associated recommendation information in the target object, and the second predicted resource amount represents the consumption resource amount corresponding to the resource consumption operation performed by the target account on the target object.

[0114] For a target object, it may correspond to multiple dimensions of resource amounts, and the resource amount of each dimension can reflect the information of the target object; in this embodiment, for each candidate recommendation information, the resource amount of multiple dimensions of the target object corresponding to the candidate recommendation information can be predicted. Specifically, the operation resource amount generated by the target account performing operations on the associated recommendation information of the target object during the use of the target object is predicted to obtain a first predicted resource amount, and the consumption resource amount corresponding to the resource consumption operation performed by the target account in the target object is predicted to obtain a second predicted resource amount.

[0115] The associated recommendation information of the target object may be information recommended based on the target object platform, and the recommendation is made during the use of the target object; the target account's operation on the associated recommendation information may include operations such as viewing the associated recommendation information; the target account's resource consumption operation in the target object may refer to an operation of consuming virtual resources to achieve the acquisition of virtual items or real items. Thus, the first predicted resource amount is predicted based on the target account's operation on the associated recommendation information, and the second predicted resource amount is predicted based on the target account's possible resource consumption operation.

[0116] S230. Determine target recommendation information based on the information sorting result; the target recommendation information is used to push to the target account.

[0117] When the information sorting result is obtained, if the recommended information in the information sorting result is sorted in descending order based on the recommendation index, in this embodiment, the higher the recommendation index, the greater the possibility that the recommended information will be recommended, so that the first preset number of recommended information ranked at the top can be determined as the target recommended information, and then the target recommended information can be pushed to the target account.

[0118] In the present disclosure, a prediction is made for the target object in each candidate recommendation information to obtain a first predicted resource amount and a second predicted resource amount of the target object in each candidate recommendation information; wherein the first predicted resource amount can represent the operation resource amount generated by the target account operating the associated recommendation information in the target object, and the second predicted resource amount can represent the consumption resource amount corresponding to the resource consumption operation performed by the target account in the target object, so that the first predicted resource amount and the second predicted resource amount can reflect the information of the target object in multiple dimensions, and then the multiple candidate recommendation information are ranked in combination with the predicted resource amount of each candidate recommendation information in multiple dimensions; different from the related art which is only based on a single The predicted resource amount of the target object in each candidate recommendation information is used to sort multiple candidate recommendation information. Since only the predicted resource amount of a single dimension is considered, a comprehensive resource amount evaluation cannot be performed on the target object in each candidate recommendation information, which affects the sorting result of the recommendation information and leads to unreasonable allocation of the information to be recommended. Accordingly, the present invention comprehensively predicts the resource amount of the target object in each candidate recommendation information in multiple dimensions, and then can sort multiple candidate recommendation information based on the predicted resource amount of the target object in each candidate recommendation information in multiple dimensions and determine the target recommendation information, thereby improving the rationality of determining the information to be recommended, and thereby improving the information recommendation effect and the utilization rate of the recommended information.

[0119] In a specific embodiment, the process of sorting multiple candidate recommendation information may include a rough sorting process and a fine sorting process; accordingly, the first predicted resource amount includes a first rough sorting predicted resource amount and a first fine sorting predicted resource amount, and the second predicted resource amount includes a second rough sorting predicted resource amount and a second fine sorting predicted resource amount; for details, please refer to Figure 3 , which shows a method for sorting information, which may include:

[0120] S310. Perform a first resource prediction based on the account information of the target account and the characteristic information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount.

[0121] In this embodiment, the account information of the target account may specifically be feature information that can characterize the features of the target account, and the feature information of the target object may specifically include feature information such as the object type that characterizes the target object, the content information of the target object, and the like; thereby, the degree of matching between the target account and the target object may be determined based on the account information of the target account and the feature information of the target object, and a first resource prediction may be performed accordingly to obtain a first rough prediction resource amount and a second rough prediction resource amount; wherein for each candidate recommendation information, the first rough prediction resource amount may be obtained by predicting the operation resource amount generated by the target account performing operations on the associated recommendation information of the target object in the candidate recommendation information, and the second rough prediction resource amount may be obtained by predicting the consumption resource amount corresponding to the resource consumption operation performed by the target account in the target object.

[0122] S320. Based on the first rough ranking predicted resource amount and the second rough ranking predicted resource amount, perform rough ranking on the plurality of candidate recommendation information to obtain a rough ranking result.

[0123] When the predicted resource amounts of the target object of each candidate recommendation information in multiple dimensions are obtained, the predicted resource amounts in multiple dimensions may be combined to roughly rank the multiple candidate recommendation information, thereby obtaining a rough ranking result.

[0124] S330. Determine a plurality of rough-ranked recommendation information from the plurality of candidate recommendation information based on the rough-ranking result.

[0125] When a rough ranking result is obtained, if the recommended information in the rough ranking result is sorted in descending order based on the recommendation index, in this embodiment, the higher the recommendation index, the greater the possibility that the recommended information is recommended, so that the second preset number of recommended information ranked at the top can be determined as the rough ranking recommended information.

[0126] S340. Perform a second resource prediction based on the account information of the target account and the characteristic information of the target object in each coarse ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount.

