Recommended object determination method and device, computer equipment and storage medium

Abstract

The invention relates to a recommended object determination method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: determining a target recommendation receiving object, a plurality of candidate recommended objects and a response time period according to a recommendation request; determining an acceptance probability value that the candidate recommended object is accepted by the target recommendation receiving object; determining a time preference value corresponding to at least one of each candidate recommended object or the target recommendation receiving object in the response time period; determining respective recommendation scores of the plurality of candidate recommended objects according to the corresponding time preference values and respective acceptance probability values of the plurality of candidate recommended objects; and according to the respective recommendation scores of the plurality of candidate recommended objects, screening out recommended objects recommended in the response time period from the plurality of candidate recommended objects. By adopting the method, the recommendation accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer device and storage medium for determining a recommended object. Background Art

[0002] With the development of science and technology, recommendation systems have emerged, through which recommended objects can be recommended to recommendation recipients. For example, items can be recommended to users through recommendation systems. Currently, when recommending recommended objects to a target object through a recommendation system, historical recommended objects that were recommended to the target object at a historical moment are referenced to predict the recommended objects that need to be recommended to the target object at the current time. However, predicting the current recommended object only by historical recommended objects will result in low recommendation accuracy. Summary of the invention

[0003] Based on this, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium and computer program product for determining a recommended object that can improve the accuracy of recommendation in response to the above technical problems.

[0004] In a first aspect, the present application provides a method for determining a recommended object, the method comprising:

[0005] When a recommendation request is received, determining a recommendation recipient, a plurality of candidate recommended objects, and a response time period according to the recommendation request;

[0006] For each of the plurality of candidate recommended objects, determining an acceptance probability value of the candidate recommended object being accepted by the recommended object according to the receiving object information of the recommended object and the recommended object information of the targeted candidate recommended object;

[0007] Determine a time preference value corresponding to the response time period for at least one of the candidate recommended objects or the recommended receiving objects; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or the corresponding recommended receiving object in the response time period;

[0008] Determine the recommendation score of each of the multiple candidate recommended objects according to the corresponding time preference value and the acceptance probability value of each of the multiple candidate recommended objects; the recommendation score is positively correlated with the acceptance probability value of the corresponding candidate recommended object and is positively correlated with the corresponding time preference value;

[0009] According to the recommendation scores of the multiple candidate recommended objects, the recommended objects for recommendation to the recommendation receiving object in the response time period are screened out from the multiple candidate recommended objects.

[0010] In a second aspect, the present application further provides a device for determining a recommended object, the device comprising:

[0011] An acceptance probability determination module is used to determine, when a recommendation request is received, a target recommendation receiving object, multiple candidate recommended objects and a response time period according to the recommendation request; for each of the multiple candidate recommended objects, based on the receiving object information of the target recommendation receiving object and the recommended object information of the targeted candidate recommended object, determine the acceptance probability value of the targeted candidate recommended object being accepted by the target recommendation receiving object.

[0012] A time preference value determination module is used to determine the time preference value corresponding to each candidate recommended object or at least one of the target recommended receiving objects in the response time period; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or at least one of the target recommended receiving objects in the response time period.

[0013] The recommendation module is used to determine the recommendation score of each of the multiple candidate recommended objects based on the corresponding time preference value and the acceptance probability value of each of the multiple candidate recommended objects; and to filter out the recommended objects to be recommended to the target recommendation receiving object within the response time period from the multiple candidate recommended objects based on the recommendation scores of each of the multiple candidate recommended objects.

[0014] In one of the embodiments, the acceptance probability determination module is also used to obtain an acceptance rate estimation decision tree when the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object is not stored; each node in the acceptance rate estimation decision tree has a corresponding first classification feature, and each leaf node has a corresponding probability value; according to the first classification feature corresponding to each node in the acceptance rate estimation decision tree, the receiving object information and the recommended object information of the targeted candidate recommended object, the first target leaf node to which the targeted candidate recommended object in the acceptance rate estimation decision tree is classified is determined; and the probability value corresponding to the first target leaf node is used as the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object.

[0015] In one of the embodiments, the acceptance probability determination module is also used to splice the receiving object information and the recommended object information of the candidate recommended object to obtain spliced ​​information; starting from the root node of the acceptance rate estimation decision tree, determine the current node, and based on the splicing information and the first classification feature of the child node connected to the current node, filter out the target child node to which the splicing information is classified from the child nodes connected to the current node; use the target child node as the new current node, return to the step of filtering out the target child node to which the splicing information is classified from the child nodes connected to the current node based on the splicing information and the first classification feature of the child node connected to the current node, and continue to execute until a leaf node in the acceptance rate estimation decision tree is reached, and the leaf node reached is used as the first target leaf node.

[0016] In one embodiment, the acceptance probability determination module is also used to obtain a first training set; the first training set includes multiple first training subsets; each of the first training subsets includes multiple first attributes; the multiple first attributes include attributes of the recommended receiving object dimension and attributes of the recommended object dimension; based on the multiple first attributes, the first training subsets in the first training set are classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the first classification feature corresponding to the corresponding node is selected from the multiple first attributes, until the first preset classification end condition is reached, and a first intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the first intermediate decision tree, the probability value corresponding to each leaf node in the first intermediate decision tree is determined, and the acceptance rate estimation decision tree is obtained by combining the first intermediate decision tree and the probability values ​​corresponding to each leaf node in the first intermediate decision tree.

[0017] In one embodiment, the acceptance probability determination module is also used to determine, for each leaf node in the first intermediate decision tree, the total number of first training subsets classified into the targeted leaf node, and determine the number of positive samples in the first training subset classified into the targeted leaf node; the positive samples are training subsets whose attribute values ​​of the target attributes are preset values; the target attribute is an attribute that characterizes whether the corresponding recommended object is accepted; and the ratio of the number of positive samples to the total number is used as the probability value corresponding to the targeted leaf node.

[0018] In one of the embodiments, the time preference value determination module is also used to determine the time preference value corresponding to the target recommendation receiving object in the response time period based on the receiving object information of the target recommendation receiving object; and to determine the time preference value corresponding to each of the multiple candidate recommended objects in the response time period based on the recommended object information of each of the multiple candidate recommended objects.

[0019] In one of the embodiments, the time preference value determination module is also used to obtain a receiving object time preference decision tree corresponding to the response time period when the time preference value corresponding to the target recommended receiving object is not stored; each non-leaf node in the receiving object time preference decision tree corresponding to the response time period has a corresponding second classification feature, and each leaf node has a corresponding probability value; according to the second classification feature corresponding to each non-leaf node in the receiving object time preference decision tree corresponding to the response time period and the receiving object information of the target recommended receiving object, determine the second target leaf node to which the target recommended receiving object is classified in the receiving object time preference decision tree corresponding to the response time period; according to the probability value corresponding to the second target leaf node, determine the time preference value corresponding to the target recommended receiving object in the response time period.

[0020] In one embodiment, the time preference value determination module is also used to obtain a second training set corresponding to the response time period; the second training set includes multiple second training subsets; each of the second training subsets is information related to the response time period collected for the corresponding recommendation object, and includes multiple second attributes; based on the multiple second attributes, the second training subsets in the second training set are classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the second classification feature corresponding to the corresponding node is selected from the multiple second attributes, until the second preset classification end condition is reached, and a second intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the second intermediate decision tree, the probability value corresponding to each leaf node in the second intermediate decision tree is determined, and the second intermediate decision tree and the probability value corresponding to each leaf node in the second intermediate decision tree are combined to obtain a receiving object time preference decision tree corresponding to the response time period.

[0021] In one of the embodiments, each of the multiple candidate recommended objects is used as a target candidate recommended object, and the time preference value determination module is also used to obtain a recommended object time preference decision tree corresponding to the response time period when the time preference value corresponding to the target candidate recommended object in the response time period is not stored; each non-leaf node in the recommended object time preference decision tree has a corresponding third classification feature, and each leaf node has a corresponding probability value; according to the third classification feature corresponding to each non-leaf node in the recommended object time preference decision tree and the recommended object information of the target candidate recommended object, determine the third target leaf node to which the target candidate recommended object should be classified in the recommended object time preference decision tree corresponding to the response time period; according to the probability value corresponding to the third target leaf node, determine the time preference value of the target candidate recommended object corresponding to the response time period.

[0022] In one embodiment, the time preference value determination module is also used to obtain a third training set corresponding to the response time period; the third training set includes multiple third training subsets; each of the third training subsets is information related to the response time period collected for the corresponding recommended object, and includes multiple third attributes; based on the multiple third attributes, the third training subsets in the third training set are classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the third classification feature corresponding to the corresponding node is selected from the multiple third attributes, until the third preset classification end condition is reached, and a third intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the third intermediate decision tree, the probability value corresponding to each leaf node in the third intermediate decision tree is determined, and the third intermediate decision tree and the probability value corresponding to each leaf node in the third intermediate decision tree are combined to obtain a recommended object time preference decision tree corresponding to the response time period.

