Business processing, method and device for generating combined gain analysis model, and electronic device

Abstract

The present disclosure relates to a service processing, combined gain analysis model generation method and apparatus, and an electronic device. The service processing method includes obtaining target object attribute data of a target object; based on a combined gain analysis model corresponding to multiple service operations and the target object attribute data, performing gain analysis on the target object to obtain a target service metric gain of the target object under the multiple service operations, where the target service metric gain represents the sensitivity of the target object to the multiple service operations; based on the target service metric gain, determining a target service operation corresponding to the target object from the multiple service operations; and performing service processing on the target object based on the target service operation. Using the embodiments of the present disclosure can effectively reduce the computational complexity in the service processing process, greatly improve the effectiveness of the target service operation determined based on the target service metric gain, and improve the accuracy of service decision-making and the efficiency of service processing.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and particularly to a method and apparatus for business processing and joint gain analysis model generation, and an electronic device. Background Art

[0002] With the development of artificial intelligence technology, artificial intelligence technology has been widely applied in multiple fields. The random forest model in artificial intelligence technology is a machine learning method that uses multiple decision trees to discriminate and classify data, and is often applied to various classification tasks.

[0003] In the related art, in business processing such as distributing coupons to users, multiple decision trees in the random forest model are often combined to determine which business operation to execute (for example, distributing coupons of which level to users); a decision tree is learned through a customized splitting criterion, so that each sub-population within the feature space divided by the tree model has as homogeneous causal effects as possible. However, the random forest model in the related art is only applicable to binary processing situations, and for multi-value processing scenarios with multiple business operations (distributing coupons of multiple levels), decision trees can only be constructed separately for each business operation, resulting in high computational complexity and low efficiency in the subsequent business processing process. Moreover, only part of the sample information is used when training each decision tree, and different samples are used to predict the sensitivity of different levels, which also brings problems such as poor effectiveness of business processing and inability to better meet user needs. Summary of the Invention

[0004] The present disclosure provides a method and apparatus for business processing and joint gain analysis model generation, and an electronic device, so as to at least solve the problems of high computational complexity, low efficiency, poor effectiveness of business processing, and inability to better meet user needs in the related art. The technical solutions of the present disclosure are as follows:

[0005] According to the first aspect of the embodiments of the present disclosure, a business processing method is provided, including:

[0006] Obtain target object attribute data of a target object;

[0007] Based on a joint gain analysis model corresponding to multiple business operations and the target object attribute data, perform gain analysis on the target object to obtain target business metric gains of the target object under multiple business operations, where the target business metric gains represent the sensitivity of the target object to multiple business operations; the multiple business operations are multiple operations for increasing business metric data corresponding to the target object;

[0008] Based on the target business metric gains, determine a target business operation corresponding to the target object from multiple business operations;

[0009] Perform business processing on the target object based on the target business operation.

[0010] In an optional embodiment, the determining the target business operation corresponding to the target object from multiple business operations based on the target business metric gain includes:

[0011] Obtain the preset resource consumption information and preset resource upper limit corresponding to multiple business operations;

[0012] Based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain, determine a target business decision parameter, where the target business decision parameter characterizes the business metric gain brought by the resource consumption information generated by allocating multiple business operations to the target object under the constraint of the preset resource upper limit;

[0013] Determine the target business operation according to the target business decision parameter.

[0014] In an optional embodiment, the determining the target business decision parameter based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain includes:

[0015] Based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain, create a dual objective function with the target business decision parameter as the dependent variable;

[0016] According to the preset resource consumption information and the target business metric gain, determine a first business decision parameter and a second business decision parameter;

[0017] Determine an initial business decision parameter according to the first business decision parameter and the second business decision parameter;

[0018] Determine the target derivative of the dual objective function when the target business decision parameter is equal to the initial business decision parameter;

[0019] Based on the target derivative, the first business decision parameter, and the second business decision parameter, determine the target business decision parameter in the dual objective function.

[0020] In an optional embodiment, the joint gain analysis model includes multiple target decision trees;

[0021] The performing gain analysis on the target object based on the joint gain analysis model corresponding to multiple business operations and the target object attribute data to obtain the target business metric gain of the target object includes:

[0022] According to the target object attribute data, determine the associated leaf nodes of the target object in multiple target decision trees from the multiple target decision trees. Each associated leaf node includes associated sample objects corresponding to multiple business operations.

[0023] Obtain the sample business metric data of multiple associated sample objects in the associated leaf nodes and the operation coding information of multiple business operations.

[0024] Based on the number of objects of each associated sample object in the associated leaf node, determine the weight information corresponding to each associated sample object.

[0025] Based on the weight information, the operation coding information, and the sample business metric data, determine the target business metric gain.

[0026] According to the second aspect of the embodiments of the present disclosure, a method for generating a joint gain analysis model is provided, including:

[0027] Obtain the sample business metric data of the sample object and the sample object attribute data of the sample object. The sample object includes objects corresponding to multiple business operations and objects corresponding to no business operation. The multiple business operations are multiple operations for increasing the business metric data corresponding to the target object.

[0028] Based on the sample object, construct the root nodes of multiple target decision trees respectively. Each root node includes a target sample object, and each target sample object includes objects corresponding to multiple business operations and objects corresponding to no business operation.

[0029] Based on the sample object attribute data of each target sample object and the sample business metric data of each target sample object, perform node splitting processing on each root node to generate a joint gain analysis model corresponding to multiple business operations.

[0030] The joint gain analysis model is used to analyze the target business metric gain of the target object under multiple business operations.

[0031] In an optional embodiment, the performing node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample business metric data of each target sample object to generate a joint gain analysis model corresponding to multiple business operations includes:

[0032] Take each root node as the current parent node corresponding to multiple target decision trees respectively.

[0033] Perform node splitting on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample business metric data of the target sample object in the current parent node, to construct target child nodes of the current parent node in multiple of the target decision trees, where the sample objects in the target child nodes include objects corresponding to multiple of the business operations and objects without business operations;

[0034] Update the target child nodes as the current parent node, and repeat the step of performing node splitting on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample business metric data of the target sample object in the current parent node, to construct target child nodes of the current parent node in multiple of the target decision trees, until a preset splitting convergence condition is reached;

[0035] Use the multiple target decision trees obtained when the preset splitting convergence condition is reached as the joint gain analysis model.

[0036] In an alternative embodiment, the sample object attribute data includes attribute data corresponding to at least one object attribute; the preset splitting convergence condition is that all of the at least one object attribute have been used for node splitting processing;

[0037] The performing node splitting on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample business metric data of the target sample object in the current parent node, to construct target child nodes of the current parent node in multiple of the target decision trees includes:

[0038] Determine target attribute data from the sample object attribute data of the sample objects in the current parent node, where the target attribute data is attribute data corresponding to a target object attribute, and the target object attribute is an object attribute among the at least one object attribute that has not been used for node splitting processing;

[0039] Obtain operation coding information corresponding to multiple of the business operations;

[0040] Based on the target attribute data, the operation coding information, and the sample business metric data of the sample objects in the current parent node, perform splitting processing on the sample objects in the current parent node to construct target child nodes of the current parent node in the target decision tree.

[0041] In an alternative embodiment, the based on the target attribute data, the operation coding information, and the sample business metric data of the sample objects in the current parent node, perform splitting processing on the sample objects in the current parent node to construct target child nodes of the current parent node in the target decision tree includes:

[0042] Perform splitting processing on the sample objects in the current parent node according to the target attribute data to obtain multiple initial child node pairs of the current parent node;

[0043] According to the operation coding information and the sample business index data of the sample objects in each pair of the initial child nodes, determine the predicted business index data of the sample objects in each pair of the initial child nodes under multiple business operations;

[0044] Based on the preset business index data, determine multiple primary candidate child node pairs from multiple initial child node pairs;

[0045] According to the operation coding information and the sample business index data of the sample objects in each pair of the primary candidate child nodes, determine the average business index gain corresponding to each primary candidate child node and the operation business index gain corresponding to the sample objects in each primary candidate child node; the average business index gain characterizes the sensitivity of the sample objects in each primary candidate child node to multiple business operations; the operation business index gain characterizes the sensitivity of the sample objects in each primary candidate child node to each business operation;

[0046] Based on the average business index gain and the operation business index gain, determine the target child nodes of the current parent node from multiple primary candidate child node pairs.

[0047] In an alternative embodiment, the determining multiple primary candidate child node pairs from multiple initial child node pairs based on the preset business index data includes:

[0048] Perform node heterogeneity analysis on multiple initial child node pairs according to the preset business index data to obtain node heterogeneity analysis results corresponding to multiple initial child node pairs, and the node heterogeneity analysis results characterize the differences between each pair of initial child nodes;

[0049] According to the node heterogeneity analysis results, determine multiple primary candidate child node pairs from multiple initial child node pairs.

[0050] In an alternative embodiment, the determining the target child nodes of the current parent node from multiple primary candidate child node pairs based on the average business index gain and the operation business index gain includes:

[0051] Perform business gain difference analysis on the sample objects in multiple primary candidate child node pairs according to the average business index gain and the operation business index gain to obtain business gain difference information corresponding to the sample objects in multiple primary candidate child node pairs, and the business gain difference information characterizes the sensitive difference degree of the sample objects in multiple primary candidate child node pairs to multiple business operations;

[0052] Based on the service gain difference information, determine the target child nodes of the current parent node from multiple pairs of the initially selected child nodes.

[0053] In an alternative embodiment, the method further includes:

[0054] Based on the sample service metric data corresponding to the leaf nodes of multiple target decision trees in the joint gain analysis model and the operation coding information of multiple types of the service operations, determine the sample service metric gain of the sample object;

[0055] Based on the sample service metric gain, determine the sample service operation corresponding to the sample object from multiple types of the service operations;

[0056] Divide the sample objects corresponding to the same sample service operation into the same sample object group;

[0057] Determine the control object group corresponding to the same service operation in the preset control group;

[0058] Determine the intersection object group between the sample object groups corresponding to each service operation;

[0059] Based on the sample service metric data corresponding to the intersection object group, the sample service metric data corresponding to the sample object group, the number of the first objects in the intersection object group, and the number of the second objects in the sample object group, determine the service measurement index data;

[0060] Wherein, the service measurement index data is used to measure the accuracy of the gain analysis of the joint gain analysis model and the rationality of the sample service operation determined based on the sample service metric gain.

[0061] In an alternative embodiment, the determining the sample service operation corresponding to the sample object from multiple types of the service operations based on the sample service metric gain includes:

[0062] Obtain the preset resource consumption information and the preset resource upper limit corresponding to multiple types of the service operations;

[0063] Based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain, determine the sample service decision parameter, where the sample service decision parameter represents the service metric gain brought by allocating multiple types of the service operations to the sample object under the constraint of the preset resource upper limit;

[0064] Determine the sample service operation according to the sample service decision parameter.

[0065] In an alternative embodiment, determining the sample service decision parameter based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain includes:

[0066] Based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain, create a dual objective function with the sample service decision parameter as the dependent variable;

[0067] According to the preset resource consumption information and the sample service metric gain, determine a first sample service decision parameter and a second sample service decision parameter;

[0068] Determine an initial sample service decision parameter based on the first sample service decision parameter and the second sample service decision parameter;

[0069] Determine the sample derivative of the dual objective function when the sample service decision parameter is equal to the initial sample service decision parameter;

[0070] Based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter, determine the sample service decision parameter in the dual objective function.