[0127] In this embodiment, the account information of the target account may specifically be feature information that can characterize the features of the target account, and the feature information of the target object may specifically include feature information such as the object type that characterizes the target object, the content information of the target object, and the like; thereby, the degree of matching between the target account and the target object may be determined based on the account information of the target account and the feature information of the target object, and a second resource prediction may be performed accordingly, thereby obtaining a first fine ranking predicted resource amount and a second fine ranking predicted resource amount; wherein for each coarse ranking recommendation information, the first fine ranking predicted resource amount may be obtained by predicting the amount of operating resources generated by the target account performing operations on the associated recommendation information of the target object in the coarse ranking recommendation information, and the second fine ranking predicted resource amount may be obtained by predicting the amount of consumed resources corresponding to the resource consumption operations performed by the target account in the target object.

[0128] S350. Based on the first fine-ranking predicted resource quantity and the second fine-ranking predicted resource quantity, fine-rank the plurality of coarse-ranked recommendation information to obtain the information ranking result.

[0129] When the predicted resource amounts of the target object of each rough-ranked recommendation information in multiple dimensions are obtained, the predicted resource amounts in multiple dimensions may be combined to perform rough ranking on the multiple rough-ranked recommendation information, thereby obtaining a rough ranking result.

[0130] In this embodiment, the process of determining the target recommended information from multiple candidate recommended information includes rough sorting and fine sorting, and in the rough sorting process and the fine sorting process, the target object in the corresponding recommended information is predicted for the amount of resources in multiple dimensions, and the rough sorting result and the fine sorting result are obtained accordingly. The rough sorting can preliminarily sort the recalled candidate recommended resources to reduce the amount of calculation and delay of the fine sorting; the fine sorting, on the basis of the rough sorting, sorts the rough sorted recommended information determined by the rough sorting more finely to determine the information sorting result; thus, on the one hand, the target object in the recommended information is predicted and sorted for the amount of resources in multiple dimensions during the sorting process, that is, the amount of resources in multiple dimensions of the target object is comprehensively predicted, thereby improving the rationality of the sorting of the recommended information; on the other hand, the combination of rough sorting and fine sorting is adopted to improve the efficiency and accuracy of the sorting of the recommended information.

[0131] In a specific embodiment, performing the first resource prediction based on the account information of the target account and the characteristic information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount includes:

[0132] Inputting the account information of the target account and the characteristic information of the target object into a first rough ranking prediction model for resource prediction to obtain the first rough ranking predicted resource amount; the first rough ranking prediction model is obtained by training a first preset rough ranking model based on the account information of the sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0133] The account information of the target account and the characteristic information of the target object are input into a second coarse-order prediction model for resource prediction to obtain the second coarse-order predicted resource amount; the second coarse-order prediction model is obtained by training a second preset coarse-order model based on the account information of the sample account, the characteristic information of the sample object, and the consumed resource amount corresponding to the resource consumption operation performed by the sample account in the sample object.

[0134] The first preset rough-cut model is different from the second preset rough-cut model. The first preset rough-cut model is a model for predicting the amount of operating resources, and the second preset rough-cut model is a model for predicting the amount of consumed resources.

[0135] For the training method of the second rough prediction model, please refer to Figure 4 , the method may include:

[0136] S410. Acquire training samples; the training samples include account information of the sample account, feature information of the sample object, and sample resource consumption corresponding to resource consumption operations performed by the sample account in the sample object.

[0137] In this embodiment, the account information of the sample account may specifically be feature information that can characterize the features of the sample account, and the feature information of the sample object may specifically include feature information such as the object type that characterizes the sample object, content information of the sample object, etc.; the sample resource consumption amount corresponding to the resource consumption operation performed by the sample account on the sample object may be regarded as a resource consumption label; the sample resource consumption amount corresponding to the resource consumption operation performed by the sample account on the sample object may be the resource consumption amount corresponding to the actual resource consumption operation performed in the sample object during the use of the sample object by the sample object within the historical time period.

[0138] S420. Input the account information of the sample account and the feature information of the sample object into the second preset coarse sorting model to perform resource prediction, and obtain the training prediction resource amount generated by the sample account operating the associated recommendation information in the sample object.

[0139] The second preset coarse sorting model receives the account information of the sample account and the feature information of the sample object and performs resource prediction to obtain the training prediction resource amount generated by the sample object's possible operation on the sample object-associated recommendation information.

[0140] S430. Perform loss processing based on the sample consumption resource amount and the training predicted resource amount to determine resource loss information.

[0141] Specifically, a difference process may be performed based on the sample consumption resource amount and the training predicted resource amount, and resource loss information may be determined based on the difference information.

[0142] S440. Update the model parameters of the second preset rough-order model based on the resource loss information to obtain the second rough-order prediction model.

[0143] In this embodiment, the second preset coarse-sorting model is trained based on the account information of the sample account, the characteristic information of the sample object, and the sample resource consumption corresponding to the resource consumption operation performed by the sample object in the sample object, so that the trained second coarse-sorting prediction model has the ability to predict the resource consumption based on the object information and the object characteristic information, thereby facilitating the prediction of the resource consumption of the target account in the target object based on the trained second coarse-sorting prediction model, thereby improving the efficiency and accuracy of the resource consumption prediction.

[0144] The training process of the first coarse ranking prediction model is similar to that of the second coarse ranking prediction model. When training the first coarse ranking prediction model, the first preset coarse ranking model can be trained based on the account information of the sample account, the characteristic information of the sample object, and the operation resources generated by the sample account operating the associated recommendation information in the sample object, wherein the operation resources generated by the sample account operating the associated recommendation information in the sample object can be regarded as the operation resource label, thereby inputting the account information of the sample account and the characteristic information of the sample object into the first preset coarse ranking model for training, and then obtaining the first coarse ranking prediction model.