[0023] In one of the embodiments, the recommendation module is further used to determine the time correction factor corresponding to each of the multiple candidate recommended objects based on the time preference value corresponding to the target recommendation receiving object, the acceptance probability values ​​of the multiple candidate recommended objects and the corresponding time preference values; obtain the acceptance value corresponding to each of the multiple candidate recommended objects, and determine the recommendation score of each of the multiple candidate recommended objects based on the acceptance value and the time correction factor corresponding to each of the multiple candidate recommended objects.

[0024] In one of the embodiments, the recommended object determination device also includes an optimization module, which is used to obtain multiple sample receiving object time preference decision trees generated based on sample object information of multiple sample recommendation receiving objects; each of the sample receiving object time preference decision trees corresponds to a preset time period; according to the multiple sample receiving object time preference decision trees, the time propensity scores corresponding to the multiple sample recommendation receiving objects in each of the preset time periods are determined; according to the time propensity scores corresponding to the multiple sample recommendation receiving objects in each of the preset time periods, the multiple sample recommendation receiving objects are grouped to obtain sample recommendation receiving object groups corresponding to each of the preset time periods; based on the sample recommendation receiving object groups corresponding to each of the preset time periods, the recommendation success rate improvement result of the preset sample recommended objects in each of the preset time periods is determined; the recommendation success rate improvement result is used to optimize the recommended object determination method.

[0025] In one of the embodiments, each of the multiple sample recommendation receiving objects is used as a target sample recommendation receiving object, and the optimization module is also used to determine, for each preset time period in a plurality of preset time periods, a fourth target leaf node to which the target sample recommendation receiving object is classified in the sample receiving object time preference decision tree corresponding to the preset time period; and determine the time propensity score of the target sample recommendation receiving object in the preset time period based on the probability value corresponding to the fourth target leaf node.

[0026] In one of the embodiments, the optimization module is also used to cluster the multiple sample recommendation receiving objects according to the time tendency score corresponding to the preset time period for each preset time period in the multiple preset time periods, so as to obtain a sample recommendation receiving object group corresponding to the preset time period.

[0027] In one of the embodiments, each of the preset time periods is used as a target preset time period respectively; the optimization module is further used to divide the sample recommendation receiving object group corresponding to the target preset time period into an experimental group and a control group; the sample recommendation receiving object in the experimental group is an object that receives the sample recommended object within the target preset time period; the sample recommendation receiving object in the control group is an object that does not receive the sample recommended object within the target preset time period; according to the acceptance result of each sample recommendation receiving object in the experimental group accepting the sample recommended object in the target preset time period, a first success probability value of the sample recommended object being successfully recommended within the target preset time period corresponding to the sample recommendation receiving object group is determined; according to the acceptance result of each sample recommendation receiving object in the control group accepting the sample recommended object in the non-target preset time period, a second success probability value of the sample recommended object being successfully recommended within the non-target preset time period corresponding to the sample recommendation receiving object group is determined; according to the first success probability value and the second success probability value corresponding to each sample recommendation receiving object group corresponding to the target preset time period, the recommendation success rate improvement result of the preset sample recommended object in the target preset time period is determined.

[0028] In one of the embodiments, the acceptance probability determination module is also used to receive a resource transfer request triggered by the target recommendation receiving object, and generate a recommendation request when performing a resource transfer operation based on the resource transfer request to obtain a resource transfer result; the recommendation module is also used to feed back a result page to the target recommendation receiving object; the result page carries the resource transfer result and the recommended object selected from the multiple candidate recommended objects.

[0029] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in any one of the methods for determining a recommended object provided in the embodiments of the present application are implemented.

[0030] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the recommended object determination methods provided in the embodiments of the present application are implemented.

[0031] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps in any one of the recommended object determination methods provided in the embodiments of the present application.

[0032] The above-mentioned method, device, computer equipment, storage medium and computer program product for determining recommended objects can determine the target recommended receiving object, multiple candidate recommended objects and response time period based on the recommendation request by receiving the recommendation request. By determining the target recommended receiving object, multiple candidate recommended objects and response time period, the acceptance probability value of each candidate recommended object accepted by the target recommended receiving object can be determined, as well as the time preference value of the target recommended receiving object in the response time period, and the time preference value of each candidate recommended object in the response time period can be determined. Since the time preference value reflects the recommendation success rate of the response time period, the recommendation score corresponding to each candidate recommended object can be determined based on the time preference value and the acceptance probability value, so as to screen out the recommended objects that are more likely to be accepted by the target recommended receiving object in the response time period based on the recommendation score. Since the present application takes the time factor into consideration when determining the recommended objects, the accuracy of the determined recommended objects can be improved based on the time factor. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A diagram of an application environment of a recommended object determination method in an embodiment;

[0034] Figure 2 A schematic diagram of a flow chart of a method for determining a recommended object in one embodiment;

[0035] Figure 3 A schematic diagram of determining a recommendation score in one embodiment;

[0036] Figure 4 A schematic diagram of an acceptance rate estimation decision tree in one embodiment;

[0037] Figure 5 is a schematic diagram of a first training set in one embodiment;

[0038] Figure 6 A schematic diagram of determining a time preference value for a recommended receiving object in one embodiment;

[0039] Figure 7 is a schematic diagram of a second training set in one embodiment;

[0040] Figure 8 is a schematic diagram of a third training set in one embodiment;

[0041] Fig. 9 A schematic diagram of a result of improving the recommendation success rate in one embodiment;

[0042] Fig.10 A schematic diagram of a sample recommendation receiving object group in one embodiment;

[0043] Fig.11A schematic diagram of a flow chart of a method for determining a recommended object in a specific embodiment;

[0044] Fig.12 A schematic diagram of the overall flow of a method for determining a recommended object in one embodiment;

[0045] Fig.13 A structural block diagram of a device for determining a recommended object in one embodiment;

[0046] Fig.14 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] The method for determining a recommended object provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other servers. Both the terminal 102 and the server 104 can be used alone to execute the recommended object determination method provided in the embodiment of the present application. The terminal 102 and the server 104 can also be used in conjunction to execute the recommended object determination method provided in the embodiment of the present application. Taking the terminal 102 and the server 104 as an example of being used in conjunction to execute the recommended object determination method provided in the embodiment of the present application, the recommended receiving object can trigger a recommendation request through the terminal 102, so that the terminal 102 can send the recommendation request to the server 104, so that the server 104 responds to the recommendation request, determines the recommended object, and returns the determined recommended object to the terminal 102. Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, etc. The portable wearable device may be a smart watch, a smart bracelet, a head-mounted device, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0049] It should be noted that the words "first", "second" and similar terms used in this application do not indicate any order, quantity or importance, but are only used to distinguish different components. Unless the context clearly indicates otherwise, the singular form "one", "an" or "the" and similar terms do not indicate a quantity restriction, but rather indicate the presence of at least one. The quantities of "multiple" or "multiple copies" mentioned in the various embodiments of the present application all refer to the quantity of "at least two", for example, "multiple" refers to "at least two", and "multiple copies" refers to "at least two copies".

[0050] It should be noted that the user information (including but not limited to receiving object information, recommended object information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0051] In one embodiment, Figure 2 As shown, a method for determining a recommended object is provided, and the method is applied to a computer device as an example for explanation. The computer device may be Figure 1 The method for determining the recommended object comprises the following steps:

[0052] Step 202: When a recommendation request is received, a target recommendation recipient, a plurality of candidate recommended objects and a response time period are determined according to the recommendation request.

[0053] The recommended recipient refers to the object that receives the recommended object; the recommended recipient may specifically be a natural person, a virtual character, or a robot, etc. The recommended object refers to the recommended object; the recommended object may specifically be a commodity, news information, or a service, etc. The response time period refers to the time period for sending the recommended object to the recommended recipient.

[0054] Specifically, when the computer device receives a recommendation request, since the recommendation request may carry an object identifier of the recommended receiving object, the computer device may extract the object identifier of the recommended receiving object from the recommendation request, and determine the corresponding recommended receiving object based on the extracted object identifier, and the determined recommended receiving object is the target recommended receiving object. The recommendation request carries a time tag, and the computer device may extract the time tag from the recommendation request, and determine the time period in which the time tag is located, and use the time period in which the time tag is located as the response time period. Among them, the time tag carried by the recommendation request may be the generation time of the recommendation request, or the reception time of the recommended receiving object. For example, when the time tag is 13:22, it can be determined that the time tag is in the time period from 13:00 to 14:00, so that the time period from 13:00 to 14:00 is used as the response time period. Further, the computer device may also obtain a plurality of preset candidate recommended objects.