[0071] According to the third aspect of the embodiments of the present disclosure, there is provided a service processing device, including:

[0072] A target object attribute data acquisition module, configured to acquire target object attribute data of a target object;

[0073] A gain analysis module, configured to perform gain analysis on the target object based on a joint gain analysis model corresponding to multiple service operations and the target object attribute data, to obtain a target service metric gain of the target object under the multiple service operations, where the target service metric gain represents the sensitivity of the target object to the multiple service operations; the multiple service operations are multiple operations for increasing service metric data corresponding to the target object;

[0074] A target service operation determination module, configured to determine a target service operation corresponding to the target object from the multiple service operations based on the target service metric gain;

[0075] A service processing module, configured to perform service processing on the target object based on the target service operation.

[0076] In an alternative embodiment, the target service operation determination module includes:

[0077] The first information acquisition unit is configured to acquire the preset resource consumption information and the preset resource upper limit corresponding to multiple said service operations;

[0078] The target service decision parameter determination unit is configured to determine target service decision parameters based on the preset resource upper limit, the preset resource consumption information, and the target service index gain. The target service decision parameters represent the service index gain brought by allocating the resource consumption information generated by multiple said service operations to the target object under the constraint of the preset resource upper limit;

[0079] The target service operation determination unit is configured to determine the target service operation according to the target service decision parameters.

[0080] In an optional embodiment, the target service decision parameter determination unit includes:

[0081] The first dual objective function creation unit is configured to create a dual objective function with the target service decision parameters as the dependent variables based on the preset resource upper limit, the preset resource consumption information, and the target service index gain;

[0082] The first service decision parameter determination unit is configured to determine a first service decision parameter and a second service decision parameter according to the preset resource consumption information and the target service index gain;

[0083] The initial service decision parameter determination unit is configured to determine the initial service decision parameters according to the first service decision parameter and the second service decision parameter;

[0084] The target derivative determination unit is configured to determine the target derivative of the dual objective function when the target service decision parameters are equal to the initial service decision parameters;

[0085] The target service decision parameter determination unit is configured to determine the target service decision parameters in the dual objective function based on the target derivative, the first service decision parameter, and the second service decision parameter.

[0086] In an optional embodiment, the joint gain analysis model includes multiple target decision trees;

[0087] The gain analysis module includes:

[0088] The associated leaf node determination unit is configured to determine the associated leaf nodes of the target object in multiple said target decision trees according to the target object attribute data. Each associated leaf node includes associated sample objects corresponding to multiple said service operations;

[0089] A sample service metric data acquisition unit, configured to execute operations of acquiring sample service metric data of multiple associated sample objects in the associated leaf nodes and operation coding information of multiple types of the service operations;

[0090] A weight information determination unit, configured to execute determining weight information corresponding to each of the associated sample objects based on the number of objects of each associated sample object in the associated leaf nodes;

[0091] A target service metric gain determination unit, configured to execute determining the target service metric gain based on the weight information, the operation coding information, and the sample service metric data.

[0092] According to a fourth aspect of the embodiments of the present disclosure, there is provided a combined gain analysis model generation device, including:

[0093] A sample data acquisition module, configured to execute operations of acquiring sample service metric data of sample objects and sample object attribute data of the sample objects, where the sample objects include objects corresponding to multiple service operations and objects corresponding to no service operations; the multiple service operations are multiple operations for increasing service metric data corresponding to the sample objects;

[0094] A root node construction module, configured to execute respectively constructing root nodes of multiple target decision trees based on the sample objects, each root node including a target sample object, and each target sample object including objects corresponding to multiple service operations and objects corresponding to no service operations;

[0095] A combined gain analysis model generation module, configured to execute performing node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample service metric data of each target sample object, and generating combined gain analysis models corresponding to multiple service operations;

[0096] The combined gain analysis model is used to analyze the target service metric gain of a target object under multiple service operations.

[0097] In an optional embodiment, the combined gain analysis model generation module includes:

[0098] A current parent node determination unit, configured to execute respectively using each root node as the current parent node corresponding to multiple target decision trees;

[0099] A node splitting processing unit, configured to perform node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in multiple target decision trees, and the sample objects in the target child nodes include objects corresponding to multiple business operations and objects without business operations;

[0100] A current parent node updating unit, configured to perform updating the target child nodes as the current parent node;

[0101] An iterative processing unit, configured to perform repeating the step of performing node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in multiple target decision trees, until a preset splitting convergence condition is reached;

[0102] A joint gain analysis model determining unit, configured to perform using multiple target decision trees obtained when the preset splitting convergence condition is reached as the joint gain analysis model.

[0103] In an optional embodiment, the sample object attribute data includes attribute data corresponding to at least one object attribute; the preset splitting convergence condition is that all of the at least one object attribute has been used for node splitting processing;

[0104] The node splitting processing unit includes:

[0105] A target attribute data determining unit, configured to perform determining target attribute data from the sample object attribute data of the sample objects in the current parent node, where the target attribute data is attribute data corresponding to a target object attribute, and the target object attribute is an object attribute that has not been used for node splitting processing among the at least one object attribute;

[0106] An operation coding information obtaining unit, configured to perform obtaining operation coding information corresponding to multiple business operations;

[0107] A sample splitting processing unit, configured to perform splitting processing on the sample objects in the current parent node based on the target attribute data, the operation coding information, and the sample service index data of the sample objects in the current parent node, so as to construct target child nodes of the current parent node in the target decision tree.

[0108] In an optional embodiment, the sample splitting processing unit includes:

[0109] Multiple initial child node pair generation units, configured to perform splitting processing on the sample objects in the current parent node according to the target attribute data to obtain multiple initial child node pairs of the current parent node;

[0110] Predicted business metric data determination unit, configured to determine the predicted business metric data of the sample objects in each pair of the initial child nodes under multiple business operations according to the operation coding information and the sample business metric data of the sample objects in each pair of the initial child nodes;

[0111] Primary selected child node pair determination unit, configured to determine multiple primary selected child node pairs from multiple initial child node pairs based on preset business metric data;

[0112] Business metric gain determination unit, configured to determine the average business metric gain corresponding to each primary selected child node and the operation business metric gain corresponding to the sample objects in each primary selected child node according to the operation coding information and the sample business metric data of the sample objects in each pair of the primary selected child nodes; the average business metric gain represents the sensitivity of the sample objects in each primary selected child node to multiple business operations; the operation business metric gain represents the sensitivity of the sample objects in each primary selected child node to each business operation;

[0113] Target child node determination unit, configured to determine the target child node of the current parent node from multiple primary selected child node pairs based on the average business metric gain and the operation business metric gain.

[0114] In an alternative embodiment, the primary selected child node pair determination unit includes:

[0115] Node heterogeneity analysis unit, configured to perform node heterogeneity analysis on multiple initial child node pairs according to preset business metric data to obtain node heterogeneity analysis results corresponding to multiple initial child node pairs, and the node heterogeneity analysis results represent the differences between each pair of initial child nodes;

[0116] Primary selected child node pair determination subunit, configured to determine multiple primary selected child node pairs from multiple initial child node pairs according to the node heterogeneity analysis results.

[0117] In an alternative embodiment, the target child node determination unit includes:

[0118] The service gain difference sub-unit is configured to perform service gain difference analysis on sample objects in multiple pairs of the primary selected sub-nodes according to the average service metric gain and the operational service metric gain, so as to obtain service gain difference information corresponding to the sample objects in multiple pairs of the primary selected sub-nodes, where the service gain difference information characterizes the sensitive difference degree of the sample objects in multiple pairs of the primary selected sub-nodes to multiple types of the service operations;

[0119] The target sub-node determination sub-unit is configured to determine the target sub-node of the current parent node from multiple pairs of the primary selected sub-nodes according to the service gain difference information.

[0120] In an optional embodiment, the apparatus further includes:

[0121] The sample service metric gain determination module is configured to determine the sample service metric gain of the sample object based on the sample service metric data corresponding to the leaf nodes of multiple target decision trees in the joint gain analysis model and the operation coding information of multiple types of the service operations;

[0122] The sample service operation determination module is configured to determine the sample service operation corresponding to the sample object from multiple types of the service operations based on the sample service metric gain;

[0123] The first object grouping processing module is configured to divide the sample objects corresponding to the same sample service operation into the same sample object group;

[0124] The second object grouping processing module is configured to determine the control object group corresponding to the same service operation in the preset control group;

[0125] The intersection object group determination module is configured to determine the intersection object group between the sample object groups corresponding to each service operation;

[0126] The service measurement index data determination module is configured to determine service measurement index data based on the sample service metric data corresponding to the intersection object group, the sample service metric data corresponding to the sample object group, the number of the first objects in the intersection object group, and the number of the second objects in the sample object group;

[0127] Wherein, the service measurement index data is used to measure the accuracy of the gain analysis of the joint gain analysis model and the rationality of the sample service operation determined based on the sample service metric gain.

[0128] In an optional embodiment, the sample service operation determination module includes:

[0129] A second information acquisition unit, configured to acquire preset resource consumption information and preset resource upper limits corresponding to a plurality of the service operations;

[0130] A sample service decision parameter determination unit, configured to determine sample service decision parameters based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain, where the sample service decision parameters represent the service metric gain brought about by allocating the resource consumption information generated by a plurality of the service operations to the sample object under the constraint of the preset resource upper limit;

[0131] A sample service operation determination unit, configured to determine the sample service operation according to the sample service decision parameters.

[0132] In an alternative embodiment, the sample service decision parameter determination unit includes:

[0133] A second dual objective function creation unit, configured to create a dual objective function with the sample service decision parameters as the dependent variables based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain;

[0134] A second service decision parameter determination unit, configured to determine a first sample service decision parameter and a second sample service decision parameter according to the preset resource consumption information and the sample service metric gain;

[0135] An initial sample service decision parameter determination unit, configured to determine initial sample service decision parameters according to the first sample service decision parameter and the second sample service decision parameter;

[0136] A sample derivative determination unit, configured to determine the sample derivative of the dual objective function when the sample service decision parameters are equal to the initial sample service decision parameters;

[0137] A sample service decision parameter determination unit, configured to determine the sample service decision parameters in the dual objective function based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter.

[0138] According to a fifth aspect of the embodiments of the present disclosure, there is provided an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to execute the instructions to implement the method according to each item in the first aspect or the second aspect as described above.

[0139] According to a sixth aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, which when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to execute the methods described in each item of the first aspect or the second aspect of the embodiments of the present disclosure.

[0140] According to a seventh aspect of the embodiments of the present disclosure, there is provided a computer program product including instructions, which when running on a computer, enables the computer to execute the methods described in each item of the first aspect or the second aspect of the embodiments of the present disclosure.

[0141] The technical solutions provided by the embodiments of the present disclosure at least bring the following beneficial effects:

[0142] During the business processing, by combining the joint gain analysis models corresponding to multiple business operations to perform gain analysis on the target object for the business, the computational complexity during the business processing can be effectively reduced, and the target business metric gain of the target object under multiple business operations can be obtained quickly and accurately. Moreover, by combining this target business metric gain, the sensitivity of the target object to multiple business operations can be accurately characterized. Furthermore, the effectiveness of the target business operation determined based on the target business metric gain can be greatly improved, the business decision-making accuracy and business processing efficiency can be enhanced, and thus the business processing can be better carried out according to the user requirements.