[0145] In this embodiment, the first coarse-ranking prediction model is used to predict the amount of operating resources generated by the target account's operations on the associated recommendation information of the target object in the candidate recommendation information, and the second coarse-ranking prediction model is used to predict the amount of consumed resources corresponding to the target account's resource-consuming operations on the target object, thereby improving the prediction efficiency and accuracy of the first coarse-ranking predicted resource amounts and the second coarse-ranking predicted resource amounts.

[0146] In an alternative embodiment, see Figure 5 , which shows a rough sorting method based on predicted resource quantity, which may include:

[0147] S510. Perform resource quantity fusion processing on the first coarse-grained predicted resource quantity and the second coarse-grained predicted resource quantity to obtain a first fused resource quantity.

[0148] Specifically, performing resource amount fusion processing on the first coarsely predicted resource amount and the second coarsely predicted resource amount to obtain a first fused resource amount includes:

[0149] The ratio of the sum of the first rough predicted resource amount and the second rough predicted resource amount to the first recommendation coefficient is determined as the first fused resource amount; the first recommendation coefficient represents the correlation information between the expected resource amount obtained for each candidate recommended information and the actually consumed resource amount of each candidate recommended information.

[0150] The first recommendation coefficient may be a coefficient pre-set by the information recommender before the information is recommended. Different recommended information may correspond to different information recommenders, and the first recommendation coefficients of corresponding different recommended information may be different. The first recommendation coefficient may specifically refer to the ratio of the expected amount of resources obtained by the candidate recommended information to the actual amount of resources consumed when the recommended information is actually recommended. Specifically, the first recommendation coefficient may be the ratio of the expected amount of resources obtained to the actual amount of resources consumed.

[0151] In this embodiment, by calculating the sum of the first rough predicted resource amount and the second rough predicted resource amount, and calculating the ratio of the sum to the first recommendation coefficient, the first recommendation coefficient is a coefficient associated with the recommendation information, so that the first fused resource amount can be matched with the recommendation logic of the corresponding recommendation information, thereby improving the rationality of the calculation of the first fused resource amount and its adaptability to the recommendation logic.

[0152] S520. Perform index prediction on each candidate recommendation information based on the account information of the target account and the feature information of each candidate recommendation information to obtain a plurality of rough ranking prediction indexes corresponding to each candidate recommendation information.

[0153] The characteristic information of the recommendation information may include the recommendation type of the recommendation information, the recommended content of the recommendation information, and other characteristic information, wherein the recommendation type of the recommendation information may include text, image, video, and other recommendation types. Based on the account information of the target account and the characteristic information of the candidate recommendation information, the indicator prediction is performed to obtain multiple rough ranking prediction indicators corresponding to each candidate recommendation information.

[0154] S530. Perform data fusion processing on the multiple rough prediction indicators based on the first fusion resource amount to obtain multiple fusion indicators corresponding to each candidate recommendation information.

[0155] Specifically, multiple rough prediction indicators can be weighted based on the first fused resource amount to obtain multiple fused indicators corresponding to each candidate recommendation information, each fused indicator fused with the original rough prediction indicator and the first fused resource amount.

[0156] S540. Based on the multiple fusion indicators corresponding to each candidate recommendation information, determine the rough ranking comprehensive indicator corresponding to each recommendation information.

[0157] When multiple fusion indicators corresponding to each candidate recommendation information are determined, the multiple fusion indicators can be further fused based on the multiple fusion indicators corresponding to each candidate recommendation information to obtain a rough ranking comprehensive indicator for subsequent sorting. Specifically, the multiple fusion indicators of each candidate recommendation information can be weighted and summed to obtain a rough ranking comprehensive indicator corresponding to each candidate recommendation information, wherein the weight corresponding to each fusion indicator can be determined based on the actual recommendation logic, that is, it can be flexibly set according to the specific recommendation scenario, and is not limited here.

[0158] S550. Sort the plurality of candidate recommendation information based on the rough ranking comprehensive indicators corresponding to each of the plurality of candidate recommendation information to obtain the rough ranking result.

[0159] When the coarse ranking comprehensive index corresponding to each candidate recommendation information is obtained, multiple candidate recommendation information can be sorted based on the coarse ranking comprehensive index to obtain a coarse ranking result; specifically, multiple candidate recommendation information can be sorted in descending order based on the coarse ranking comprehensive index, and then multiple arranged candidate recommendation information, i.e., a coarse ranking result, can be obtained.

[0160] In this embodiment, by respectively fusing the first fused resource amount of each candidate recommendation information into multiple coarse ranking prediction indicators of the candidate recommendation information, and then sorting the multiple candidate recommendation information based on the fused multiple fusion indicators of each candidate recommendation information, it is possible to integrate the information of multiple dimensions of the candidate recommendation information as the basis for sorting for each candidate recommendation information, thereby avoiding the problem of deviation in the sorting results caused by not sorting based on comprehensive dimensional information, thereby improving the rationality and accuracy of the coarse ranking results.

[0161] In an optional embodiment, performing the second resource prediction based on the account information of the target account and the characteristic information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount includes:

[0162] Inputting the account information of the target account and the characteristic information of the target object into a first precise ranking prediction model for resource prediction to obtain a first precise ranking predicted resource amount; the first precise ranking prediction model is obtained by training a first preset precise ranking model based on the account information of a sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommended information in the sample object;

[0163] The account information of the target account and the characteristic information of the target object are input into a second precise ranking prediction model for resource prediction to obtain the second precise ranking predicted resource quantity; the second precise ranking prediction model is obtained by training a second preset precise ranking model based on the account information of the sample account, the characteristic information of the sample object, and the resource consumption quantity corresponding to the resource consumption operation performed by the sample account in the sample object.