[0055] In one of the embodiments, the recommendation receiving object may trigger a recommendation request through a terminal, so that the terminal may generate a recommendation request carrying a time tag according to the object identifier of the recommendation receiving object, and send the recommendation request to the server.

[0056] Step 204, for each of the multiple candidate recommended objects, determine the acceptance probability value of the candidate recommended object being accepted by the target recommended receiving object based on the receiving object information of the target recommended receiving object and the recommended object information of the targeted candidate recommended object.

[0057] Among them, the receiving object information refers to information related to the recommended receiving object, for example, the receiving object information includes at least one of the basic information and behavior information of the recommended receiving object. Basic information refers to the attribute information of the receiving object itself. For example, the basic information may include the gender, age, etc. of the receiving object. Behavior information refers to information related to the behavior triggered by the recommended receiving object. For example, the behavior information may include the acceptance information of the recommended receiving object recently accepted by the recommended receiving object, etc. The recommended object information of the candidate recommended object refers to information related to the candidate recommended object, for example, the recommended object information may include information such as the category to which the candidate recommended object belongs, the historical recommendation situation, the recommended value, etc.

[0058] Specifically, for each of the multiple candidate recommended objects, the computer device processes in the following manner. For the candidate recommended object, the computer device can determine the probability value of the candidate recommended object being accepted by the target recommended object based on the receiving object information of the target recommended object and the recommended object information of the candidate recommended object, and the probability value is the acceptance probability value.

[0059] In one embodiment, the computer device may determine similar recommendation receiving objects with similar information as the target recommendation receiving object based on the receiving object information of the target recommendation receiving object, and count the first total number of times the similar recommendation receiving objects receive the targeted candidate recommended objects in the historical time period, and count the second total number of times the similar recommendation receiving objects accept the targeted candidate recommended objects after receiving the targeted candidate recommended objects in the historical time period, and divide the second total number of times by the first total number of times to obtain the acceptance probability value of the targeted candidate recommended objects being accepted by the target recommendation receiving object. For example, the targeted candidate recommended objects may specifically be target commodities, and the computer device may count the total number of times the similar recommendation objects receive advertisements of the target commodities in the historical time period to obtain the first total number of times. The computer device may also count the total number of times the similar recommendation objects click on the link in the advertisement to jump to the specific details page of the target commodity after receiving the advertisement of the target commodity in the historical time period to obtain the second total number of times. And divide the second total number of times by the first total number of times to obtain the acceptance probability value of the targeted candidate recommended objects being accepted by the target recommendation receiving object. The acceptance here refers to the recommendation recipient clicking the link in the advertisement to jump to the specific detail page of the target product.

[0060] In one of the embodiments, the computer device may also obtain an acceptance rate estimation decision tree, and determine the acceptance probability value of the candidate recommended object being accepted by the target recommended receiving object based on the acceptance rate estimation decision tree. The acceptance rate estimation decision tree is specifically a binary tree, and each node in the acceptance rate estimation decision tree has one or more first classification features, so that starting from the root node of the acceptance rate estimation decision tree, based on the first classification features of each node, the node that matches the receiving object information of the target recommended receiving object and the recommended object information of the candidate recommended object can be determined until the leaf node, and the probability value corresponding to the leaf node is used as the acceptance probability value.

[0061] Step 206, determine the time preference value corresponding to each candidate recommended object or at least one of the target recommended receiving objects in the response time period; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or at least one of the corresponding target recommended receiving objects in the response time period.

[0062] Step 208, determining the recommendation scores of the multiple candidate recommended objects according to the corresponding time preference values ​​and the acceptance probability values ​​of the multiple candidate recommended objects; the recommendation scores are positively correlated with the acceptance probability values ​​of the corresponding candidate recommended objects and with the corresponding time preference values.

[0063] Among them, the time preference value corresponding to the candidate recommended object in the response time period represents the probability that the corresponding candidate recommended object is successfully recommended in the response time period. For example, the time preference value corresponding to the candidate recommended object in the response time period represents the probability value that after an advertisement including the candidate recommended object is sent to the corresponding recommendation receiving object in the response time period, the corresponding recommendation receiving object will click on the link in the advertisement to jump to the details page of the candidate recommended object. The time preference value corresponding to the target recommendation receiving object in the response time period represents the probability value that the target recommendation receiving object accepts the recommended object. For example, the time preference value corresponding to the recommendation receiving object in the response time period represents the probability value that after an advertisement including the corresponding recommended object is sent to the target recommendation receiving object in the response time period, the target recommendation receiving object will click on the link in the advertisement to jump to the details page of the corresponding recommended object.

[0064] Specifically, the computer device may determine the time preference value corresponding to the target recommendation receiving object in the response time period, and determine the recommendation scores of the multiple candidate recommended objects based on the time preference value corresponding to the target recommendation receiving object in the response time period and the acceptance probability values ​​corresponding to the multiple candidate recommended objects. Figure 3 For the candidate recommended object, the computer device multiplies the time preference value corresponding to the target recommendation receiving object in the response time period by the acceptance probability value of the candidate recommended object being accepted by the target recommendation receiving object to obtain the recommendation score of the candidate recommended object.

[0065] Alternatively, the computer device may determine the time preference value corresponding to each candidate recommended object in the response time period, and obtain the recommendation score of each candidate recommended object based on the corresponding reception probability value and the time preference value corresponding to the response time period of each candidate recommended object. Figure 3 The computer device may multiply the time preference value corresponding to the response time period of the candidate recommended object by the reception probability value of the candidate recommended object being accepted by the recommended receiving object to obtain the recommendation score of the candidate recommended object.

[0066] Alternatively, the computer device may determine the time preference value corresponding to each candidate recommended object in the response time period, and determine the time preference value corresponding to the target recommended receiving object in the response time period, and obtain the recommendation score of each candidate recommended object based on the time preference value corresponding to each candidate recommended object in the response time period, the time preference value corresponding to the target recommended receiving object in the response time period, and the corresponding reception probability value of each candidate recommended object. Figure 3The computer device may multiply the time preference value corresponding to the response time period of the candidate recommended object, the time preference value corresponding to the response time period of the target recommendation receiving object, and the acceptance probability value of the candidate recommended object by the recommended receiving object to obtain the recommendation score of the candidate recommended object. Figure 3 A schematic diagram of determining a recommendation score in one embodiment is shown.

[0067] Step 210 , based on the recommendation scores of the plurality of candidate recommended objects, select the recommended objects that are recommended to the target recommendation receiving object within the response time period from the plurality of candidate recommended objects.

[0068] Specifically, the computer device sorts the multiple candidate recommended objects in descending order to obtain a candidate recommended object sequence. The computer device selects a preset number of candidate recommended objects from the head of the candidate recommended object, and uses the selected candidate recommended objects as recommended objects to be recommended to the target recommendation receiving object.

[0069] In the above-mentioned method for determining the recommended object, by receiving the recommendation request, the target recommended receiving object, multiple candidate recommended objects and the response time period can be determined based on the recommendation request. By determining the target recommended receiving object, multiple candidate recommended objects and the response time period, the acceptance probability value of each candidate recommended object being accepted by the target recommended receiving object can be determined, as well as the time preference value of the target recommended receiving object in the response time period, and the time preference value of each candidate recommended object in the response time period can be determined. Since the time preference value reflects the recommendation success rate in the response time period, the recommendation score corresponding to each candidate recommended object can be determined based on the time preference value and the acceptance probability value, so as to screen out the recommended objects that are more likely to be accepted by the target recommended receiving object in the response time period based on the recommendation score. Since the present application takes the time factor into consideration when determining the recommended objects, the accuracy of the determined recommended objects can be improved based on the time factor.

[0070] In one of the embodiments, based on the receiving object information of the target recommended receiving object and the recommended object information of the targeted candidate recommended object, the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object is determined, including: in the case where the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object is not stored, an acceptance rate estimation decision tree is obtained; each node in the acceptance rate estimation decision tree has a corresponding first classification feature, and each leaf node has a corresponding probability value; based on the first classification feature, the receiving object information and the recommended object information of the targeted candidate recommended object corresponding to each node in the acceptance rate estimation decision tree, the first target leaf node to which the targeted candidate recommended object is classified in the acceptance rate estimation decision tree is determined; and the probability value corresponding to the first target leaf node is used as the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object.