[0143] 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

[0144] The drawings herein are incorporated into the specification and constitute a part of the 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 of the present disclosure.

[0145] Figure 1 is a schematic diagram of an application environment shown according to an exemplary embodiment;

[0146] Figure 2 is a flowchart of a method for generating a joint gain analysis model shown according to an exemplary embodiment;

[0147] Figure 3 is a flowchart of splitting each root node based on the sample object attribute data of each target sample object and the sample business metric data of each target sample object to generate a joint gain analysis model corresponding to multiple business operations;

[0148] Figure 4A flowchart for splitting a current parent node based on sample object attribute data of a target sample object and sample service metric data of the target sample object in the current parent node to construct target child nodes of the current parent node in multiple target decision trees;

[0149] Figure 5 A flowchart for splitting sample objects in a current parent node based on target attribute data, operation coding information, and sample service metric data of the sample objects in the current parent node to construct target child nodes of the current parent node in a target decision tree;

[0150] Figure 6 A schematic diagram of a target decision tree provided according to an exemplary embodiment;

[0151] Figure 7 A flowchart for model measurement and analysis shown according to an exemplary embodiment;

[0152] Figure 8 A flowchart for determining a sample service operation corresponding to a sample object from multiple service operations based on sample service metric gain shown according to an exemplary embodiment;

[0153] Figure 9 A flowchart for determining sample service decision parameters based on a preset resource upper limit, preset resource consumption information, and sample service metric gain shown according to an exemplary embodiment;

[0154] Figure 10 A flowchart for a service processing method shown according to an exemplary embodiment;

[0155] Figure 11 A flowchart for performing gain analysis on a target object based on a joint gain analysis model corresponding to multiple service operations and target object attribute data to obtain target service metric gain of the target object shown according to an exemplary embodiment;

[0156] Figure 12 A flowchart for determining a target service operation corresponding to a target object from multiple service operations based on target service metric gain shown according to an exemplary embodiment;

[0157] Figure 13 A schematic diagram of a service platform performing service processing provided according to an embodiment;

[0158] Figure 14 A block diagram of a service processing device shown according to an exemplary embodiment;

[0159] Figure 15It is a block diagram of a combined gain analysis model generation device shown according to an exemplary embodiment;

[0160] Figure 16 It is a block diagram of an electronic device for business processing shown according to an exemplary embodiment;

[0161] Figure 17 It is a block diagram of an electronic device for combined gain analysis model generation shown according to an exemplary embodiment. Detailed implementation manners

[0162] In order to enable those of ordinary skill 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 with reference to the accompanying drawings.

[0163] 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 do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances 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. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0164] It should be noted that the user information (including but not limited to user device information, user personal attribute data, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in the present disclosure are all information and data authorized by the user or fully authorized by all parties.

[0165] Please refer to Figure 1 , Figure 1 It is a schematic diagram of an application environment shown according to an exemplary embodiment. As Figure 1 shown, the application environment may include a server 100 and a terminal 200.

[0166] In a specific embodiment, the server 100 may be used to generate a combined gain analysis model corresponding to multiple service operations under a target service, and the terminal 200 may provide a service to a user in combination with the combined gain analysis model.

[0167] In an alternative embodiment, the above-mentioned server 100 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0168] In an alternative embodiment, the above-mentioned terminal 200 may include, but is not limited to, electronic devices such as smart phones, desktop computers, tablet computers, laptop computers, smart speakers, digital assistants, augmented reality (AR) / virtual reality (VR) devices, smart wearable devices, etc., or may also be software running on the above-mentioned electronic devices, such as application programs, etc. Optionally, the operating systems running on the electronic devices may include, but are not limited to, Android system, IOS system, Linux, Windows, etc.

[0169] In addition, it should be noted that Figure 1 The application environment shown is only one provided by the present disclosure. In actual applications, other application environments may also be included, for example, more terminals may be included.

[0170] In the embodiments of this specification, the above-mentioned server 100 and terminal 200 may be directly or indirectly connected through wired or wireless communication methods, and the present disclosure does not limit this.

[0171] Figure 2 It is a flowchart of a method for generating a joint gain analysis model shown according to an exemplary embodiment. As Figure 2 shown, this service processing method is used in electronic devices such as terminals and servers, and includes the following steps.

[0172] In step S201, obtain the sample service metric data of the sample object and the sample object attribute data of the sample object.

[0173] In a specific embodiment, the above-mentioned sample object may include objects corresponding to multiple service operations and objects without service operations. Specifically, the multiple service operations may be multiple operations for increasing the service metric data corresponding to the object in the target service. Specifically, the object in the target service may include the above-mentioned sample object. Optionally, the object may be a user account in the service platform corresponding to the target service. The service metric data may be metric data representing the situation of the target service. The sample object attribute data may be information capable of describing the sample object, such as age, gender, etc.

[0174] In a specific embodiment, each object corresponding to the multiple service operations is assigned a service operation during the actual service execution process. Each object without a service operation is not assigned a service operation during the actual service execution process.

[0175] In a specific embodiment, taking the target service as the multimedia resource push service as an example, multiple service operations can be operations of distributing different quantities of virtual resources. The service metric data can be the duration of browsing multimedia resources. Specifically, the multimedia resources can include static media resources such as pictures and texts, and can also include dynamic media resources such as videos.

[0176] In addition, it should be noted that the above target service is only an example, and in actual applications, it can also include other services, such as the preset item acquisition service. Correspondingly, the service metric data can be the number of times of executing the acquisition of the preset item, and multiple service operations can be operations of reducing different quantities of virtual resources required for acquiring the preset item.

[0177] In step S203, based on the sample objects, the root nodes of multiple target decision trees are respectively constructed.

[0178] In a specific embodiment, each root node (i.e., the root node of each target decision tree) includes a target sample object, and each target sample object includes objects corresponding to multiple service operations and objects without service operations.

[0179] In a specific embodiment, a certain number of objects (target sample objects) can be randomly selected with replacement from the sample objects for each decision tree, and the root nodes of each target decision tree are constructed.

[0180] In the embodiments of this specification, by allocating objects corresponding to multiple service operations and objects without service operations in the root nodes of each decision tree, during the process of generating the joint gain analysis model, based on the same sample objects, the service metric gain analysis corresponding to different service operations can be performed, thereby improving the accuracy of the gain analysis of the subsequent joint gain analysis model, and at the same time reducing the computational complexity of the gain analysis and improving the processing efficiency.

[0181] In step S205, based on the sample object attribute data of each target sample object and the sample service metric data of each target sample object, each root node is split to generate a joint gain analysis model corresponding to multiple service operations.

[0182] In an alternative embodiment, the above joint gain analysis model can be used to analyze the target service metric gain of the target object under multiple service operations. Specifically, the joint gain analysis model can include multiple target decision trees. Each node in each target decision tree includes objects corresponding to multiple service operations and objects corresponding to no service operations. Specifically, the target object can be a user account that needs to perform service operations in the service platform.

[0183] In an alternative embodiment, such as Figure 3As shown, the above-mentioned node splitting process for each root node based on the sample object attribute data of each target sample object and the sample business index data of each target sample object to generate a joint gain analysis model corresponding to multiple business operations may include the following steps:

[0184] In step S301, each root node is respectively used as the current parent node corresponding to multiple target decision trees;

[0185] In step S303, based on the sample object attribute data of the target sample object in the current parent node and the sample business index data of the target sample object in the current parent node, the current parent node is split to construct the target child nodes of the current parent node in multiple target decision trees.

[0186] In a specific embodiment, the sample objects in the above-mentioned target child nodes may include objects corresponding to multiple business operations and objects without business operations;

[0187] In an optional embodiment, as Figure 4 shown, the above-mentioned sample object attribute data may include attribute data corresponding to at least one object attribute; the above-mentioned preset splitting convergence condition may be that at least one object attribute has been used for node splitting processing; correspondingly, the above-mentioned splitting the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample business index data of the target sample object in the current parent node to construct the target child nodes of the current parent node in multiple target decision trees may include the following steps:

[0188] In step S3031, determine the target attribute data from the sample object attribute data of the sample objects in the current parent node.

[0189] In a specific embodiment, the above-mentioned target attribute data may be attribute data corresponding to a target object attribute; specifically, the target object attribute may be an object attribute that has not been used for node splitting processing among at least one object attribute. Optionally, the object attributes that have not been used for node splitting processing may be determined in turn according to the preset object attribute splitting order corresponding to each target decision tree.

[0190] In step S3033, obtain the operation coding information corresponding to multiple business operations.

[0191] In a specific embodiment, the operation coding information corresponding to multiple business operations may be generated in combination with the one-hot encoding algorithm.

[0192] In step S3035, based on the target attribute data, the operation coding information, and the sample business index data of the sample objects in the current parent node, split the sample objects in the current parent node to construct the target child nodes of the current parent node in the target decision tree.

[0193] In an optional embodiment, as Figure 5 shown, the above-mentioned splitting process of the sample objects in the current parent node based on the target attribute data, operation coding information, and sample business metric data of the sample objects in the current parent node to construct the target child nodes of the current parent node in the target decision tree may include the following steps:

[0194] In step S501, according to the target attribute data, split the sample objects in the current parent node to obtain multiple pairs of initial child nodes of the current parent node;

[0195] In a specific embodiment, assume that the target object attribute is age. Correspondingly, the target attribute data may be the age data of the corresponding sample objects. Optionally, multiple age thresholds may be preset in advance, and a pair of initial child nodes may be divided in combination with each age threshold (for example, the sample objects with age information greater than the age threshold are used as the left child nodes, and the sample objects with age information less than or equal to the age threshold are used as the right child nodes).

[0196] In step S503, according to the operation coding information and the sample business metric data of the sample objects in each pair of initial child nodes, determine the predicted business metric data of the sample objects in each pair of initial child nodes under multiple business operations.

[0197] In a specific embodiment, the above-mentioned determination of the predicted business metric data of the sample objects in each pair of initial child nodes under multiple business operations based on the operation coding information and the sample business metric data of the sample objects in each pair of initial child nodes may be combined with the following formula:

[0198]

[0199] where ρ represents the predicted business metric data of the sample objects in a certain pair of initial child nodes under multiple business operations, Y i represents the sample business metric data of the i-th sample object in a certain pair of initial child nodes; m is the number of sample objects in a certain pair of initial child nodes; ∑ is the summation symbol; W i represents the operation coding information of the actual business operation assigned to the i-th sample object in a certain pair of initial child nodes (wherein, the operation coding information corresponding to the object without business operation may be 0).

[0200] In step S505, based on the preset business metric data, determine multiple pairs of primary selected child nodes from multiple pairs of initial child nodes.

[0201] In a specific embodiment, determining multiple primary candidate sub-node pairs from multiple initial sub-node pairs based on the preset service metric data may include: performing node heterogeneity analysis on the multiple initial sub-node pairs according to the preset service metric data to obtain node heterogeneity analysis results corresponding to the multiple initial sub-node pairs, where the node heterogeneity analysis results characterize the differences between each pair of initial sub-nodes; and determining multiple primary candidate sub-node pairs from the multiple initial sub-node pairs according to the node heterogeneity analysis results.