[0164] The first preset fine ranking model, the second preset fine ranking model, the first preset coarse ranking model and the second preset coarse ranking model are all different. The first preset coarse ranking model is a model for predicting the operating resource quantity, and the second preset coarse ranking model is a model for predicting the consumed resource quantity. The first preset fine ranking model is a model for predicting the operating resource quantity, and the second preset fine ranking model is a model for predicting the consumed resource quantity.

[0165] The training process of the first fine ranking prediction model is similar to that of the first coarse ranking prediction model. When training the first fine ranking prediction model, the first preset fine ranking model can be trained based on the account information of the sample account, the characteristic information of the sample object, and the amount of operation resources generated by the sample account operating the associated recommendation information in the sample object, wherein the operation resources generated by the sample account operating the associated recommendation information in the sample object can be regarded as operation resource labels, thereby inputting the account information of the sample account and the characteristic information of the sample object into the first preset fine ranking model for training, and then obtaining the first fine ranking prediction model.

[0166] The training process of the second fine ranking prediction model is similar to that of the second coarse ranking prediction model. When training the second fine ranking prediction model, the second preset fine ranking model can be trained based on the account information of the sample account, the characteristic information of the sample object, and the amount of consumed resources corresponding to the resource consumption operations performed by the sample object in the sample object. The amount of consumed resources corresponding to the resource consumption operations performed by the sample object in the sample object can be regarded as the resource consumption label, thereby inputting the account information of the sample account and the characteristic information of the sample object into the second preset fine ranking model for training, thereby obtaining the second fine ranking prediction model.

[0167] In this embodiment, the first precise ranking prediction model is used to predict the amount of operating resources generated by the target account's operations on the associated recommendation information of the target object in the candidate recommendation information, and the second precise ranking prediction model is used to predict the amount of consumed resources corresponding to the target account's resource consumption operations on the target object, thereby improving the prediction efficiency and accuracy of the first precise ranking predicted resource amount and the second precise ranking predicted resource amount.

[0168] In an alternative embodiment, see Figure 6 , which shows a method for precise sorting based on predicted resource quantities, which may include:

[0169] S610. Perform resource quantity fusion processing on the first refined predicted resource quantity and the second refined predicted resource quantity to obtain a second fused resource quantity.

[0170] Specifically, performing resource amount fusion processing on the first refined predicted resource amount and the second refined predicted resource amount to obtain a second fused resource amount includes:

[0171] The ratio of the sum of the first fine-rank predicted resource quantity and the second fine-rank predicted resource quantity to the second recommendation coefficient is determined as the second fused resource quantity; the second recommendation coefficient represents the correlation information between the expected resource quantity of each rough-rank recommended information and the actually consumed resource quantity of each rough-rank recommended information.

[0172] The second recommendation coefficient may be a coefficient pre-set by the information recommender before the information is recommended. Different recommended information may correspond to different information recommenders, and the second recommendation coefficients of different recommended information may be different. The second recommendation coefficient may specifically refer to the ratio of the expected amount of resources obtained by the rough recommended information to the actual amount of resources consumed when the recommended information is actually recommended. Specifically, the second recommendation coefficient may be the ratio of the expected amount of resources obtained to the actual amount of resources consumed.

[0173] In this embodiment, by calculating the sum of the first precise predicted resource amount and the second precise predicted resource amount, and calculating the ratio with the second recommended coefficient, the second recommended coefficient is a coefficient associated with the recommended information, so that the second fused resource amount can be matched with the recommendation logic of the corresponding recommended information, thereby improving the rationality of the calculation of the second fused resource amount and its adaptability to the recommendation logic.

[0174] S620. Perform an index prediction on each rough-ranked recommendation information based on the account information of the target account and the feature information of the candidate recommendation information to obtain a fine-ranking prediction index corresponding to each rough-ranked recommendation information.

[0175] The characteristic information of the recommendation information may include the recommendation type of the recommendation information, the recommended content of the recommendation information, and other characteristic information, wherein the recommendation type of the recommendation information may include text, picture, video, and other recommendation types. Based on the account information of the target account and the characteristic information of the rough ranking recommendation information, the index prediction is performed to obtain the fine ranking prediction index corresponding to each rough ranking recommendation information.

[0176] S630. Perform data fusion processing on the refined ranking prediction index based on the second fused resource quantity to obtain a refined ranking comprehensive index corresponding to each rough ranking recommendation information.

[0177] Specifically, the refined ranking prediction indicators can be weighted based on the second fused resource amount to obtain the refined ranking comprehensive indicator corresponding to each candidate recommendation information, so that the refined ranking comprehensive indicator integrates the original refined ranking prediction indicator and the second fused resource amount.

[0178] S640. Sorting the plurality of rough-ranked recommendation information based on the respective corresponding comprehensive indices of fine-ranking to obtain the information sorting result.

[0179] When the fine ranking comprehensive index corresponding to each rough ranking recommendation information is obtained, the multiple rough ranking recommendation information can be sorted based on the fine ranking comprehensive index to obtain the information sorting result; specifically, the multiple rough ranking recommendation information can be sorted in descending order based on the fine ranking comprehensive index, and then the multiple arranged rough ranking recommendation information can be obtained, that is, the information sorting result.