[0071] Specifically, the computer device may check whether there is an acceptance probability value of the candidate recommended object being accepted by the target recommended receiving object. If not, a pre-created acceptance rate estimation decision tree may be obtained to determine the acceptance probability value based on the acceptance rate estimation decision tree. More specifically, the acceptance rate estimation decision tree may be obtained from the reference Figure 4 ,The acceptance rate estimation decision tree is a tree structure, in which each non-leaf node has a corresponding first classification feature, and each leaf node has a corresponding probability value. Figure 4 A schematic diagram of an acceptance rate estimation decision tree in one embodiment is shown. The first classification feature corresponding to the node in the acceptance rate estimation decision tree refers to a feature used to classify the splicing information, and the first classification feature corresponding to the node in the acceptance rate estimation decision tree may include at least one of an attribute of a recommendation object dimension and an attribute of a recommended object dimension.

[0072] Furthermore, for each of the multiple candidate recommended objects, the computer device may splice the recommendation receiving information of the target recommended receiving object and the recommended object information of the targeted recommended object to obtain spliced ​​information. The computer device classifies the spliced ​​information according to the first classification feature in the acceptance rate estimation decision tree to determine the node to which the spliced ​​information is classified, until the leaf node, and uses the leaf node to which the spliced ​​information is classified as the first target leaf node, and uses the probability value corresponding to the first target leaf node as the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object.

[0073] In this embodiment, since the acceptance probability value can be obtained by performing a classification operation based on the acceptance rate estimation decision tree, the efficiency of determining the acceptance probability value is improved.

[0074] In one of the embodiments, the first target leaf node to which the candidate recommended object is classified in the acceptance rate estimation decision tree is determined based on the first classification feature corresponding to each node in the acceptance rate estimation decision tree, the receiving object information and the recommended object information of the candidate recommended object, including: splicing the receiving object information and the recommended object information of the candidate recommended object to obtain splicing information; determining the current node from the root node of the acceptance rate estimation decision tree, and filtering out the target child node to which the splicing information is classified from the child nodes connected to the current node based on the splicing information, the first classification feature of the child node connected to the current node and the first classification feature value corresponding to each child node connected to the current node; taking the target child node as the new current node, returning the step of filtering out the target child node to which the splicing information is classified from the child nodes connected to the current node based on the splicing information, the first classification feature of the child node connected to the current node and the first classification feature value corresponding to each child node connected to the current node, and continuing to execute until a leaf node in the acceptance rate estimation decision tree is reached, and the arrived leaf node is taken as the first target leaf node.

[0075] Specifically, the splicing information may be, for example, [gender of the recommended recipient: female; age group of the recommended recipient: middle-aged; product category preferred by the recommended recipient: clothing; product category to which the candidate recommended recipient belongs: clothing; number of times the candidate recommended recipient has been viewed: 100 times…]. Among them, "gender of the recommended recipient", "age group of the recommended recipient", etc. are called attributes. Starting from the root node, the computer device can extract information matching the first classification feature corresponding to the current node from the splicing information, and based on the first classification feature value corresponding to the child node connected to the current node, determine the target child node matching the extracted information, and classify the splicing information to the target child node. And so on, until the leaf node. The first classification feature value refers to the value set for the first classification feature.

[0076] For example, refer to Figure 4, the first classification feature corresponding to the root node in the acceptance rate estimation decision tree is "gender", and the first classification feature value corresponding to the child node 1 connected to the root node is "male", and the first classification feature value corresponding to the child node 2 connected to the root node is "female". When the root node is used as the current node, the computer device can extract the information "gender of the recommended recipient: female" that matches the first classification feature "gender" from the splicing information. Since the first classification feature value corresponding to the child node 2 connected to the current node is "female", it matches the extracted information. Therefore, the splicing information can be classified into the child node 2 connected to the current node. Further, when subnode 2 is used as the current node, and the first classification feature corresponding to the second subnode connected to the root node is "age group", and the first classification feature value corresponding to subnode 3 connected to the current node is "youth", the first classification feature value corresponding to subnode 4 connected to the current node is "middle-aged", and the first classification feature value corresponding to subnode 5 connected to the current node is "old", the computer device can extract the information "recommended recipient age group: middle-aged" that matches the first classification feature "age group" from the spliced ​​information. Since the first classification feature value corresponding to subnode 4 connected to the current node is "middle-aged", it matches the extracted information, and therefore, the spliced ​​information can be classified into subnode 4 connected to the current node. And so on, until the leaf node to which the spliced ​​information is classified is determined, and the leaf node to which the spliced ​​information is classified is used as the first target leaf node.

[0077] In the above embodiment, the corresponding acceptance probability value can be determined by sequentially determining the nodes to which the splicing information is classified. Compared with the traditional method of determining the acceptance probability value through complex calculation, the efficiency of determining the acceptance probability value can be greatly improved.

[0078] In one of the embodiments, the steps for generating an acceptance rate prediction decision tree include: obtaining a first training set; the first training set includes multiple first training subsets; each first training subset includes multiple first attributes; the multiple first attributes include attributes of the recommended receiving object dimension and attributes of the recommended object dimension; based on the multiple first attributes, classifying the first training subset in the first training set layer by layer, corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, selecting the first classification feature corresponding to the corresponding node from the multiple first attributes, until the first preset classification end condition is reached, and obtaining a first intermediate decision tree; according to the classification results corresponding to each leaf node in the first intermediate decision tree, determining the probability value corresponding to each leaf node in the first intermediate decision tree, and combining the first intermediate decision tree and the probability value corresponding to each leaf node in the first intermediate decision tree to obtain an acceptance rate prediction decision tree.

[0079] Specifically, when it is necessary to generate a reception rate estimation decision tree, the computer device may obtain a first training set. The first training set includes multiple first training subsets, and each first training subset may be a subset obtained by collecting information for a corresponding recommendation receiving object. For example, it may be multiple subsets collected for a preset plurality of users. Each first training subset may include multiple first attributes, and data corresponding to each first attribute; the multiple first attributes include the first attribute in the above-mentioned splicing information. For example, referring to Figure 5 , the first training set can be Figure 5 The exposure feature matrix shown in the figure. Each row in the exposure feature matrix is ​​a first training subset. The "gender of the recommended recipient", "age group of the recommended recipient", "product category preferred by the recommended recipient", "product category to which the recommended recipient belongs", "whether to accept the recommended recipient", etc. in the exposure feature matrix are the first attributes; "female", "middle-aged", "1", etc. in the exposure feature matrix are the data corresponding to the first attributes. The last column in the exposure feature matrix is ​​the target variable, which indicates whether the recommended recipient accepts the recommended recipient. If the recommended recipient accepts the recommended recipient, the value is "1", and if the recommended recipient does not accept the recommended recipient, the value is "0". Figure 5 A schematic diagram of a first training set in one embodiment is shown.

[0080] Furthermore, the computer device can construct a reception rate estimation decision tree based on the first training set. The reception rate estimation decision tree is a decision tree, and the decision tree can implement a classification algorithm based on tree rules. The algorithm uses a tree structure to represent different decision paths and decision results, and classifies by continuously segmenting the data and using the first attribute that can maximize the classification purity. When constructing the reception rate estimation decision tree, the data in the first training set will be divided according to the first attribute value, and each node represents a first attribute or a first attribute set. The purpose of constructing the reception rate estimation decision tree is achieved by recursively selecting the first attribute. During the construction process, the reception rate estimation decision tree will select the best attribute to classify multiple first training subsets to form new branch nodes, and at the same time assign the classified first training subsets to the corresponding branch nodes. This process will iterate until the first preset classification end condition is reached, and the first intermediate decision tree is obtained, and a complete tree structure model is obtained based on the first intermediate decision tree. The first preset classification end condition may be freely set according to demand. For example, when the first classification feature cannot be determined for the end node in each path in the decision tree, it is considered that the first preset classification end condition is met.

[0081] In one of the embodiments, the first training set is collected only after obtaining the authorization of the corresponding recommendation recipient. The multiple first attributes in the first training set include attributes of the recommendation recipient dimension and attributes of the recommended object dimension. The attributes of the recommendation recipient dimension include, for example, "the gender of the recommendation recipient", "the age group of the recommendation recipient", and "the commodity category preferred by the recommendation recipient". The attributes of the recommended object dimension include, for example, "the commodity category to which the recommended object belongs", "the number of times the recommended object has been recommended", etc.