[0202] In a specific embodiment, performing node heterogeneity analysis on the multiple initial sub-node pairs according to the preset service metric data to obtain the node heterogeneity analysis results corresponding to the multiple initial sub-node pairs may be combined with the following formula:

[0203]

[0204] Wherein, represents the node heterogeneity analysis result corresponding to a certain initial sub-node pair, C1 and C2 respectively represent two sub-nodes in a certain initial sub-node pair, |{i: X i ∈C l}| represents the number of sample objects in the l-th initial sub-node C l , X i represents the i-th sample object in C l ; ρ ij represents the preset service metric data of the i-th sample object in C l under the j-th service operation; and k represents the number of operations of multiple service operations.

[0205] In an alternative embodiment, determining multiple primary candidate sub-node pairs from the multiple initial sub-node pairs according to the node heterogeneity analysis results may include: sorting the multiple initial sub-node pairs in descending order according to the node heterogeneity analysis results, and taking the preset number of pairs of initial sub-nodes before sorting as the multiple primary candidate sub-node pairs. Specifically, the preset number can be set according to the actual application.

[0206] In another alternative embodiment, determining multiple primary candidate sub-node pairs from the multiple initial sub-node pairs according to the node heterogeneity analysis results may include: selecting the initial sub-node pairs whose node heterogeneity analysis results are greater than or equal to a preset threshold as the multiple primary candidate sub-node pairs. Specifically, the preset threshold can be set according to the actual application.

[0207] In the above embodiments, by combining the predicted service metric data of the sample objects in each pair of initial sub-nodes under multiple service operations to analyze the differences between each pair of initial sub-nodes, the accuracy of the node heterogeneity analysis and the unity of the sample objects in the same node can be effectively improved, thereby improving the effectiveness of the multiple primary candidate sub-node pairs selected.

[0208] In step S507, according to the operation coding information and the sample service index data of the sample objects in each pair of primary candidate child nodes, determine the average service index gain corresponding to each primary candidate child node and the operation service index gain corresponding to the sample objects in each primary candidate child node.

[0209] In a specific embodiment, the above average service index gain can characterize the sensitivity degree of the sample objects in each primary candidate child node to various service operations; the above operation service index gain can characterize the sensitivity degree of the sample objects in each primary candidate child node to each service operation.

[0210] In a specific embodiment, according to the operation coding information and the sample service index data of the sample objects in each pair of primary candidate child nodes, determining the average service index gain corresponding to each primary candidate child node and the operation service index gain corresponding to the sample objects in each primary candidate child node can be combined with the following formula:

[0211]

[0212]

[0213] Wherein, represents the operation service index gain corresponding to the sample objects in the primary candidate child node C l ; represents the average service index gain corresponding to the primary candidate child node C l ; Y i represents the sample service index data of the i-th sample object in the primary candidate child node C l ; m is the number of sample objects in the primary candidate child node C l ; W i represents the operation coding information of the service operation actually assigned to the i-th sample object in the primary candidate child node C l (wherein, the operation coding information corresponding to the object without service operation can be 0).

[0214] In step S509, based on the average service index gain and the operation service index gain, determine the target child node of the current parent node from multiple pairs of primary candidate child nodes.

[0215] In an alternative embodiment, determining the target child node of the current parent node from multiple primary candidate child node pairs based on the average service metric gain and the operational service metric gain may include: performing service gain difference analysis on the sample objects in the multiple primary candidate child node pairs according to the average service metric gain and the operational service metric gain to obtain service gain difference information corresponding to the sample objects in the multiple primary candidate child node pairs; and determining the target child node of the current parent node from the multiple primary candidate child node pairs according to the service gain difference information.

[0216] In a specific embodiment, the above service gain difference information may characterize the sensitivity difference degree of the sample objects in the multiple primary candidate child node pairs to various service operations.

[0217] In a specific embodiment, performing service gain difference analysis on the sample objects in the multiple primary candidate child node pairs according to the average service metric gain and the operational service metric gain to obtain service gain difference information corresponding to the sample objects in the multiple primary candidate child node pairs may be combined with the following formula:

[0218]

[0219] where represents the service gain difference information corresponding to the sample object in a certain primary candidate child node pair; represents the operational service metric gain of the j-th service operation corresponding to the sample object in the l-th initial child node C l ; represents the average service metric gain corresponding to the primary candidate child node C l ; and k represents the number of operations of various service operations.

[0220] In a specific embodiment, the primary candidate child node pair corresponding to the maximum service gain difference information may be used as the target child node.

[0221] In the above embodiment, by combining the average service metric gain corresponding to each primary candidate child node and the operational service metric gain corresponding to the sample object in each primary candidate child node to perform service gain difference analysis on the sample objects in the multiple primary candidate child node pairs, the distinctiveness of the analysis of the service metric gains corresponding to various service operations can be effectively improved, and further the effectiveness of the determined target child node can be improved.

[0222] In the above embodiments, during the node splitting process, by combining the operation coding information corresponding to multiple service operations, the predicted service index data of each sample object under multiple service operations can be predicted. First, the initial child nodes are screened based on the predicted service index data, and then the average service index gain and operation service index gain corresponding to the multiple initially selected child nodes after screening are combined to determine the target child nodes of the current parent node. This can greatly improve the processing efficiency of node splitting while enhancing the effectiveness of the split nodes.

[0223] In step S305, update the target child node to the current parent node, and repeat the step of performing node splitting on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node to construct the target child nodes of the current parent node in multiple target decision trees until the preset splitting convergence condition is reached.

[0224] In a specific embodiment, after determining the target child nodes of the current parent node, update the target child node to the current parent node, and repeat the step of performing node splitting on the current parent node until at least one type of object attribute has been used for node splitting (preset splitting convergence condition).

[0225] In step S307, the multiple target decision trees obtained when the preset splitting convergence condition is reached are used as the combined gain analysis model.

[0226] In a specific embodiment, assume that the target object attribute data includes age information and income information; as Figure 6 shown, Figure 6 is a schematic diagram of a target decision tree provided according to an exemplary embodiment. Specifically, the sample objects corresponding to multiple service operations and the sample objects without service operations can be split into different nodes by combining age information and income information.

[0227] In the above embodiments, during the node splitting process, by continuously updating the current parent node, nodes including the objects corresponding to multiple service operations and the objects without service operations can be quickly split, which can ensure that the combined gain analysis model is based on the same sample objects during the generation process, and perform the service index gain analysis corresponding to different service operations, thereby effectively improving the gain analysis accuracy of the subsequent combined gain analysis model.

[0228] In an alternative embodiment, the above method may further include: a step of model measurement and analysis. Specifically, as Figure 7 shown, the step of model measurement and analysis may include:

[0229] In step S701, based on the sample service metric data corresponding to the leaf nodes of multiple target decision trees in the joint gain analysis model and the operation coding information of multiple service operations, determine the sample service metric gain of the sample object.

[0230] In a specific embodiment, after obtaining the joint gain analysis model, the sample service metric gain of the sample object can be determined by combining the sample service metric data corresponding to the leaf nodes of multiple target decision trees. Specifically, the sample service metric data corresponding to the leaf nodes can be the sample service metric data of the sample objects in the leaf nodes. Optionally, the least squares regression algorithm can be combined to fit the sample service metric data of the sample objects in the leaf nodes and the operation coding information of multiple service operations to determine the sample service metric gain of the sample object. Specifically, the least squares regression algorithm can be combined to fit the sample service metric data of the experimental objects (objects corresponding to multiple service operations) in the sample object and the operation coding information of multiple service operations, and subtract the sample service metric data of the control object (object without service operation) in the sample object from the fitted data to obtain the sample service metric gain.

[0231] In step S703, based on the sample service metric gain, determine the sample service operation corresponding to the sample object from multiple service operations;

[0232] In an optional embodiment, the greedy algorithm can be combined in the process of determining the sample service operation corresponding to the sample object based on the sample service metric gain.

[0233] In another specific embodiment, as Figure 8 shown, the determination of the sample service operation corresponding to the sample object from multiple service operations based on the sample service metric gain may include:

[0234] In step S801, obtain the preset resource consumption information and preset resource upper limit corresponding to multiple service operations.

[0235] In step S803, based on the preset resource upper limit, preset resource consumption information, and sample service metric gain, determine the sample service decision parameter.

[0236] In step S805, determine the sample service operation according to the sample service decision parameter.

[0237] In a specific embodiment, the preset resource consumption information corresponding to each service operation may represent the virtual resources required to execute each service operation. The preset resource upper limit may be the upper limit of virtual resources that can be consumed by the business platform for executing the target service as set in advance. The above sample service decision parameters may represent the service metric gain brought about by allocating various service operations to the sample object under the constraint of the preset resource upper limit;

[0238] In practical applications, under the virtual resource budget constraint (i.e., under the constraint of the preset resource upper limit), in order to maximize the service metric gain, the following mathematical representation can be obtained:

[0239]

[0240]

[0241]

[0242] w ij ∈ {0, 1}, i = 1,..., m; j = 1,..., k.

[0243] Among them, v ij represents the sample service metric gain of the i-th sample object under the j-th service operation, and c ij represents the resource consumption information for allocating the j-th service operation to the i-th sample object (i.e., the preset resource consumption information corresponding to the j-th service operation); w ij represents allocating the j-th service operation to the i-th sample object; B is the preset resource upper limit, m is the number of sample objects; k is the number of service operation types.

[0244] In a specific embodiment, assume there are m different sample objects and k different service operations. The first constraint limits that each object can be assigned to at most one service operation. The second constraint limits the total virtual resources that the target service can consume. The third constraint w ij ∈ {0, 1} means whether the object is assigned to a service operation (0: not assigned to a service operation, 1: assigned to a service operation).

[0245] In a specific embodiment, in order to maximize the service metric gain under the above three constraints, the Lagrangian dual algorithm can be combined to convert the above problem of solving for the maximum value into a problem of solving for the optimal solution of λ (sample service decision parameter) in the dual objective function. Specifically, the dual objective function can be:

[0246]

[0247] Among them, L(λ) is a convex function with respect to λ.

[0248] In a specific embodiment, the dual gradient descent method can be combined to solve the sample service decision parameters. However, this algorithm requires selecting an appropriate learning rate and usually has low computational efficiency. To improve the computational performance, the dual gradient bisection algorithm can be combined to solve the sample service decision parameters.

[0249] In a specific embodiment, taking the combination of the dual gradient bisection algorithm to solve the sample service decision parameters as an example, as Figure 9 shown, determining the sample service decision parameters based on the preset resource upper limit, preset resource consumption information, and sample service metric gain may include the following steps:

[0250] In step S901, based on the preset resource upper limit, preset resource consumption information, and sample service metric gain, a dual objective function with the sample service decision parameters as the dependent variable is created.

[0251] In a specific embodiment, the creation of the dual objective function with the sample service decision parameters as the dependent variable can refer to the above relevant description and will not be elaborated here.

[0252] In step S903, according to the preset resource consumption information and sample service metric gain, the first sample service decision parameter and the second sample service decision parameter are determined;

[0253] In an alternative embodiment, it can be set that where is the label set of sample objects, is the label set of multiple service operations. Correspondingly, V represents the sample service metric gain matrix of sample objects, and C represents the preset resource consumption information matrix corresponding to multiple service operations.