[0180] In the present embodiment, by respectively fusing the second fused resource amount of each rough-ranked recommendation information into the refined ranking prediction index of the rough-ranked recommendation information, and then sorting the multiple rough-ranked recommendation information based on the fused refined ranking comprehensive index of each rough-ranked recommendation information, it is possible to integrate the information of multiple dimensions of the rough-ranked recommendation information as the sorting basis for each rough-ranked recommendation information, thereby avoiding the problem of deviation in the sorting results caused by not sorting based on comprehensive dimensional information, thereby improving the rationality and accuracy of the refined ranking results.

[0181] In another optional embodiment, when recommending target recommendation information to the target account, the information recommender can also adjust the amount of resources consumed by the recommendation in real time and flexibly; Figure 7 , which shows an information recommendation method, which may include:

[0182] S710. When the target recommendation information has been pushed to the target account, a first feedback event and a second feedback event are received within a preset time period; the first feedback event includes the amount of operation resources generated by the target account performing operations on the associated recommendation information in the target object of the target recommendation information, and the second feedback event includes the amount of consumed resources corresponding to the resource consumption operations performed by the target account in the target object.

[0183] The preset time period in this embodiment may be the initial stage of pushing the target recommendation information to the target account, and specifically may be a time period starting from the time when the target recommendation information is pushed to the target account and lasting for a preset time period. The first return event is an event related to the amount of operation resources, and the corresponding first return event may include the operation time of the target account performing the operation on the associated recommendation information in the target object of the target recommendation information, and the amount of operation resources generated by the operation; the second return event is an event related to the amount of consumed resources, and the corresponding second return event may include the consumption time of the target account performing the resource consumption operation in the target object of the target recommendation information, and the amount of consumed resources consumed.

[0184] S720. Determine the target consumed resource amount based on the ratio of the sum of the operating resource amount and the consumed resource amount to a third recommendation coefficient; the third recommendation coefficient represents the correlation information between the expected obtained resource amount of the target recommended information and the actual consumed resource amount of the target recommended information.

[0185] The third recommendation coefficient may be a coefficient preset by the information recommender before the information recommendation. Different recommendation information may correspond to different information recommenders, and the third recommendation coefficients of different recommendation information may be different. The third recommendation coefficient may specifically refer to the ratio of the expected amount of resources obtained by the target recommendation information to the actual amount of resources consumed when the recommendation information is actually recommended. Specifically, the third recommendation coefficient may be the ratio of the expected amount of resources obtained to the actual amount of resources consumed.

[0186] S730. Recommend the target recommendation information based on the target resource consumption amount.

[0187] In this embodiment, before recommending the target recommendation information, a preset consumption resource can be set for the target recommendation information, and at the starting moment of recommending the target recommendation information to the target account, the information recommendation is made based on the preset consumption resource; when the target recommendation information is recommended to the target account, feedback information from the target account regarding the target recommendation information can be received, which may specifically include a first feedback event related to the amount of operating resources and a second feedback event related to the amount of consumed resources, thereby facilitating the information recommender to dynamically adjust the consumption resources for the recommendation based on real-time feedback events, so that the actual consumption resources recommended by the information recommender match the resources obtained by the recommendation, thereby improving the flexibility of resource recommendation.

[0188] The following is a specific example to illustrate the implementation process of the present disclosure. Figure 8 , which may include the recall stage, rough sorting stage, fine sorting stage and data attribution:

[0189] Recall stage: The main recall model can be used to recall candidate recommendation information from the recommendation information set.

[0190] Rough ranking stage: Rough ranking prediction indicators of candidate recommendation information can be predicted based on the first path, the second path and the third path, and the first fused resource amount of the candidate recommendation information can be determined based on the resource processing module, wherein the first fused resource amount is obtained by fusing the first rough ranking prediction resource amount and the second rough ranking prediction resource amount, the first rough ranking prediction resource amount is predicted by the first rough ranking prediction model, and the second rough ranking prediction resource amount is predicted by the second rough ranking prediction model; further, multiple candidate recommendation information can be sorted based on the rough ranking prediction indicators and the first fused resource amount to obtain a rough ranking result, and multiple rough ranking recommendation information are determined to enter the fine ranking stage based on the rough ranking result.

[0191] Refined ranking stage: based on the fourth path, the refined ranking prediction index of the rough ranking recommendation information can be predicted, and based on the resource processing module, the second fused resource amount of the rough ranking recommendation information can be determined, wherein the second fused resource amount is obtained by fusing the first refined ranking prediction resource amount and the second refined ranking prediction resource amount, the first refined ranking prediction resource amount is predicted by the first refined ranking prediction model, and the second refined ranking prediction resource amount is predicted by the second refined ranking prediction model; further, based on the refined ranking prediction index and the second fused resource, multiple rough ranking recommendation information can be sorted to obtain information sorting results.

[0192] Data attribution: The consumption resources set by the information recommender are adjusted through real-time feedback data, and historical data is used as samples to train the model used in the information recommendation process.

[0193] Among them, for the resource processing module, multiple resource processing methods can be further adopted; in addition, for cold start accounts, it can be processed by the default resource processing method of the information recommender; and in this embodiment, for objects that have been activated or have consumed resources in the target object within the historical time period, recommended information for target objects of the same or similar topics can be recommended to them; for target type recommendation information, it can be supported in the recall stage or rough sorting stage to increase its possibility of being recommended, and the target objects in the target type recommendation information support resource consumption operations in the target object, as well as operations on related recommendation information in the target object.