[0082] In one of the embodiments, in the process of constructing a reception rate estimation decision tree, when it is necessary to determine the first classification feature corresponding to the current node, the computer device determines the first training subset that has been classified to the current node. The computer device classifies the first training subset that has been classified to the current node based on each first attribute of the multiple first attributes to obtain a classification result corresponding to each first attribute. The computer device selects the best classification result from the multiple classification results, and uses the first attribute of the best classification result as the first classification feature corresponding to the current node. The computer device generates a child node connected to the current node, and classifies the first training subset that has been classified to the current node based on the first classification feature corresponding to the current node to obtain the first training subset classified to the child node connected to the current node. For example, when the first classification feature corresponding to the current node is "age group", the first training subsets classified to the current node are first training subsets 1 to 5, and subnodes 1 and 2 connected to the current node are generated, the first training subsets 1 to 5 can be classified based on the first classification feature corresponding to the current node, and the first training subsets classified to subnode 1 are first training subsets 1, 3, and 4, and the first training subsets classified to subnode 2 are first training subsets 2 and 5. Among them, in the first training subsets 1, 3, and 4, the data corresponding to the first attribute "age group" is "middle-aged"; in the first training subsets 2 and 5, the data corresponding to the first attribute "age group" is "youth".

[0083] The computer device uses the child node connected to the current node as the new current node, and iterates in this way until the first training subset that has been classified into the current node cannot be classified based on the first attribute or the combination of the first attributes, that is, until the first classification feature cannot be determined, and the current node is used as a leaf node. Among them, the process of classifying the first training subset based on the first attribute can refer to the above-mentioned process of classifying the splicing information according to the first classification feature in the acceptance rate estimation decision tree. This embodiment will not be repeated here. It is easy to understand that each leaf node in the first intermediate decision tree can no longer select the first classification feature based on the above steps. A leaf node refers to a node without child nodes.

[0084] In one embodiment, when it is necessary to select the optimal classification result, for each first attribute, the number of classification clusters obtained after classifying the corresponding multiple first training subsets based on the first attribute can be determined, wherein the data corresponding to the first attribute of the first training subsets in a single classification cluster are the same or similar. When determining the number of classification clusters corresponding to each first attribute, the first attribute with the least number of classification clusters can be used as the first classification feature, and the classification result with the least number of classification clusters can be used as the optimal classification result.

[0085] In one embodiment, when the first intermediate decision tree is obtained, the computer device may perform the following steps for each leaf node in the first intermediate decision tree. For the leaf node, the computer device may determine the first training subset classified to the leaf node, and determine the probability value corresponding to the leaf node based on the first training subset classified to the leaf node. When the probability value corresponding to each leaf node is determined, the computer device obtains the acceptance rate estimation decision tree.

[0086] In the above embodiment, by selecting the first classification feature from the plurality of first attributes, the node to which the splicing information belongs can be subsequently determined based on the first classification feature.

[0087] In one of the embodiments, according to the classification results corresponding to each leaf node in the first intermediate decision tree, the probability value corresponding to each leaf node in the first intermediate decision tree is determined, including: for each leaf node in the first intermediate decision tree, determining the total number of the first training subset classified to the targeted leaf node, and determining the number of positive samples in the first training subset classified to the targeted leaf node; the positive sample is a training subset whose attribute value of the target attribute is a preset value; the target attribute is an attribute that characterizes whether to accept the corresponding recommended object; and taking the ratio of the number of positive samples to the total number as the probability value corresponding to the targeted leaf node.

[0088] Specifically, when it is necessary to determine the probability value of the leaf node in the first intermediate decision tree, for each leaf node in the first intermediate decision tree, the computer device determines the first training subset classified to the leaf node, and counts the number of the first training subset classified to the leaf node, and the number is referred to as the total number. The computer device determines the number of positive samples in the first training subset classified to the leaf node, wherein the positive sample refers to the training subset whose attribute value of the target attribute is a preset value. The target attribute may specifically be "whether to accept the recommended object", and the preset value may specifically be "1", wherein "1" represents the acceptance of the recommended object. Further, the computer device uses the ratio of the number of positive samples to the total number as the probability value corresponding to the leaf node. For the convenience of description, the target attribute in the first training set is referred to as the first target attribute below. The target attribute in the second training set is referred to as the second target attribute, and the target attribute in the third training set is referred to as the third target attribute.

[0089] In this embodiment, since the positive samples are samples that have accepted the corresponding recommended objects, the ratio of the number of positive samples to the total number can reflect the probability that the corresponding recommendation recipient will accept the corresponding recommended object, so that the acceptance probability value can be quickly determined based on this probability.

[0090] In one of the embodiments, a time preference value corresponding to a response time period for each candidate recommended object or at least one of a target recommended receiving object is determined, including: determining a time preference value corresponding to the target recommended receiving object in the response time period based on receiving object information of the target recommended receiving object; determining a time preference value corresponding to each of a plurality of candidate recommended objects in the response time period based on respective recommended object information of a plurality of candidate recommended objects.

[0091] Specifically, the computer device can determine the time preference value corresponding to the response time period of the recommended receiving object based on the receiving object information of the target recommended receiving object, and determine the time preference value corresponding to each candidate recommended object in the response time period based on the recommended object information of each candidate recommended object.

[0092] In one of the embodiments, based on the receiving object information of the target recommended receiving object, determining the time preference value corresponding to the target recommended receiving object in the response time period includes: in the case where the time preference value corresponding to the target recommended receiving object is not stored, obtaining the receiving object time preference decision tree corresponding to the response time period; each non-leaf node in the receiving object time preference decision tree corresponding to the response time period has a corresponding second classification feature, and each leaf node has a corresponding probability value; based on the second classification feature corresponding to each non-leaf node in the receiving object time preference decision tree corresponding to the response time period, and the receiving object information of the target recommended receiving object, determining the second target leaf node to which the target recommended receiving object is classified in the receiving object time preference decision tree corresponding to the response time period; based on the probability value corresponding to the second target leaf node, determining the time preference value corresponding to the target recommended receiving object in the response time period.

[0093] Specifically, the target recommendation receiving object may be an object used to construct a receiving object time preference tree, so that after the receiving object time preference tree is constructed based on the target recommendation receiving object, the time preference value corresponding to the target recommendation receiving object in the response time period can be obtained, and the computer device can store the time preference value for subsequent use. When the target recommendation receiving object is not an object used to construct a receiving object time preference tree, at this time, it is necessary to obtain the time preference value of the target recommendation receiving object through the already constructed receiving object time preference tree.

[0094] Further, when you need to get the time preference value by receiving the object time preference tree, refer to Figure 6 , a receiving object time preference tree corresponding to each of the different time periods may be pre-set, so that the computer device can obtain the receiving object time preference tree corresponding to the response time period. Each non-leaf node in the receiving object time preference tree has a corresponding second classification feature, and each leaf node has a corresponding probability value. The computer device may start from the root node, and classify the target recommended receiving object based on the second classification feature of the current node and the receiving object information of the target recommended receiving object, so as to determine the child node of the current node to which the target recommended receiving object should be classified, and use the child node to be classified as the new current node. Iterate in this way until the leaf node. The computer device calls the leaf node to which the target recommended receiving object is finally classified as the second target leaf node, and uses the probability value corresponding to the second target leaf node as the time preference value corresponding to the target recommended receiving object in the response time period. Figure 6 A schematic diagram of determining a time preference value for a recommended receiving object in one embodiment is shown.

[0095] In one embodiment, for the current node in the receiving object time preference tree, the computer device can extract the target information from the receiving object information according to the second classification feature of the current node, and determine the child node to which the target information should be classified according to the second classification feature value corresponding to each child node connected to the current node. The child node is the child node to which the target recommended receiving object should be classified. Figure 6 , the specific receiving object information can be, for example, [gender of the recommended receiving object: female; age group of the recommended receiving object: middle-aged; category of goods preferred by the recommended receiving object: clothing; ...]. When the root node is the current node, the second classification feature of the current node is "gender of the recommended receiving object", and the classification values ​​corresponding to the child nodes connected to the current node are "male" and "female", the target information extracted from the receiving object information is "gender of the recommended receiving object: female". Since the classification value corresponding to the child node 1 connected to the current node is "female", the target recommended receiving object is classified into the child node 1.

[0096] In the above embodiment, the corresponding time preference value can be determined by only determining the nodes to which the target recommendation receiving object is classified in sequence. Compared with the traditional method of determining the time preference value through complex calculations, the efficiency of determining the time preference value can be greatly improved.

[0097] In one of the embodiments, the step of generating a time preference decision tree for a receiving object corresponding to a response time period includes: obtaining a second training set corresponding to the response time period; the second training set includes multiple second training subsets; each second training subset is information related to the response time period collected for a corresponding recommended object and includes multiple second attributes; based on the multiple second attributes, classifying the second training subsets in the second training set layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, selecting a second classification feature corresponding to the corresponding node from multiple second attributes, until the second preset classification end condition is reached, and obtaining a second intermediate decision tree; according to the classification results corresponding to each leaf node in the second intermediate decision tree, determining the probability value corresponding to each leaf node in the second intermediate decision tree, and combining the second intermediate decision tree and the probability value corresponding to each leaf node in the second intermediate decision tree to obtain a time preference decision tree for the receiving object corresponding to the response time period.