[0254] In a specific embodiment, determining the first sample service decision parameter and the second sample service decision parameter according to the preset resource consumption information and sample service metric gain may include:

[0255] Let λ l = 0, λ r = max i，j (V / C), where λ l represents the first sample service decision parameter, and λ r represents the second sample service decision parameter.

[0256] In step S905, the initial service decision parameter is determined according to the first sample service decision parameter and the second sample service decision parameter;

[0257] In a specific embodiment, determining the initial sample service decision parameter based on the first sample service decision parameter and the second sample service decision parameter may be combined with the following formula:

[0258]

[0259] where λ * represents the initial sample service decision parameter; λ l represents the first sample service decision parameter, and λ r represents the second sample service decision parameter.

[0260] In step S907, determine the sample derivative of the dual objective function when the sample service decision parameter is equal to the initial sample service decision parameter;

[0261] In step S909, based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter, determine the sample service decision parameter in the dual objective function.

[0262] In a specific embodiment, determining the sample service decision parameter in the dual objective function based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter may include:

[0263] Determine the sample parameter difference between the first sample service decision parameter and the second sample service decision parameter;

[0264] When the sample parameter difference is greater than the preset parameter threshold and the sample derivative is greater than zero, use the second sample service decision parameter as the sample service decision parameter;

[0265] When the sample parameter difference is greater than the preset parameter threshold and the sample derivative is less than or equal to zero, use the first sample service decision parameter as the sample service decision parameter.

[0266] In a specific embodiment, when the sample service decision parameter corresponding to the sample object is equal to 0, it can be determined that no service processing is performed on the sample object; conversely, when the sample service decision parameter corresponding to the sample object is not equal to 0, the service operation corresponding to the maximum sample service decision parameter can be used as the sample service operation of the sample object. Specifically, the preset parameter threshold can be set in combination with the actual application and can be a relatively small positive number to ensure that the service index gain is not 0.

[0267] In the above embodiments, a dual objective function with the sample service decision parameter as the dependent variable is created by combining the preset resource upper limit, the preset resource consumption information, and the sample service index gain; and the sample service decision parameter in the dual objective function is determined by combining the sample derivative of the dual objective function when the sample service decision parameter is equal to the initial sample service decision parameter, as well as the first sample service decision parameter and the second sample service decision parameter, which can effectively improve the efficiency of calculating the sample service decision parameter and improve the calculation performance.

[0268] In a specific embodiment, when the sample service decision parameter is non-zero, the service operation corresponding to the maximum sample service decision parameter can be selected as the final sample service operation.

[0269] In step S705, the sample objects corresponding to the same sample service operation are divided into the same sample object group;

[0270] In step S707, the control object group corresponding to the same service operation in the preset control group is determined;

[0271] In a specific embodiment, the preset control group can be a combination of objects randomly assigned to record operations in the actual online environment.

[0272] In step S709, the intersection object group between the sample object group corresponding to each service operation and the sample object groups is determined;

[0273] In step S711, based on the sample service index data corresponding to the intersection object group, the sample service index data corresponding to the sample object group, the first number of objects in the intersection object group, and the second number of objects in the sample object group, the service measurement index data is determined;

[0274] In a specific embodiment, the above service measurement index data can be used to measure the accuracy of the gain analysis of the joint gain analysis model and the rationality of the sample service operation determined based on the sample service index gain. Specifically, the service measurement index data can represent the percentage of the service index gain brought by the joint gain analysis model. Optionally, when determining the sample service operation corresponding to the sample object in combination with the preset resource upper limit, the above service measurement index data can represent the percentage of the service index gain brought by the joint gain analysis model under the limitation of the preset resource upper limit.

[0275] In a specific embodiment, determining the service measurement index data based on the sample service index data corresponding to the intersection object group, the sample service index data corresponding to the sample object group, the first number of objects in the intersection object group, and the second number of objects in the sample object group can be combined with the following formula:

[0276]

[0277] Among them, represents business measurement index data, A(j) represents the set of sample objects that will be assigned the j-th business operation determined based on the sample business index gain determined by the combined gain analysis model; S(j) represents the intersection object group between the set of sample objects (sample object group) assigned the j-th business operation and the set of control objects (control object group) assigned the j-th business operation in the preset control group determined based on the sample business index gain determined by the combined gain analysis model; |{i|t i = 0}| represents the number of objects among the sample objects that are not assigned business operations; represents the sum of the sample business index data of the objects among the sample objects that are not assigned business operations; represents the sum of the sample business index data of the objects in the intersection object group corresponding to the j-th business operation; m is the number of objects of the sample objects; k is the number of operations of multiple business operations.

[0278] In the above embodiments, based on the sample business index gain determined by the combined gain analysis model, the sample business operation corresponding to the sample object is determined, and through the intersection object group between the sample object group assigned the same sample business operation based on the combined gain analysis model and the control object group of the same business operation in the preset control group, the business measurement index data can be determined, which can solve the problem of the lack of real business index data corresponding to multiple business operations in practical applications based on overlapping sample objects, and further can greatly improve the effectiveness of the determined business measurement index data when measuring the gain analysis accuracy of the combined gain analysis model and the rationality of the sample business operation determined based on the sample business index gain.

[0279] As can be seen from the technical solutions provided in the embodiments of this specification above, in the process of generating the combined gain analysis model in this specification, the objects corresponding to multiple business operations and the objects without business operations are used as sample objects, and the target sample objects including the objects corresponding to multiple business operations and the objects without business operations are configured for the root nodes of multiple target decision trees. During the process of generating the combined gain analysis model, based on the same sample objects, the business index gain analysis corresponding to different business operations can be performed, which can improve the gain analysis accuracy of the subsequent combined gain analysis model, reduce the computational complexity of the gain analysis, and improve the business processing efficiency.

[0280] Figure 10 is a flowchart of a business processing method shown according to an exemplary embodiment. As Figure 10 shown, this business processing method is used in a terminal electronic device and includes the following steps.

[0281] In step S1001, the target object attribute data of the target object is obtained.

[0282] In a specific embodiment, the target object may be a user account that needs to perform business operations in a business platform.

[0283] In step S1003, based on the joint gain analysis model corresponding to multiple business operations and the target object attribute data, a gain analysis is performed on the target object to obtain the target business metric gain of the target object under multiple business operations.

[0284] In a specific embodiment, the above-mentioned target business metric gain may characterize the sensitivity of the target object to multiple business operations.

[0285] In a specific embodiment, the above-mentioned joint gain analysis model may include multiple target decision trees; optionally, as Figure 11 shown, the above-mentioned performing a gain analysis on the target object based on the joint gain analysis model corresponding to multiple business operations and the target object attribute data to obtain the target business metric gain of the target object may include the following steps:

[0286] In step S1101, according to the target object attribute data, the associated leaf nodes of the target object in multiple target decision trees are determined from the multiple target decision trees.

[0287] In a specific embodiment, each associated leaf node may include associated sample objects corresponding to multiple business operations. Specifically, the leaf node where the target object is located can be determined in combination with the target object attribute data. Correspondingly, the leaf node where the target object is located in each target decision tree can be the associated leaf node corresponding to the target object.

[0288] In step S1103, the sample business metric data of multiple associated sample objects in the associated leaf node and the operation coding information of multiple business operations are obtained;

[0289] In step S1105, based on the number of objects of each associated sample object in the associated leaf node, the weight information corresponding to each associated sample object is determined;

[0290] In a specific embodiment, multiple target decisions may include multiple associated leaf nodes corresponding to the target object; there are the same sample objects in different associated leaf nodes. Correspondingly, the number of objects corresponding to each sample object (associated sample object) in multiple associated leaf nodes can be counted, and the corresponding weight information can be determined in combination with the number of objects.

[0291] Specifically, the number of objects of each associated sample object is positively correlated with the corresponding weight information. The number of objects can be quantified into the corresponding weight information in combination with a certain rule. Optionally, the number of objects can also be directly used as the weight information.

[0292] In step S1107, based on the weight information, the operation coding information, and the sample business metric data of multiple associated sample objects, determine the target business metric gain.

[0293] In a specific embodiment, the sample business metric data of the experimental objects (objects corresponding to multiple business operations) among the multiple associated sample objects and the operation coding information of the multiple business operations can be fitted by combining the weighted least squares regression algorithm, and the data after fitting is subtracted from the sample business metric data of the control objects (objects without business operations) among the multiple associated sample objects to obtain the target business metric gain.

[0294] In the above embodiment, by combining the target object attribute data, find the associated leaf node where the target object is located from multiple decision trees, and combining the sample business metric data of the associated sample objects corresponding to multiple business operations in the associated leaf node, the operation coding information of the multiple business operations, and the corresponding weight information of each associated sample object in the associated leaf node, the target business metric gain can be determined quickly and accurately.

[0295] In step S1005, based on the target business metric gain, determine the target business operation corresponding to the target object from multiple business operations;

[0296] In an alternative embodiment, as Figure 12 shown, the above-mentioned determining the target business operation corresponding to the target object from multiple business operations based on the target business metric gain may include the following steps:

[0297] In step S1201, obtain the preset resource consumption information and the preset resource upper limit corresponding to multiple business operations;

[0298] In step S1203, based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain, determine the target business decision parameter.

[0299] In a specific embodiment, the above-mentioned determining the target business decision parameter based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain may include: creating a dual objective function with the target business decision parameter as the dependent variable based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain; determining the first business decision parameter and the second business decision parameter according to the preset resource consumption information and the target business metric gain; determining the initial business decision parameter according to the first business decision parameter and the second business decision parameter; determining the target derivative of the dual objective function when the target business decision parameter is equal to the initial business decision parameter; and determining the target business decision parameter in the dual objective function based on the target derivative, the first business decision parameter, and the second business decision parameter.

[0300] In a specific embodiment, for the specific refinement of determining the target service decision parameter based on the preset resource limit, preset resource consumption information, and target service metric gain, reference may be made to the determination of the sample service decision parameter based on the preset resource limit, preset resource consumption information, and sample service metric gain, which will not be elaborated herein.

[0301] In a specific embodiment, the target service decision parameter may represent the service metric gain brought about by the resource consumption information generated by allocating multiple service operations to the target object under the constraint of the preset resource limit.

[0302] In step S1205, the target service operation is determined according to the target service decision parameter.

[0303] In a specific embodiment, when the target service decision parameter is non-zero, the service operation corresponding to the maximum target service decision parameter may be selected as the target service operation. Optionally, when the target service decision parameter is 0, no service processing is performed on the target object.

[0304] In the above embodiment, by combining the preset resource limit, preset resource consumption information, and target service metric gain, a dual objective function with the target service decision parameter as the dependent variable is created; and by combining the objective derivative of the dual objective function when the target service decision parameter is equal to the initial service decision parameter, as well as the first service decision parameter and the second service decision parameter, to determine the target service decision parameter in the dual objective function, the efficiency of calculating the target service decision parameter can be effectively improved, and the calculation performance can be enhanced.

[0305] In step S1007, service processing is performed on the target object based on the target service operation.

[0306] In a specific embodiment, after determining the target service operation, the target service operation may be executed on the target object to achieve service processing.