[0194] Fig. 9 is a block diagram of an information processing device according to an exemplary embodiment. Fig. 9 , the device comprises:

[0195] The candidate recommendation information acquisition unit 910 is configured to acquire a plurality of candidate recommendation information; each candidate recommendation information is recommendation information for a target object;

[0196] The information sorting unit 920 is configured to sort the plurality of candidate recommendation information based on a first predicted resource amount corresponding to a target object in each candidate recommendation information and a second predicted resource amount corresponding to the target object, to obtain an information sorting result; the first predicted resource amount represents an operation resource amount generated by the target account performing an operation on the associated recommendation information in the target object, and the second predicted resource amount represents a consumption resource amount corresponding to the resource consumption operation performed by the target account on the target object;

[0197] The target recommendation information determining unit 930 is configured to determine the target recommendation information based on the information sorting result; the target recommendation information is used to push to the target account.

[0198] In an exemplary embodiment, the first predicted resource amount includes a first coarse predicted resource amount and a first fine predicted resource amount, and the second predicted resource amount includes a second coarse predicted resource amount and a second fine predicted resource amount;

[0199] The information sorting unit comprises:

[0200] A first prediction unit is configured to perform a first resource prediction based on the account information of the target account and the feature information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount;

[0201] A rough sorting unit is configured to perform rough sorting on the plurality of candidate recommendation information based on the first rough sorting predicted resource amount and the second rough sorting predicted resource amount to obtain a rough sorting result;

[0202] A first determining unit is configured to determine a plurality of coarse-ranked recommendation information from the plurality of candidate recommendation information based on the coarse-ranking result;

[0203] A second prediction unit is configured to perform a second resource prediction based on the account information of the target account and the feature information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount;

[0204] The fine ranking unit is configured to perform fine ranking on the plurality of coarse ranking recommendation information based on the first fine ranking predicted resource quantity and the second fine ranking predicted resource quantity to obtain the information ranking result.

[0205] In an exemplary embodiment, the first prediction unit is configured to perform:

[0206] Inputting the account information of the target account and the characteristic information of the target object into a first rough ranking prediction model for resource prediction to obtain the first rough ranking predicted resource amount; the first rough ranking prediction model is obtained by training a first preset rough ranking model based on the account information of the sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0207] The account information of the target account and the characteristic information of the target object are input into a second coarse-order prediction model for resource prediction to obtain the second coarse-order predicted resource amount; the second coarse-order prediction model is obtained by training a second preset coarse-order model based on the account information of the sample account, the characteristic information of the sample object, and the consumed resource amount corresponding to the resource consumption operation performed by the sample account in the sample object.

[0208] In an exemplary embodiment, the device model training unit, the model training unit, is configured to perform:

[0209] Acquire a training sample; the training sample includes account information of the sample account, feature information of the sample object, and sample resource consumption amount corresponding to resource consumption operation performed by the sample account in the sample object;

[0210] Inputting the account information of the sample account and the feature information of the sample object into the second preset coarse sorting model to perform resource prediction, and obtaining the training prediction resource amount generated by the sample account operating the associated recommendation information in the sample object;

[0211] Perform loss processing based on the sample consumption resource amount and the training predicted resource amount to determine resource amount loss information;

[0212] The model parameters of the second preset rough-order model are updated based on the resource loss information to obtain the second rough-order prediction model.

[0213] In an exemplary embodiment, the coarse arrangement unit comprises:

[0214] A first resource quantity fusion unit is configured to perform resource quantity fusion processing on the first coarse-order predicted resource quantity and the second coarse-order predicted resource quantity to obtain a first fused resource quantity;

[0215] A first indicator prediction unit is configured to perform indicator prediction on each candidate recommendation information based on the account information of the target account and the feature information of each candidate recommendation information, so as to obtain a plurality of rough ranking prediction indicators corresponding to each candidate recommendation information;

[0216] A first data fusion unit is configured to perform data fusion processing on the plurality of rough ranking prediction indicators based on the first fusion resource amount to obtain a plurality of fusion indicators corresponding to each candidate recommendation information;

[0217] A second determining unit is configured to determine a rough ranking comprehensive index corresponding to each recommendation information based on a plurality of fusion indexes corresponding to each candidate recommendation information;

[0218] The first sorting unit is configured to perform sorting based on the rough sorting comprehensive indicators corresponding to each of the plurality of candidate recommendation information to obtain the rough sorting result.

[0219] In an exemplary embodiment, the first resource quantity fusion unit is configured to execute:

[0220] The ratio of the sum of the first rough predicted resource amount and the second rough predicted resource amount to the first recommendation coefficient is determined as the first fused resource amount; the first recommendation coefficient represents the correlation information between the expected resource amount obtained for each candidate recommended information and the actually consumed resource amount of each candidate recommended information.

[0221] In an exemplary embodiment, the second prediction unit is configured to perform:

[0222] Inputting the account information of the target account and the characteristic information of the target object into a first precise ranking prediction model for resource prediction to obtain a first precise ranking predicted resource amount; the first precise ranking prediction model is obtained by training a first preset precise ranking model based on the account information of a sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommended information in the sample object;

[0223] The account information of the target account and the characteristic information of the target object are input into a second precise ranking prediction model for resource prediction to obtain the second precise ranking predicted resource quantity; the second precise ranking prediction model is obtained by training a second preset precise ranking model based on the account information of the sample account, the characteristic information of the sample object, and the resource consumption quantity corresponding to the resource consumption operation performed by the sample account in the sample object.