[0098] Specifically, when it is necessary to generate a time preference decision tree for a receiving object corresponding to a response time period, the computer device may obtain a second training set. The second training set includes a plurality of second training subsets, each of which may be a subset related to the response time period obtained by collecting information for the corresponding recommended receiving object. For example, when it is necessary to generate a time preference decision tree for a receiving object from 1:00 to 2:00, the computer device may determine the recommended object received by the corresponding recommended receiving object from 1:00 to 2:00, and determine whether the corresponding recommended receiving object accepted the received recommended object from 1:00 to 2:00, thereby generating a second training subset corresponding to the corresponding recommended receiving object based on the determined information and the corresponding recommended receiving object's own attribute information. At this time, the recommended object recommended to the corresponding recommended receiving object may be a randomly recommended object, or an object recommended by a traditional method.

[0099] Each second training subset may include multiple second attributes and data corresponding to each second attribute; the multiple second attributes include the attributes in the receiving object information of the target recommendation receiving object. Figure 7 , the second training set can be a receiving object feature matrix. Each row in the receiving object feature matrix is ​​a second training subset. The "gender of the recommended receiving object", "age group of the recommended receiving object", "category of goods preferred by the recommended receiving object", "time period for receiving the recommended object", "whether to accept the recommended object" and so on in the receiving object feature matrix are the second attributes; "female", "middle-aged", "1" and so on in the receiving object feature matrix are the data corresponding to the second attributes. The last column in the receiving object feature matrix is ​​the target variable, which indicates whether the recommended receiving object accepts the recommended object. If the recommended object is accepted, the value is "1", and if the recommended object is not accepted, the value is "0". Figure 7 A schematic diagram of a second training set in one embodiment is shown.

[0100] Further, the computer device can construct a receiving object time preference decision tree corresponding to the response time period based on the second training set. The receiving object time preference decision tree is a decision tree, and the decision tree can implement a classification algorithm based on tree rules. The algorithm uses a tree structure to represent different decision paths and decision results, and classifies by continuously segmenting data and using the second attribute that can maximize the classification purity. When constructing the receiving object time preference decision tree, the data in the second training set will be divided according to the second attribute value, and each node represents a second attribute or a second attribute set. The purpose of constructing the receiving object time preference decision tree is achieved by recursively selecting the second attribute. During the construction process, the receiving object time preference decision tree selects the best second attribute to classify multiple second training subsets, forming new branch nodes, and at the same time assigning the classified second training subsets to the corresponding branch nodes. This process will be iterated until the second preset classification end condition is reached, and the second intermediate decision tree is obtained, and a complete tree structure model is obtained based on the second intermediate decision tree. The second preset classification end condition can be freely set according to needs. For example, when the second classification feature cannot be determined for the end node in each path in the decision tree, it is considered that the second preset classification end condition is met.

[0101] Further, when the second intermediate decision tree is obtained, the computer device can perform the following steps for each leaf node in the second intermediate decision tree. For the targeted leaf node, the computer device can determine the second training subset classified to the targeted leaf node, and the computer device counts the number of the second training subsets classified to the targeted leaf node, and the number is referred to as the total number. The computer device determines the number of positive samples in the second training subset classified to the targeted leaf node, wherein the positive sample refers to the training subset whose attribute value of the second target attribute is a preset value. The second target attribute can specifically be "whether to accept the recommended object", and the preset value can specifically be "1", wherein "1" represents the acceptance of the recommended object. Further, the computer device uses the ratio of the number of positive samples to the total number as the probability value corresponding to the targeted leaf node. It is easy to understand that the specific process of constructing the receiving object time preference decision tree can refer to the above-mentioned process of constructing the acceptance rate estimation decision tree.

[0102] In one of the embodiments, each of a plurality of candidate recommended objects is used as a target candidate recommended object, and the step of determining the time preference value corresponding to the target candidate recommended object in the response time period includes: in the case where the time preference value corresponding to the target candidate recommended object in the response time period is not stored, obtaining a recommended object time preference decision tree corresponding to the response time period; each non-leaf node in the recommended object time preference decision tree has a corresponding third classification feature, and each leaf node has a corresponding probability value; according to the third classification feature corresponding to each non-leaf node in the recommended object time preference decision tree and the recommended object information of the target candidate recommended object, determining the third target leaf node to which the target candidate recommended object should be classified in the recommended object time preference decision tree corresponding to the response time period; and determining the time preference value corresponding to the target candidate recommended object in the response time period according to the probability value corresponding to the third target leaf node.

[0103] Specifically, each candidate recommended object among multiple candidate recommended objects is taken as an example for explanation. The target candidate recommended object can be an object used to construct a time preference decision tree for the recommended object, so that after the recommended object time preference decision tree is constructed based on the target candidate recommended object, the time preference value of the target candidate recommended object corresponding to the response time period can be obtained, and the computer device can store the time preference value for subsequent use. When the target candidate recommended object is not an object used to construct a recommended object time preference decision tree, at this time, it is necessary to obtain the time preference value of the target candidate recommended object through the recommended object time preference decision tree that has been constructed.

[0104] Further, when it is necessary to obtain the time preference value through the recommended object time preference decision tree, the recommended object time preference decision trees corresponding to different time periods can be pre-set, so that the computer device can obtain the recommended object time preference decision tree corresponding to the response time period. Each non-leaf node in the recommended object time preference decision tree has a corresponding third classification feature, and each leaf node has a corresponding probability value. The computer device can start from the root node, and classify the target candidate recommended object based on the third classification feature of the current node and the recommended object information of the target candidate recommended object to determine the child node of the current node to which the target candidate recommended object should be classified, and use the child node to which it should be classified as the new current node. Iterate in this way until the leaf node. The computer device calls the leaf node to which the target recommendation receiving object is finally classified as the third target leaf node, and uses the probability value corresponding to the third target leaf node as the time preference value corresponding to the target candidate recommended object in the response time period.

[0105] In one embodiment, for example, the recommended object information may specifically include [recommended object category: entity; recommended object recommended time period: 12:00 to 13:00; whether the recommended object is accepted: 1; ...]. When the root node is the current node, the third classification feature of the current node is "recommended object category", and the classification values ​​corresponding to the child nodes connected to the current node are "entity" and "virtual item", the target information extracted from the received object information is "recommended object category: entity". Since the classification value corresponding to the child node 1 connected to the current node is "entity", the target candidate recommended object is classified into the child node 1.

[0106] In the above embodiment, the corresponding time preference value can be determined by only determining the nodes to which the target candidate is classified by the recommended object in sequence. Compared with the traditional method of determining the time preference value through complex calculations, the efficiency of determining the time preference value can be greatly improved.

[0107] In one of the embodiments, the step of generating a time preference decision tree for a recommended object corresponding to a response time period includes: obtaining a third training set corresponding to the response time period; the third training set includes multiple third training subsets; each third training subset is information related to the response time period collected for the corresponding recommended object and includes multiple third attributes; based on the multiple third attributes, classifying the third training subsets in the third training set layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, selecting a third classification feature corresponding to the corresponding node from the multiple third attributes, until the third preset classification end condition is reached, and a third intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the third intermediate decision tree, the probability value corresponding to each leaf node in the third intermediate decision tree is determined, and the third intermediate decision tree and the probability value corresponding to each leaf node in the third intermediate decision tree are combined to obtain the time preference decision tree for the recommended object corresponding to the response time period.

[0108] Specifically, when it is necessary to generate a time preference decision tree of a recommended object corresponding to a response time period, the computer device may obtain a third training set. The third training set includes a plurality of third training subsets, each of which may be a subset related to the response time period obtained by collecting information for the corresponding recommended object. For example, when it is necessary to generate a time preference decision tree of a recommended object from 1 o'clock to 2 o'clock, the computer device may determine the recommendation information of the corresponding recommended object from 1 o'clock to 2 o'clock, and determine that the corresponding recommended object is accepted by the recommended receiving object from 1 o'clock to 2 o'clock, thereby generating a third training subset corresponding to the corresponding recommended object based on the determined information and the corresponding recommended object's own attribute information.