[0307] In a specific embodiment, as Figure 13 shown Figure 13It is a schematic diagram of business processing by a business platform provided according to an embodiment. Specifically, in the first stage, the determination of the model and the business operation allocation scheme can be performed offline first. The model construction and optimization module can construct a joint gain analysis model corresponding to multiple business operations by combining the business index data and object attribute information of users (sample objects) obtained under online random control allocation stored in the data storage module. Then, the business operation allocation module performs business operation allocation under the preset resource upper limit control by combining the sample business index gains obtained from the constructed joint gain analysis model, and performs model evaluation and optimization by combining the allocation results. Then, in the second stage, online business processing is performed. The target business index gain of the target object (user) can be determined by combining the joint gain model constructed offline. Then, the target business operation of the target object is determined by combining the allocation control scheme of the business operation, so as to implement the business processing of the target object.

[0308] As can be seen from the technical solutions provided in the embodiments of this specification above, in the process of business processing in this specification, by combining the joint gain analysis model corresponding to multiple business operations to perform gain analysis on the target object, the computational complexity in the business processing process can be effectively reduced, and the target business index gain of the target object under multiple business operations can be obtained quickly and accurately. Moreover, the target business index gain can accurately characterize the sensitivity of the target object to multiple business operations. Furthermore, the effectiveness of the target business operation determined based on the target business index gain can be greatly improved, the accuracy of business decision-making and the efficiency of business processing can be improved, and thus business processing can be better carried out according to user requirements.

[0309] Figure 14 It is a block diagram of a business processing device shown according to an exemplary embodiment. Referring to Figure 14 , the device includes:

[0310] The target object attribute data acquisition module 1410 is configured to acquire the target object attribute data of the target object;

[0311] The gain analysis module 1420 is configured to perform gain analysis on the target object based on the joint gain analysis model corresponding to multiple business operations and the target object attribute data, and obtain the target business index gain of the target object under multiple business operations. The target business index gain characterizes the sensitivity of the target object to multiple business operations; the multiple business operations are multiple operations for increasing the business index data corresponding to the target object;

[0312] The target business operation determination module 1430 is configured to determine the target business operation corresponding to the target object from multiple business operations based on the target business index gain;

[0313] The service processing module 1440 is configured to perform service processing on the target object based on the target service operation.

[0314] In an alternative embodiment, the target service operation determination module 1430 includes:

[0315] The first information acquisition unit is configured to acquire the preset resource consumption information and the preset resource upper limit corresponding to multiple service operations.

[0316] The target service decision parameter determination unit is configured to determine the target service decision parameter based on the preset resource upper limit, the preset resource consumption information, and the target service index gain. The target service decision parameter represents the service index gain brought by allocating the resource consumption information generated by multiple service operations to the target object under the constraint of the preset resource upper limit.

[0317] The target service operation determination unit is configured to determine the target service operation according to the target service decision parameter.

[0318] In an alternative embodiment, the target service decision parameter determination unit includes:

[0319] The first dual objective function creation unit is configured to create a dual objective function with the target service decision parameter as the dependent variable based on the preset resource upper limit, the preset resource consumption information, and the target service index gain.

[0320] The first service decision parameter determination unit is configured to determine the first service decision parameter and the second service decision parameter according to the preset resource consumption information and the target service index gain.

[0321] The initial service decision parameter determination unit is configured to determine the initial service decision parameter according to the first service decision parameter and the second service decision parameter.

[0322] The target derivative determination unit is configured to determine the target derivative of the dual objective function when the target service decision parameter is equal to the initial service decision parameter.

[0323] The target service decision parameter determination unit is configured to determine the target service decision parameter in the dual objective function based on the target derivative, the first service decision parameter, and the second service decision parameter.

[0324] In an alternative embodiment, the joint gain analysis model includes multiple target decision trees.

[0325] The gain analysis module 1420 includes:

[0326] The associated leaf node determination unit is configured to determine, according to the target object attribute data, the associated leaf nodes of the target object in multiple target decision trees from multiple target decision trees, and each associated leaf node includes associated sample objects corresponding to multiple service operations;

[0327] The sample service metric data acquisition unit is configured to acquire the sample service metric data of multiple associated sample objects in the associated leaf nodes and the operation coding information of multiple service operations;

[0328] The weight information determination unit is configured to determine the weight information corresponding to each associated sample object based on the number of objects of each associated sample object in the associated leaf node;

[0329] The target service metric gain determination unit is configured to determine the target service metric gain based on the weight information, the operation coding information, and the sample service metric data.

[0330] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0331] Figure 15 It is a block diagram of a combined gain analysis model generation device shown according to an exemplary embodiment. Refer to Figure 15 , the device includes:

[0332] The sample data acquisition module 1510 is configured to acquire the sample service metric data of the sample object and the sample object attribute data of the sample object without service operations, and the sample object includes the object corresponding to multiple service operations and the object; the multiple service operations are multiple operations for increasing the service metric data corresponding to the sample object

[0333] The root node construction module 1520 is configured to construct the root nodes of multiple target decision trees respectively based on the sample object, and each root node includes a target sample object, and each target sample object includes the object corresponding to multiple service operations and the object without service operations;

[0334] The combined gain analysis model generation module 1530 is configured to perform node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample service metric data of each target sample object, and generate a combined gain analysis model corresponding to multiple service operations;

[0335] The combined gain analysis model is used to analyze the target service metric gain of the target object under multiple service operations.

[0336] In an optional embodiment, the combined gain analysis model generation module 1530 includes:

[0337] A current parent node determination unit, configured to execute using each node as the current parent node corresponding to a plurality of target decision trees respectively;

[0338] A node splitting processing unit, configured to execute node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in the plurality of target decision trees, and the sample objects in the target child nodes include objects corresponding to various service operations and objects without service operations;

[0339] A current parent node update unit, configured to execute updating the target child node as the current parent node;

[0340] An iterative processing unit, configured to execute repeating the step of performing node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in the plurality of target decision trees, until a preset splitting convergence condition is reached;

[0341] A joint gain analysis model determination unit, configured to execute using the plurality of target decision trees obtained when the preset splitting convergence condition is reached as the joint gain analysis model.

[0342] In an optional embodiment, the sample object attribute data includes attribute data corresponding to at least one object attribute; the preset splitting convergence condition is that at least one object attribute has been used for node splitting processing;

[0343] The node splitting processing unit includes:

[0344] A target attribute data determination unit, configured to execute determining target attribute data from the sample object attribute data of the sample objects in the current parent node, where the target attribute data is the attribute data corresponding to the target object attribute, and the target object attribute is the object attribute among at least one object attribute that has not been used for node splitting processing;

[0345] An operation coding information acquisition unit, configured to execute acquiring operation coding information corresponding to various service operations;

[0346] A sample splitting processing unit, configured to execute splitting processing on the sample objects in the current parent node based on the target attribute data, the operation coding information, and the sample service index data of the sample objects in the current parent node, so as to construct target child nodes of the current parent node in the target decision tree.

[0347] In an optional embodiment, the sample splitting processing unit includes:

[0348] Multiple initial child node pair generation units, configured to perform splitting processing on sample objects in the current parent node according to target attribute data to obtain multiple initial child node pairs of the current parent node;

[0349] Predicted business metric data determination unit, configured to perform determining predicted business metric data of sample objects in each pair of initial child nodes under multiple business operations according to operation coding information and sample business metric data of sample objects in each pair of initial child nodes;

[0350] Initial selected child node pair determination unit, configured to perform determining multiple initial selected child node pairs from multiple initial child node pairs based on preset business metric data;

[0351] Business metric gain determination unit, configured to perform determining the average business metric gain corresponding to each initial selected child node and the operation business metric gain corresponding to sample objects in each initial selected child node according to operation coding information and sample business metric data of sample objects in each pair of initial selected child nodes; the average business metric gain represents the sensitivity degree of sample objects in each initial selected child node to multiple business operations; the operation business metric gain represents the sensitivity degree of sample objects in each initial selected child node to each business operation;

[0352] Target child node determination unit, configured to perform determining the target child node of the current parent node from multiple initial selected child node pairs based on the average business metric gain and the operation business metric gain.

[0353] In an optional embodiment, the initial selected child node pair determination unit includes:

[0354] Node heterogeneity analysis unit, configured to perform node heterogeneity analysis on multiple initial child node pairs according to preset business metric data to obtain node heterogeneity analysis results corresponding to multiple initial child node pairs, and the node heterogeneity analysis results represent the differences between each pair of initial child nodes;

[0355] Initial selected child node pair determination subunit, configured to perform determining multiple initial selected child node pairs from multiple initial child node pairs according to the node heterogeneity analysis results.

[0356] In an optional embodiment, the target child node determination unit includes:

[0357] Business gain difference analysis unit, configured to perform business gain difference analysis on sample objects in multiple initial selected child node pairs according to the average business metric gain and the operation business metric gain to obtain business gain difference information corresponding to sample objects in multiple initial selected child node pairs, and the business gain difference information represents the sensitive difference degree of sample objects in multiple initial selected child node pairs to multiple business operations;

[0358] A target child node determination subunit, configured to determine a target child node of a current parent node from multiple primary selected child node pairs according to service gain difference information.

[0359] In an optional embodiment, the above device further includes:

[0360] A sample service metric gain determination module, configured to determine a sample service metric gain of a sample object based on sample service metric data corresponding to leaf nodes of multiple target decision trees in a joint gain analysis model and operation coding information of multiple service operations;

[0361] A sample service operation determination module, configured to determine a sample service operation corresponding to a sample object from multiple service operations based on the sample service metric gain;

[0362] A first object grouping processing module, configured to divide sample objects corresponding to the same sample service operation into the same sample object group;

[0363] A second object grouping processing module, configured to determine a control object group corresponding to the same service operation in a preset control group;

[0364] An intersection object group determination module, configured to determine an intersection object group between sample object groups corresponding to each service operation;

[0365] A service measurement metric data determination module, configured to determine service measurement metric data based on sample service metric data corresponding to the intersection object group, sample service metric data corresponding to the sample object group, the number of first objects in the intersection object group, and the number of second objects in the sample object group;

[0366] Wherein, the service measurement metric data is used to measure the accuracy of the gain analysis of the joint gain analysis model and the rationality of the sample service operation determined based on the sample service metric gain.

[0367] In an optional embodiment, the sample service operation determination module includes:

[0368] A second information acquisition unit, configured to acquire preset resource consumption information and a preset resource upper limit corresponding to multiple service operations;

[0369] A sample service decision parameter determination unit, configured to determine a sample service decision parameter based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain, where the sample service decision parameter represents the service metric gain brought by the resource consumption information generated by allocating multiple service operations to the sample object under the constraint of the preset resource upper limit;

[0370] A sample service operation determination unit, configured to determine a sample service operation according to sample service decision parameters.

[0371] In an optional embodiment, the sample service decision parameter determination unit includes:

[0372] A second dual objective function creation unit, configured to create a dual objective function with sample service decision parameters as dependent variables based on a preset resource upper limit, preset resource consumption information, and sample service metric gain;

[0373] A second service decision parameter determination unit, configured to determine a first sample service decision parameter and a second sample service decision parameter according to preset resource consumption information and sample service metric gain;

[0374] An initial sample service decision parameter determination unit, configured to determine an initial sample service decision parameter according to the first sample service decision parameter and the second sample service decision parameter;

[0375] A sample derivative determination unit, configured to determine the sample derivative of the dual objective function when the sample service decision parameter is equal to the initial sample service decision parameter;

[0376] A sample service decision parameter determination unit, configured to determine the sample service decision parameter in the dual objective function based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter.