[0224] In an exemplary embodiment, the fine sorting unit comprises:

[0225] A second resource quantity fusion unit is configured to perform resource quantity fusion processing on the first refined ranking predicted resource quantity and the second refined ranking predicted resource quantity to obtain a second fused resource quantity;

[0226] A second indicator prediction unit is configured to perform indicator prediction on each rough-ranked recommendation information based on the account information of the target account and the feature information of the candidate recommendation information, so as to obtain a fine-ranking prediction indicator corresponding to each rough-ranked recommendation information;

[0227] A second data fusion unit is configured to perform data fusion processing on the refined ranking prediction index based on the second fusion resource amount to obtain a refined ranking comprehensive index corresponding to each rough ranking recommendation information;

[0228] The second sorting unit is configured to perform sorting based on the respective corresponding comprehensive indicators of the rough sorting recommendation information to obtain the information sorting result.

[0229] In an exemplary embodiment, the second resource quantity fusion unit is configured to execute:

[0230] The ratio of the sum of the first fine-rank predicted resource quantity and the second fine-rank predicted resource quantity to the second recommendation coefficient is determined as the second fused resource quantity; the second recommendation coefficient represents the correlation information between the expected resource quantity of each rough-rank recommended information and the actually consumed resource quantity of each rough-rank recommended information.

[0231] In an exemplary embodiment, the apparatus further comprises:

[0232] a feedback event receiving unit, configured to receive a first feedback event and a second feedback event within a preset time period when the target recommendation information has been pushed to the target account; the first feedback event includes an amount of operation resources generated by the target account performing an operation on the associated recommendation information in the target object of the target recommendation information, and the second feedback event includes an amount of consumed resources corresponding to the resource consumption operation performed by the target account in the target object;

[0233] a target resource consumption amount determination unit, configured to determine the target resource consumption amount based on the ratio of the sum of the operation resource amount and the consumption resource amount to a third recommendation coefficient; the third recommendation coefficient represents the correlation information between the expected resource amount obtained by the target recommendation information and the actual resource consumption amount of the target recommendation information;

[0234] The information recommendation unit is configured to recommend the target recommendation information based on the target resource consumption amount.

[0235] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0236] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided. Optionally, the computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a tape, a floppy disk, an optical data storage device, etc.; when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute any of the methods described above.

[0237] In an exemplary embodiment, a computer program product is also provided, which includes a computer program stored in a readable storage medium, and at least one processor of a computer device reads and executes the computer program from the readable storage medium, so that the device performs any of the above methods.

[0238] Fig.10 is a block diagram of an electronic device for recommending information processing according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as shown in FIG. Fig.10As shown. The electronic device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic 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 electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an information processing method is implemented.

[0239] Those skilled in the art will understand that Fig.10 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present disclosure, and does not constitute a limitation on the electronic device to which the scheme of the present disclosure is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0240] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0241] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An information processing method, characterized in that: include: Acquire multiple candidate recommendation information; each candidate recommendation information is recommendation information for a target object; sorting the plurality of candidate recommendation information based on a first predicted resource amount corresponding to a target object in each candidate recommendation information and a second predicted resource amount corresponding to the target object to obtain an information sorting result; The first predicted resource amount represents the amount of operation resources generated by the target account operating the associated recommendation information in the target object, and the second predicted resource amount represents the amount of consumed resources corresponding to the resource consumption operation performed by the target account on the target object; Determining target recommended information based on the information sorting result; The target recommendation information is used to push to the target account.

2. The method according to claim 1, characterized in that The first predicted resource amount includes a first rough predicted resource amount and a first fine predicted resource amount, and the second predicted resource amount includes a second rough predicted resource amount and a second fine predicted resource amount; The sorting of the plurality of candidate recommendation information based on the first predicted resource amount corresponding to the target object in each candidate recommendation information and the second predicted resource amount corresponding to the target object to obtain the information sorting result includes: Performing a first resource prediction based on the account information of the target account and the characteristic information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount; Based on the first rough ranking predicted resource amount and the second rough ranking predicted resource amount, the plurality of candidate recommendation information are roughly ranked to obtain a rough ranking result; Determining a plurality of rough-ranked recommendation information from the plurality of candidate recommendation information based on the rough-ranking result; Performing a second resource prediction based on the account information of the target account and the characteristic information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount; Based on the first fine ranking predicted resource quantity and the second fine ranking predicted resource quantity, the plurality of coarse ranking recommendation information are finely ranked to obtain the information ranking result.

3. The method according to claim 2, characterized in that The performing a first resource prediction based on the account information of the target account and the feature information of the target object in each candidate recommendation information to obtain the first rough prediction resource amount and the second rough prediction resource amount includes: Inputting the account information of the target account and the characteristic information of the target object into a first rough ranking prediction model for resource prediction to obtain the first rough ranking predicted resource amount; the first rough ranking prediction model is obtained by training a first preset rough ranking model based on the account information of the sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommendation information in the sample object; The account information of the target account and the characteristic information of the target object are input into a second coarse-order prediction model for resource prediction to obtain the second coarse-order predicted resource amount; the second coarse-order prediction model is obtained by training a second preset coarse-order model based on the account information of the sample account, the characteristic information of the sample object, and the consumed resource amount corresponding to the resource consumption operation performed by the sample account in the sample object.

4. The method according to claim 3, characterized in that The method further comprises: Acquire a training sample; the training sample includes account information of the sample account, feature information of the sample object, and sample resource consumption amount corresponding to resource consumption operation performed by the sample account in the sample object; Inputting the account information of the sample account and the feature information of the sample object into the second preset coarse sorting model to perform resource prediction, and obtaining the training prediction resource amount generated by the sample account operating the associated recommendation information in the sample object; Perform loss processing based on the sample consumption resource amount and the training predicted resource amount to determine resource amount loss information; The model parameters of the second preset rough-order model are updated based on the resource loss information to obtain the second rough-order prediction model.