[0109] Each third training subset may include multiple third attributes and data corresponding to each second attribute; the multiple second attributes include the attributes in the recommended object information of the target candidate recommended object. Figure 8 , the third training set can be a recommended object feature matrix. Each row in the recommended object feature matrix is ​​a third training subset. The "category to which the recommended object belongs", "recommended time period", "whether accepted by the recommended recipient", etc. in the recommended object feature matrix are the third attributes; "entity", "12:00 to 13:00", "1", etc. in the recommended object feature matrix are the data corresponding to the third attributes. The last column in the recommended object feature matrix is ​​the target variable, which indicates whether the recommended recipient accepts it. If accepted by the recommended recipient, the value is "1", and if not accepted by the recommended recipient, the value is "0". Figure 8 A schematic diagram of a third training set in one embodiment is shown.

[0110] Further, the computer device may construct a receiving object time preference decision tree corresponding to the response time period based on the second training set. The recommended object time preference decision tree corresponding to the response time period is constructed. The construction process of the recommended object time preference decision tree may refer to the above-mentioned process of constructing the receiving object time preference decision tree corresponding to the response time period.

[0111] In one of the embodiments, when the third intermediate decision tree is obtained, the computer device may perform the following steps for each leaf node in the third intermediate decision tree. For the targeted leaf node, the computer device may determine the third training subset classified into the targeted leaf node, and the computer device counts the number of the third training subset classified into the targeted leaf node, and refers to the number as the total number. The computer device determines the number of positive samples in the third training subset classified into the targeted leaf node, wherein the positive sample refers to the training subset whose attribute value of the third target attribute is a preset value. The third target attribute may specifically be "whether it is accepted by the recommended recipient", and the preset value may specifically be "1", wherein "1" represents acceptance by the recommended recipient. Further, the computer device uses the ratio of the number of positive samples to the total number as the probability value corresponding to the targeted leaf node.

[0112] In one of the embodiments, the recommendation scores of multiple candidate recommended objects are determined according to the corresponding time preference values ​​and the acceptance probability values ​​of the multiple candidate recommended objects, including: determining the time correction factors corresponding to the multiple candidate recommended objects according to the time preference value corresponding to the target recommendation receiving object, the acceptance probability values ​​of the multiple candidate recommended objects and the corresponding time preference values; obtaining the acceptance values ​​corresponding to the multiple candidate recommended objects, and determining the recommendation scores of the multiple candidate recommended objects according to the acceptance values ​​and time correction factors corresponding to the multiple candidate recommended objects.

[0113] Specifically, for each of the multiple candidate recommended objects, the computer device processes in the following manner. The computer device obtains the time preference value of the candidate recommended object in the response time period, and multiplies the time preference value of the candidate recommended object in the response time period with the time preference value of the target recommendation receiving object in the response time period to obtain a time correction factor corresponding to the candidate recommended object. Further, the computer device multiplies the time correction factor with the accepted value of the candidate recommended object to obtain a recommendation score for the candidate recommended object. The accepted value can specifically be the price per click of the advertisement provided by the advertiser.

[0114] In this embodiment, the recommendation score is determined by a time correction factor, so that the determined recommendation score is strongly correlated with the response time period, and thus the recommendation score can accurately reflect the probability value of the corresponding candidate recommended object being accepted within the response time period and the probability value of the target recommendation recipient accepting the corresponding candidate recommended object, thereby improving the accuracy of the determined recommendation score.

[0115] In one of the embodiments, the method further includes: obtaining multiple sample receiving object time preference decision trees generated based on sample object information of multiple sample recommendation receiving objects; each sample receiving object time preference decision tree corresponds to a preset time period; determining the time propensity scores corresponding to the multiple sample recommendation receiving objects in each preset time period according to the multiple sample receiving object time preference decision trees; grouping the multiple sample recommendation receiving objects according to the time propensity scores corresponding to the multiple sample recommendation receiving objects in each preset time period to obtain sample recommendation receiving object groups corresponding to each preset time period; determining the recommendation success rate improvement result of the preset sample recommended objects in each preset time period based on the sample recommendation receiving object groups corresponding to each preset time period; the recommendation success rate improvement result is used to optimize the recommended object determination method.

[0116] Specifically, in order to determine the optimization method of the above-mentioned recommended object determination method, the result of improving the recommendation success rate can be determined. When it is necessary to determine the result of improving the recommendation success rate, the computer device can obtain multiple sample recommendation receiving objects, wherein the sample recommendation receiving objects are recommended receiving objects recommended by the above-mentioned recommended object determination method. Further, the computer device can generate multiple sample receiving object time preference decision trees based on the sample object information of the multiple sample recommendation receiving objects, wherein each sample receiving object time preference decision tree corresponds to a preset time period. The process of generating a sample receiving object time preference decision tree corresponding to the corresponding preset time period based on the sample object information can refer to the above-mentioned process of generating a receiving object time preference decision tree corresponding to the response time period.

[0117] Furthermore, according to the time preference decision trees of the sample receiving objects corresponding to the multiple preset time periods, the time propensity scores corresponding to the multiple sample recommendation receiving objects in each preset time period are determined. For example, if each hour of a day is regarded as a preset time period, then for each sample recommendation receiving object, 24 time propensity scores of the sample recommendation receiving objects can be determined based on the 24 sample receiving object time preference decision trees. Among them, the time propensity score is used to characterize the activity level of the corresponding sample recommendation receiving object in the corresponding preset time period, that is, to characterize the probability value of the corresponding sample recommendation receiving object accepting the recommended object in the corresponding preset time period.

[0118] Further, when the time tendency score of each sample recommendation receiving object in each preset time period is obtained, for each preset time period in multiple preset time periods, the computer device obtains the time tendency score of each sample recommendation receiving object in the preset time period, and clusters the obtained time tendency score, and divides the sample recommendation receiving objects with similar time tendency scores into one group, so as to obtain multiple sample recommendation receiving object groups corresponding to the preset time period. When the sample recommendation receiving object group corresponding to each preset time period is obtained, the computer device can determine the recommendation success rate improvement result of the preset sample recommended object in each preset time period according to the sample recommendation receiving object group corresponding to each preset time period, and determine the optimization method for optimizing the above-mentioned recommended object determination method through the recommendation success rate improvement result. Among them, the recommendation success rate improvement result is a result that indicates whether the recommendation success rate has been improved after the recommendation of the corresponding recommended object in the corresponding time period. The recommendation success can be that after the corresponding recommended object is recommended to the recommendation receiving object, the recommended receiving object accepts the recommended object.

[0119] For example, refer to Fig. 9 , Fig. 9A schematic diagram showing the improvement results of recommendation success rate corresponding to each preset time period in an embodiment is shown. Fig. 9 It can be found that the recommendation success rate is improved when the recommendation of the corresponding recommended object is made between 1:30 and 7:30, while the recommendation success rate is not improved when the recommendation of the corresponding recommended object is made between 7:50 and 10:50. At this time, when the algorithm is optimized, the recommendation of the corresponding recommended object is not made between 7:30 and 10:30.

[0120] In the above embodiment, by determining the recommendation success rate improvement result, the recommended object determination method can be optimized based on the recommendation success rate improvement result, thereby further improving the recommendation success rate based on the optimized recommended object determination method.

[0121] In one of the embodiments, each of the multiple sample recommendation receiving objects is respectively used as the target sample recommendation receiving object, and the step of determining the time propensity score corresponding to each of the target sample recommendation receiving objects in each preset time period includes: for each preset time period in the multiple preset time periods, determining the fourth target leaf node to which the target sample recommendation receiving object is classified in the sample receiving object time preference decision tree corresponding to the preset time period; and determining the time propensity score of the target sample recommendation receiving object in the preset time period according to the probability value corresponding to the fourth target leaf node.

[0122] Specifically, each sample recommendation receiving object among multiple sample recommendation receiving objects is used as a target sample recommendation receiving object as an example for explanation. For each preset time period among multiple preset time periods, the computer device obtains the sample receiving object time preference decision tree for the preset time period, and obtains the receiving object information of the target sample recommendation receiving object, which is called sample receiving object information. Based on the sample receiving object information, the computer device determines the leaf node to which the target sample recommendation receiving object should be classified in the corresponding sample receiving object time preference decision tree, and uses the leaf node as the fourth target leaf node. Among them, the corresponding sample receiving object time preference decision tree is the sample receiving object time preference decision tree for the preset time period. It is easy to understand that the process of determining the fourth target leaf node can refer to the above-mentioned process of determining the second target leaf node. This embodiment will not be repeated here.

[0123] Further, the computer device determines the probability value corresponding to the fourth target leaf node, and uses the probability value as the time preference score of the target sample recommended object in the preset time period. The process of determining the probability value corresponding to the leaf node in the sample receiving object time preference decision tree can refer to the process of determining the probability value corresponding to the leaf node in the receiving object time preference decision tree of Shanshu.