[0377] Regarding the device in the above embodiment, the specific manners in which each module performs operations have been described in detail in the embodiment related to the method, and will not be elaborated herein.

[0378] Figure 16 is a block diagram of an electronic device for service processing shown according to an exemplary embodiment. The electronic device may be a terminal, and its internal structure diagram may be as Figure 16As shown. The electronic device includes a processor, a memory, a network interface, a display screen, and an input device connected via 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 via a network connection. When the computer program is executed by the processor, it implements a business processing method. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the outer shell of the electronic device, or an external keyboard, touchpad, or mouse, etc.

[0379] Figure 17 is a block diagram of an electronic device for generating a joint gain analysis model shown according to an exemplary embodiment. The electronic device can be a server, and its internal structure diagram can be as Figure 17 shown. The electronic device includes a processor, a memory, and a network interface connected via 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 via a network connection. When the computer program is executed by the processor, it implements a joint gain analysis model generation method.

[0380] Those skilled in the art can understand that Figure 16 or Figure 17 the structures shown in do not constitute a limitation on the electronic devices to which the present disclosure solution is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0381] In an exemplary embodiment, an electronic device is further provided, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to execute the instructions to implement the business processing method or the joint gain analysis model generation method as in the embodiments of the present disclosure.

[0382] In an exemplary embodiment, a computer-readable storage medium is further provided. When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute the business processing method or the joint gain analysis model generation method in the embodiments of the present disclosure.

[0383] In an exemplary embodiment, a computer program product including instructions is further provided. When it runs on a computer, it causes the computer to execute the service processing method or the joint gain analysis model generation method in the embodiments of the present disclosure.

[0384] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0385] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present disclosure. 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 known common knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0386] It should be understood that the present disclosure is not limited to the exact structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. A service processing method, characterized in that, Including: Obtain the target object attribute data of the target object; Based on the joint gain analysis models corresponding to multiple business operations and the target object attribute data, perform gain analysis on the target object to obtain the target business metric gains of the target object under multiple business operations, where the target business metric gains represent the sensitivity of the target object to multiple business operations; The multiple business operations are multiple operations for increasing the business metric data corresponding to the target object; the joint gain analysis model is generated in the following manner: Obtain the sample business metric data of the sample object and the sample object attribute data of the sample object, where the sample object includes objects corresponding to multiple business operations and objects corresponding to no business operation; Based on the sample object, respectively construct the root nodes of multiple target decision trees, and each target sample object includes objects corresponding to multiple business operations and objects with no business operation; Based on the sample object attribute data of each target sample object and the sample business metric data of each target sample object, perform node splitting processing on each root node to generate the joint gain analysis model corresponding to multiple business operations; Based on the target business metric gains, determine the target business operation corresponding to the target object from multiple business operations; Based on the target business operation, perform business processing on the target object.

2. The service processing method according to claim 1, wherein, The determining the target business operation corresponding to the target object from multiple business operations based on the target business metric gains includes: Obtain the preset resource consumption information and preset resource upper limit corresponding to multiple business operations; Based on the preset resource upper limit, the preset resource consumption information, and the target business metric gains, determine the target business decision parameter, where the target business decision parameter represents the business metric gain brought by the resource consumption information generated by allocating multiple business operations to the target object under the constraint of the preset resource upper limit; Determine the target business operation according to the target business decision parameter.

3. The service processing method according to claim 2, wherein The determining the target business decision parameter based on the preset resource upper limit, the preset resource consumption information, and the target business metric gains includes: Based on the preset resource upper limit, the preset resource consumption information, and the target business metric gains, create a dual objective function with the target business decision parameter as the dependent variable; According to the preset resource consumption information and the target business metric gains, determine the first business decision parameter and the second business decision parameter; Determine the initial business decision parameter according to the first business decision parameter and the second business decision parameter; Determine the target derivative of the dual objective function when the target business decision parameter is equal to the initial business decision parameter; Based on the target derivative, the first business decision parameter, and the second business decision parameter, determine the target business decision parameter in the dual objective function.

4. The service processing method according to any one of claims 1 to 3, characterized in that The joint gain analysis model includes multiple target decision trees; Performing gain analysis on the target object based on the joint gain analysis model corresponding to multiple service operations and the target object attribute data to obtain the target service metric gain of the target object includes: According to the target object attribute data, determining the associated leaf nodes of the target object in multiple target decision trees from multiple target decision trees, and each associated leaf node includes associated sample objects corresponding to multiple service operations; Obtaining the sample service metric data of multiple associated sample objects in the associated leaf nodes and the operation coding information of multiple service operations; Based on the number of objects of each associated sample object in the associated leaf node, determining the weight information corresponding to each associated sample object; Based on the weight information, the operation coding information, and the sample service metric data, determining the target service metric gain.

5. A method for generating a combined gain analysis model, characterized in that, Including: Obtaining the sample service metric data of the sample object and the sample object attribute data of the sample object, where the sample object includes objects corresponding to multiple service operations and objects corresponding to no service operation; the multiple service operations are multiple operations for increasing the service metric data corresponding to the sample object; Based on the sample object, respectively constructing the root nodes of multiple target decision trees, and each root node includes a target sample object, and each target sample object includes objects corresponding to multiple service operations and objects corresponding to no service operation; Performing node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample service metric data of each target sample object to generate a joint gain analysis model corresponding to multiple service operations; The joint gain analysis model is used to analyze the target service metric gain of the target object under multiple service operations.

6. The method for generating a combined gain analysis model according to claim 5, wherein, The performing node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample service metric data of each target sample object to generate a joint gain analysis model corresponding to multiple service operations includes: Taking each root node as the current parent node corresponding to multiple target decision trees; Performing node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service metric data of the target sample object in the current parent node to construct the target child nodes of the current parent node in multiple target decision trees, and the sample objects in the target child nodes include objects corresponding to multiple service operations and objects corresponding to no service operation; Updating the target child nodes as the current parent node, and repeating the step of performing node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service metric data of the target sample object in the current parent node to construct the target child nodes of the current parent node in multiple target decision trees until a preset splitting convergence condition is reached; Taking the multiple target decision trees obtained when the preset splitting convergence condition is reached as the joint gain analysis model.

7. The method for generating the combined gain analysis model according to claim 6, wherein The sample object attribute data includes attribute data corresponding to at least one object attribute; the preset splitting and convergence condition is that all of the at least one object attribute has been used for node splitting processing; The node splitting processing of the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node to construct target child nodes of the current parent node in multiple target decision trees includes: Determine target attribute data from the sample object attribute data of the sample objects in the current parent node, where the target attribute data is attribute data corresponding to a target object attribute, and the target object attribute is an object attribute among the at least one object attribute that has not been used for node splitting processing; Obtain operation coding information corresponding to multiple types of the service operations; Based on the target attribute data, the operation coding information, and the sample service index data of the sample objects in the current parent node, perform splitting processing on the sample objects in the current parent node to construct target child nodes of the current parent node in the target decision tree.

8. The method for generating the combined gain analysis model according to claim 7, wherein The performing splitting processing on the sample objects in the current parent node based on the target attribute data, the operation coding information, and the sample service index data of the sample objects in the current parent node to construct target child nodes of the current parent node in the target decision tree includes: According to the target attribute data, perform splitting processing on the sample objects in the current parent node to obtain multiple initial child node pairs of the current parent node; According to the operation coding information and the sample service index data of the sample objects in each pair of the initial child nodes, determine the predicted service index data of the sample objects in each pair of the initial child nodes under multiple types of the service operations; Based on preset service index data, determine multiple primary selection child node pairs from multiple initial child node pairs; According to the operation coding information and the sample service index data of the sample objects in each pair of the primary selection child nodes, determine the average service index gain corresponding to each primary selection child node and the operation service index gain corresponding to the sample objects in each primary selection child node; the average service index gain represents the sensitivity of the sample objects in each primary selection child node to multiple types of the service operations; the operation service index gain represents the sensitivity of the sample objects in each primary selection child node to each type of the service operations; Based on the average service index gain and the operation service index gain, determine the target child nodes of the current parent node from multiple primary selection child node pairs.

9. The method for generating the combined gain analysis model according to claim 8, wherein, The determining multiple primary selection child node pairs from multiple initial child node pairs based on preset service index data includes: Perform node heterogeneity analysis on multiple initial child node pairs according to preset service index data to obtain node heterogeneity analysis results corresponding to multiple initial child node pairs, where the node heterogeneity analysis results represent the differences between each pair of the initial child nodes; According to the node heterogeneity analysis results, determine multiple primary selection child node pairs from multiple initial child node pairs.

10. The method for generating the combined gain analysis model according to claim 8, characterized in that, Determining the target child node of the current parent node from multiple pairs of primary candidate child nodes based on the average service metric gain and the operational service metric gain includes: Performing service gain difference analysis on sample objects in multiple pairs of primary candidate child nodes according to the average service metric gain and the operational service metric gain, to obtain service gain difference information corresponding to the sample objects in multiple pairs of primary candidate child nodes, where the service gain difference information characterizes the sensitive difference degrees of the sample objects in multiple pairs of primary candidate child nodes to multiple types of service operations; Determining the target child node of the current parent node from multiple pairs of primary candidate child nodes according to the service gain difference information.

11. The method for generating a combined gain analysis model according to any one of claims 5 to 10, characterized in that, The method further includes: Determining the sample service metric gain of the sample object based on the sample service metric data corresponding to the leaf nodes of multiple target decision trees in the joint gain analysis model and the operation coding information of multiple types of service operations; Determining the sample service operation corresponding to the sample object from multiple types of service operations based on the sample service metric gain; Dividing the sample objects corresponding to the same sample service operation into the same sample object group; Determining the control object group corresponding to the same service operation in the preset control group; Determining the intersection object group between the sample object groups corresponding to each service operation; Determining service measurement index data based on the sample service metric data corresponding to the intersection object group, the sample service metric data corresponding to the sample object group, the number of first objects in the intersection object group, and the number of second objects in the sample object group; where the service measurement index data is used to measure the accuracy of the gain analysis of the joint gain analysis model and the rationality of the sample service operation determined based on the sample service metric gain.

12. The method for generating the combined gain analysis model according to claim 11, wherein Determining the sample service operation corresponding to the sample object from multiple types of service operations based on the sample service metric gain includes: Obtaining the preset resource consumption information and the preset resource upper limit corresponding to multiple types of service operations; Determining sample service decision parameters based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain, where the sample service decision parameters characterize the service metric gain brought by allocating multiple types of service operations to the sample object under the constraint of the preset resource upper limit; Determining the sample service operation according to the sample service decision parameters.

13. The method for generating a combined gain analysis model according to claim 12, wherein Determining sample service decision parameters based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain includes: Creating a dual objective function with the sample service decision parameters as the dependent variable based on the preset resource upper limit, the preset resource consumption information, and the sample service metric gain; Determining a first sample service decision parameter and a second sample service decision parameter according to the preset resource consumption information and the sample service metric gain; Determining an initial sample service decision parameter according to the first sample service decision parameter and the second sample service decision parameter; Determine the sample derivative of the dual objective function when the sample business decision parameter is equal to the initial sample business decision parameter; Based on the sample derivative, the first sample business decision parameter, and the second sample business decision parameter, determine the sample business decision parameter in the dual objective function.