5. The method according to claim 2, characterized in that: The step of roughly ranking the candidate recommendation information based on the first roughly ranked predicted resource amount and the second roughly ranked predicted resource amount to obtain a roughly ranked result includes: Performing resource quantity fusion processing on the first coarsely predicted resource quantity and the second coarsely predicted resource quantity to obtain a first fused resource quantity; Based on the account information of the target account and the feature information of each candidate recommendation information, an index prediction is performed on each candidate recommendation information to obtain a plurality of rough ranking prediction indexes corresponding to each candidate recommendation information; Based on the first fusion resource amount, data fusion processing is performed on the multiple rough prediction indicators respectively to obtain multiple fusion indicators corresponding to each candidate recommendation information; Based on the multiple fusion indicators corresponding to each candidate recommendation information, determine the rough ranking comprehensive indicator corresponding to each recommendation information; The rough ranking result is obtained by sorting the plurality of candidate recommendation information based on the rough ranking comprehensive indicators corresponding to each of the plurality of candidate recommendation information.

6. The method according to claim 5, characterized in that The performing resource amount fusion processing on the first coarsely predicted resource amount and the second coarsely predicted resource amount to obtain a first fused resource amount includes: The ratio of the sum of the first rough predicted resource amount and the second rough predicted resource amount to the first recommendation coefficient is determined as the first fused resource amount; the first recommendation coefficient represents the correlation information between the expected resource amount obtained for each candidate recommended information and the actually consumed resource amount of each candidate recommended information.

7. The method according to claim 2, characterized in that The performing second resource prediction based on the account information of the target account and the characteristic information of the target object in each rough ranking recommendation information to obtain the first fine ranking predicted resource amount and the second fine ranking predicted resource amount includes: Inputting the account information of the target account and the characteristic information of the target object into a first precise ranking prediction model for resource prediction to obtain a first precise ranking predicted resource amount; the first precise ranking prediction model is obtained by training a first preset precise ranking model based on the account information of a sample account, the characteristic information of the sample object, and the operation resource amount generated by the sample account operating the associated recommended information in the sample object; The account information of the target account and the characteristic information of the target object are input into a second precise ranking prediction model for resource prediction to obtain the second precise ranking predicted resource quantity; the second precise ranking prediction model is obtained by training a second preset precise ranking model based on the account information of the sample account, the characteristic information of the sample object, and the resource consumption quantity corresponding to the resource consumption operation performed by the sample account in the sample object.

8. The method according to claim 7, characterized in that The step of finely sorting the plurality of coarsely ranked recommendation information based on the first finely ranked predicted resource amount and the second finely ranked predicted resource amount to obtain the information sorting result includes: Performing resource quantity fusion processing on the first refined predicted resource quantity and the second refined predicted resource quantity to obtain a second fused resource quantity; Based on the account information of the target account and the feature information of the candidate recommendation information, an index prediction is performed on each rough-ranked recommendation information to obtain a fine-ranking prediction index corresponding to each rough-ranked recommendation information; Performing data fusion processing on the refined ranking prediction index based on the second fused resource quantity to obtain a refined ranking comprehensive index corresponding to each rough ranking recommendation information; The information sorting result is obtained by sorting the information based on the respective corresponding comprehensive indicators of the rough ranking of the plurality of recommended information.

9. The method according to claim 8, characterized in that The performing resource amount fusion processing on the first refined predicted resource amount and the second refined predicted resource amount to obtain a second fused resource amount includes: The ratio of the sum of the first fine-rank predicted resource quantity and the second fine-rank predicted resource quantity to the second recommendation coefficient is determined as the second fused resource quantity; the second recommendation coefficient represents the correlation information between the expected resource quantity of each rough-rank recommended information and the actually consumed resource quantity of each rough-rank recommended information.

10. The method according to claim 1, characterized in that The method further comprises: In the case where the target recommendation information has been pushed to the target account, a first return event and a second return event are received within a preset time period; the first return event includes the amount of operation resources generated by the target account performing an operation on the associated recommendation information in the target object of the target recommendation information, and the second return event includes the amount of consumed resources corresponding to the resource consumption operation performed by the target account in the target object; Determine a target consumed resource amount based on a ratio of the sum of the operating resource amount and the consumed resource amount to a third recommendation coefficient; the third recommendation coefficient represents correlation information between the expected acquired resource amount of the target recommended information and the actual consumed resource amount of the target recommended information; The target recommendation information is recommended based on the target resource consumption amount.

11. An information processing device, characterized in that: include: A candidate recommendation information acquisition unit is configured to acquire a plurality of candidate recommendation information; each candidate recommendation information is recommendation information for a target object; an information sorting unit, configured to sort the plurality of candidate recommendation information based on a first predicted resource amount corresponding to a target object in each candidate recommendation information and a second predicted resource amount corresponding to the target object, to obtain an information sorting result; The first predicted resource amount represents the amount of operation resources generated by the target account operating the associated recommendation information in the target object, and the second predicted resource amount represents the amount of consumed resources corresponding to the resource consumption operation performed by the target account on the target object; a target recommendation information determining unit, configured to determine the target recommendation information based on the information sorting result; The target recommendation information is used to push to the target account.

12. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the information processing method according to any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the information processing method according to any one of claims 1 to 10.

14. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the information processing method according to any one of claims 1 to 10 is implemented.

Images