14. A service processing device, characterized in that, Including: A target object attribute data acquisition module configured to acquire target object attribute data of a target object; A gain analysis module configured to perform gain analysis on the target object based on a joint gain analysis model corresponding to multiple business operations and the target object attribute data, to obtain a target business metric gain of the target object under the multiple business operations, where the target business metric gain characterizes the sensitivity of the target object to the multiple business operations; The multiple business operations are multiple operations for increasing business metric data corresponding to the target object; the joint gain analysis model is generated in the following manner: acquire sample business metric data of a sample object and sample object attribute data of the sample object, where the sample object includes objects corresponding to multiple business operations and an object corresponding to no business operation; Based on the sample object, respectively construct root nodes of multiple target decision trees, where each target sample object includes objects corresponding to multiple business operations and an object corresponding to no business operation; Perform node splitting processing on each root node based on the sample object attribute data of each target sample object and the sample business metric data of each target sample object to generate a joint gain analysis model corresponding to the multiple business operations; A target business operation determination module configured to determine a target business operation corresponding to the target object from the multiple business operations based on the target business metric gain; A service processing module configured to perform service processing on the target object based on the target business operation.

15. The service processing device according to claim 14, characterized in that, The target business operation determination module includes: A first information acquisition unit configured to acquire preset resource consumption information and a preset resource upper limit corresponding to the multiple business operations; A target business decision parameter determination unit configured to determine a target business decision parameter based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain, where the target business decision parameter characterizes the business metric gain brought by allocating resource consumption information generated by the multiple business operations to the target object under the constraint of the preset resource upper limit; A target business operation determination unit configured to determine the target business operation according to the target business decision parameter.

16. The service processing device according to claim 15, wherein The target business decision parameter determination unit includes: A first dual objective function creation unit configured to create a dual objective function with the target business decision parameter as the dependent variable based on the preset resource upper limit, the preset resource consumption information, and the target business metric gain; A first business decision parameter determination unit configured to determine a first business decision parameter and a second business decision parameter according to the preset resource consumption information and the target business metric gain; An initial service decision parameter determination unit, configured to execute the determination of initial service decision parameters according to the first service decision parameter and the second service decision parameter; A target derivative determination unit, configured to execute the determination of the target derivative of the dual objective function in the case that the target service decision parameter is equal to the initial service decision parameter; A target service decision parameter determination unit, configured to execute the determination of the target service decision parameter in the dual objective function based on the target derivative, the first service decision parameter, and the second service decision parameter.

17. The service processing apparatus according to any one of claims 14 to 16, characterized in that, The joint gain analysis model includes a plurality of target decision trees; The gain analysis module includes: An associated leaf node determination unit, configured to execute the determination of the associated leaf nodes of the target object in a plurality of the target decision trees according to the target object attribute data, each of the associated leaf nodes including associated sample objects corresponding to a plurality of the service operations; A sample service index data acquisition unit, configured to execute the acquisition of the sample service index data of a plurality of the associated sample objects in the associated leaf nodes and the operation coding information of a plurality of the service operations; A weight information determination unit, configured to execute the determination of the weight information corresponding to each of the associated sample objects based on the number of objects of each associated sample object in the associated leaf nodes; A target service index gain determination unit, configured to execute the determination of the target service index gain based on the weight information, the operation coding information, and the sample service index data.

18. An apparatus for generating a combined gain analysis model, characterized in that It includes: A sample data acquisition module, configured to execute the acquisition of the sample service index data of a sample object and the sample object attribute data of the sample object, the sample object including objects corresponding to a plurality of service operations and an object corresponding to no service operation; the plurality of service operations being a plurality of operations for increasing the service index data corresponding to the sample object; A root node construction module, configured to execute the construction of the root nodes of a plurality of target decision trees respectively based on the sample object, each of the root nodes including a target sample object, each of the target sample objects including objects corresponding to a plurality of the service operations and an object corresponding to no service operation; A joint gain analysis model generation module, configured to execute the node splitting process on each of the root nodes based on the sample object attribute data of each of the target sample objects and the sample service index data of each of the target sample objects, and generate a joint gain analysis model corresponding to a plurality of the service operations; The joint gain analysis model is used to analyze the target service index gain of the target object under a plurality of the service operations.

19. The apparatus for generating a combined gain analysis model according to claim 18, wherein The joint gain analysis model generation module includes: A current parent node determination unit, configured to execute using each of the root nodes as the current parent node corresponding to a plurality of the target decision trees respectively; The node splitting processing unit is configured to perform node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in multiple target decision trees, and the sample objects in the target child nodes include objects corresponding to multiple types of the service operations and objects without service operations; The current parent node updating unit is configured to perform updating the target child nodes as the current parent node; The iterative processing unit is configured to perform repeating the step of performing node splitting processing on the current parent node based on the sample object attribute data of the target sample object in the current parent node and the sample service index data of the target sample object in the current parent node, so as to construct target child nodes of the current parent node in multiple target decision trees, until a preset splitting convergence condition is reached; The joint gain analysis model determining unit is configured to perform using multiple target decision trees obtained when the preset splitting convergence condition is reached as the joint gain analysis model.

20. The combined gain analysis model generation device according to claim 19, characterized in that The sample object attribute data includes attribute data corresponding to at least one object attribute; the preset splitting convergence condition is that the at least one object attribute has been used for node splitting processing; The node splitting processing unit includes: The target attribute data determining unit is configured to perform determining target attribute data from the sample object attribute data of the sample objects in the current parent node, where the target attribute data is attribute data corresponding to a target object attribute, and the target object attribute is an object attribute in the at least one object attribute that has not been used for node splitting processing; The operation coding information obtaining unit is configured to perform obtaining operation coding information corresponding to multiple types of the service operations; The sample splitting processing unit is configured to perform splitting processing on the sample objects in the current parent node based on the target attribute data, the operation coding information, and the sample service index data of the sample objects in the current parent node, so as to construct target child nodes of the current parent node in the target decision tree.

21. The apparatus for generating a combined gain analysis model according to claim 20, wherein The sample splitting processing unit includes: Multiple initial child node pair generating units are configured to perform splitting processing on the sample objects in the current parent node according to the target attribute data, so as to obtain multiple initial child node pairs of the current parent node; The predicted service index data determining unit is configured to perform determining the predicted service index data of the sample objects in each pair of the initial child nodes under multiple types of the service operations according to the operation coding information and the sample service index data of the sample objects in each pair of the initial child nodes; The primary selection child node pair determining unit is configured to perform determining multiple primary selection child node pairs from multiple initial child node pairs based on preset service index data; A service metric gain determination unit, configured to determine, according to the operation coding information and the sample service metric data of the sample objects in each pair of the primary selected child nodes, the average service metric gain corresponding to each of the primary selected child nodes and the operation service metric gain corresponding to the sample objects in each of the primary selected child nodes; the average service metric gain represents the sensitivity degree of the sample objects in each of the primary selected child nodes to various service operations; the operation service metric gain represents the sensitivity degree of the sample objects in each of the primary selected child nodes to each service operation; A target child node determination unit, configured to determine the target child node of the current parent node from multiple pairs of the primary selected child nodes based on the average service metric gain and the operation service metric gain.

22. The combined gain analysis model generation device according to claim 21, wherein The primary selected child node pair determination unit includes: A node heterogeneity analysis unit, configured to perform node heterogeneity analysis on multiple pairs of the initial child nodes according to preset service metric data, and obtain node heterogeneity analysis results corresponding to the multiple pairs of the initial child nodes, where the node heterogeneity analysis results represent the differences between each pair of the initial child nodes; A primary selected child node pair determination subunit, configured to determine multiple pairs of the primary selected child nodes from the multiple pairs of the initial child nodes according to the node heterogeneity analysis results.

23. The apparatus for generating a combined gain analysis model according to claim 21, wherein The target child node determination unit includes: A service gain difference analysis unit, configured to perform service gain difference analysis on the sample objects in multiple pairs of the primary selected child nodes according to the average service metric gain and the operation service metric gain, and obtain service gain difference information corresponding to the sample objects in the multiple pairs of the primary selected child nodes, where the service gain difference information represents the sensitive difference degree of the sample objects in the multiple pairs of the primary selected child nodes to various service operations; A target child node determination subunit, configured to determine the target child node of the current parent node from the multiple pairs of the primary selected child nodes according to the service gain difference information.

24. The combined gain analysis model generation device according to any one of claims 18 to 23, characterized in that The apparatus further includes: A sample service metric gain determination module, configured to determine the sample service metric gain of the sample object based on the sample service metric data corresponding to the leaf nodes of multiple target decision trees in the joint gain analysis model and the operation coding information of various service operations; A sample service operation determination module, configured to determine the sample service operation corresponding to the sample object from various service operations based on the sample service metric gain; A first object grouping processing module, configured to divide the sample objects corresponding to the same sample service operation into the same sample object group; A second object grouping processing module, configured to determine the control object group corresponding to the same service operation in the preset control group; An intersection object group determination module, configured to determine the intersection object group between the sample object groups corresponding to each service operation. A service measurement index data determination module, configured to determine service measurement index data based on the sample service index data corresponding to the intersection object group, the sample service index data corresponding to the sample object group, the number of first objects in the intersection object group, and the number of second objects in the sample object group; wherein the service measurement index data is used to measure the accuracy of the gain analysis of the combined gain analysis model and the rationality of the sample service operation determined based on the sample service index gain.

25. The apparatus for generating a combined gain analysis model according to claim 24, wherein The sample service operation determination module includes: A second information acquisition unit, configured to acquire the preset resource consumption information and the preset resource upper limit corresponding to multiple service operations; A sample service decision parameter determination unit, configured to determine sample service decision parameters based on the preset resource upper limit, the preset resource consumption information, and the sample service index gain, where the sample service decision parameters represent the service index gain brought by allocating multiple service operations to the sample objects under the constraint of the preset resource upper limit; A sample service operation determination unit, configured to determine the sample service operation according to the sample service decision parameters.

26. The joint gain analysis model generation device according to claim 25, wherein The sample service decision parameter determination unit includes: A second dual objective function creation unit, configured to create a dual objective function with the sample service decision parameters as the dependent variables based on the preset resource upper limit, the preset resource consumption information, and the sample service index gain; A second service decision parameter determination unit, configured to determine a first sample service decision parameter and a second sample service decision parameter according to the preset resource consumption information and the sample service index gain; An initial sample service decision parameter determination unit, configured to determine the initial sample service decision parameters according to the first sample service decision parameter and the second sample service decision parameter; A sample derivative determination unit, configured to determine the sample derivative of the dual objective function when the sample service decision parameters are equal to the initial sample service decision parameters; A sample service decision parameter determination unit, configured to determine the sample service decision parameters in the dual objective function based on the sample derivative, the first sample service decision parameter, and the second sample service decision parameter.

27. An electronic device, characterized in that, It includes: A processor; A memory for storing the executable instructions of the processor; wherein the processor is configured to execute the instructions to implement the service processing method according to any one of claims 1 to 4 or the combined gain analysis model method according to any one of claims 5 to 13.

28. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is enabled to execute the service processing method according to any one of claims 1 to 4 or the combined gain analysis model method according to any one of claims 5 to 13.

29. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, the service processing method according to any one of claims 1 to 4 or the combined gain analysis model method according to any one of claims 5 to 13 is implemented.

Images