[0124] In one of the embodiments, a plurality of sample recommendation receiving objects are grouped according to their respective time propensity scores in each preset time period to obtain a sample recommendation receiving object group corresponding to each preset time period, including: for each preset time period in the plurality of preset time periods, a plurality of sample recommendation receiving objects are clustered according to the time propensity scores corresponding to the preset time period to obtain a sample recommendation receiving object group corresponding to the preset time period.

[0125] Specifically, when the time propensity score of each sample recommendation receiving object in each preset time period is obtained, for each preset time period in multiple preset time periods, the computer device obtains the time propensity score of each sample recommendation receiving object in the preset time period, and clusters the obtained time propensity scores, and divides the sample recommendation receiving objects with similar time propensity scores into one group, so as to obtain multiple sample recommendation receiving object groups corresponding to the preset time periods. For example, referring to Fig.10 , sample recommendation receiving objects 1 to 8 all have 24 time propensity scores. For the preset time period from 1 o'clock to 2 o'clock, the computer device can use receiving objects 1, 3, and 4 as a sample recommendation receiving object group corresponding to 1 o'clock to 2 o'clock, and use receiving objects 2, 5, 6, 7, and 8 as another sample recommendation receiving object group corresponding to 1 o'clock to 2 o'clock. Among them, the difference between the time propensity scores of receiving objects 1, 3, and 4 from 1 o'clock to 2 o'clock is less than or equal to the preset difference threshold. The difference between the time propensity scores of receiving objects 2, 5, 6, 7, and 8 from 1 o'clock to 2 o'clock is less than or equal to the preset difference threshold. Fig.10 A schematic diagram of a sample recommendation receiving object group in one embodiment is shown.

[0126] In one embodiment, for a preset time period, a preset number of sample recommendation receiving object groups corresponding to the preset time period may be generated. For example, for each preset time period, 10 sample recommendation receiving object groups corresponding to the preset time period may be generated.

[0127] In this embodiment, if the recommendation success rate improvement result is determined directly by the acceptance of the recommended object by multiple sample recommendation recipients, it is impossible to determine whether the recommendation success rate improvement result is caused by the difference caused by time or the difference caused by the recommendation algorithm. Therefore, it is necessary to group multiple sample recipients based on the time propensity score and divide similar sample recipients into one group to eliminate the interference of the recommendation algorithm. In this way, the degree of enhancement or attenuation of the recommended object by time as an interference factor can be understood from the sample recipient information of the grouped sample recipients, and the improvement direction of the algorithm after adding the time dimension modeling can be determined based on the determined enhancement or attenuation degree.

[0128] In one of the embodiments, each preset time period is respectively used as a target preset time period; the step of determining the recommendation success rate improvement result of the sample recommended objects preset in the target preset time period includes: for each sample recommendation receiving object group corresponding to the target preset time period, dividing the sample recommendation receiving object group into an experimental group and a control group; the sample recommendation receiving objects in the experimental group are objects that receive the sample recommended objects within the target preset time period; the sample recommendation receiving objects in the control group are objects that do not receive the sample recommended objects within the target preset time period; according to the acceptance result of each sample recommendation receiving object in the experimental group accepting the sample recommended object in the target preset time period, determining the first success probability value of the sample recommended object being successfully recommended within the target preset time period corresponding to the sample recommendation receiving object group; according to the acceptance result of each sample recommendation receiving object in the control group accepting the sample recommended object in the non-target preset time period, determining the second success probability value of the sample recommended object being successfully recommended within the non-target preset time period corresponding to the sample recommendation receiving object group; according to the first success probability value and the second success probability value corresponding to each sample recommendation receiving object group corresponding to the target preset time period, determining the recommendation success rate improvement result of the sample recommended objects preset in the target preset time period.

[0129] Specifically, when it is necessary to determine the recommendation success rate improvement result of the preset sample recommended object in the target preset time period, each preset time period can be used as the target preset time period for example. For each sample recommendation receiving object group corresponding to the target preset time period, the sample recommendation receiving object group targeted is divided into an experimental group and a control group. Among them, the sample recommendation receiving object in the experimental group is the object that receives the sample recommended object in the target preset time period; the sample recommendation receiving object in the control group is the object that does not receive the sample recommended object in the preset time period targeted. For example, when the sample recommended object is clothes, the target preset time period is from 1 o'clock to 2 o'clock, and there are two sample recommendation receiving object groups 1 to 2 from 1 o'clock to 2 o'clock, the computer device can divide the sample recommendation receiving object group 1 into the experimental group 1 and the control group 1, and divide the sample recommendation receiving object group 2 into the experimental group 2 and the control group 2, wherein the sample recommendation receiving objects in the experimental group 1 and the experimental group 2 are all recommended receiving objects that receive the recommended clothes from 1 o'clock to 2 o'clock; the sample recommendation receiving objects in the control group 1 and the control group 2 are recommended receiving objects that do not receive the recommended clothes from 1 o'clock to 2 o'clock.

[0130] Further, the computer device determines the acceptance result of each sample recommendation receiving object in the experimental group obtained based on the division of the sample recommendation receiving object group in the target preset time period, and determines the first success probability value of the sample recommended object being successfully recommended within the target preset time period corresponding to the experimental group according to the acceptance result, that is, determines the first success probability value of the sample recommended object being successfully recommended within the target preset time period corresponding to the sample recommendation receiving object group. For example, in the above example, when the sample recommendation receiving object group is sample recommendation receiving object group 1, the computer device determines the number of sample recommended objects in experimental group 1 who accept the recommended clothes between 1 o'clock and 2 o'clock, and divides the number of sample recommended objects who accept the recommended clothes between 1 o'clock and 2 o'clock by the total number of sample recommended objects in experimental group 1 to obtain the first success probability value.

[0131] Further, the computer device determines the acceptance result of each sample recommendation receiving object in the control group obtained based on the division of the sample recommendation receiving object group in the non-target preset time period, and determines the first success probability value of the sample recommended object being successfully recommended in the target preset time period corresponding to the control group according to the acceptance result. That is, the second success probability value of the sample recommended object being successfully recommended in the non-target preset time period corresponding to the sample recommendation receiving object group is determined. For example, in the above example, when the sample recommendation receiving object group is sample recommendation receiving object group 1, the computer device determines the number of sample recommended objects in the control group 1 who accept the recommended clothes in the remaining time except from 1 o'clock to 2 o'clock, and divides the number of sample recommended objects who accept the recommended clothes in the remaining time except from 1 o'clock to 2 o'clock in 1 day by the total number of sample recommended objects in the control group 1 to obtain the second success probability value.

[0132] Further, when the first success probability value and the second success probability value corresponding to each sample recommendation receiving object group corresponding to the target preset time period are obtained, the computer device averages the first success probability values ​​to obtain the first success probability mean value, and averages the second success probability values ​​to obtain the second success probability mean value. The second success probability mean value is subtracted from the first success probability mean value to obtain the recommendation success rate improvement result of the preset sample recommended objects in the target preset time period, for example, the recommendation success rate improvement value of the recommended clothes accepted by the sample recommendation receiving objects from 1 point to 2 points is obtained.

[0133] In one embodiment, the recommendation success rate improvement result can be characterized by ATE, referring to Fig.10If ATE=0, there will be no significant impact on the time-sharing delivery of the sample recommended objects. If ATE>0, the delivery will be improved within the time period. If ATE<0, the opposite is true.

[0134] In this embodiment, by dividing the sample recommendation receiving object group into an experimental group and a control group, the first success probability mean of accepting the corresponding sample recommended object within the target preset time period can be determined based on the experimental group, and the second success probability of accepting the sample recommended object within the non-target preset time period can be determined based on the control group. Therefore, based on the first success rate mean and the second success probability, it can be determined whether the success rate of recommending the sample recommended object to the sample recommended object within the target preset time period will increase, and if it will not increase, avoid recommending the sample recommended object to other recommendation receiving objects within the target preset time period. In this way, the recommendation success rate can be improved.

[0135] In one of the embodiments, when a recommendation request is received, before determining a target recommendation receiving object, multiple candidate recommended objects and a response time period according to the recommendation request, the method further includes: receiving a resource transfer request triggered by the target recommendation receiving object, and generating a recommendation request when performing a resource transfer operation based on the resource transfer request to obtain a resource transfer result; after screening out recommended objects recommended to the target recommendation receiving object within the response time period from the multiple candidate recommended objects according to the recommendation scores of the multiple candidate recommended objects, the method further includes: feeding back a result page to the target recommendation receiving object; the result page carries the resource transfer result and the recommended objects screened out from the multiple candidate recommended objects.

[0136] Specifically, when the target recommendation receiving object triggers a resource transfer request, the computer device may respond to the resource transfer request, trigger a resource transfer operation, and generate a recommendation request when the resource transfer result is obtained. Among them, the resources may be physical resources and virtual resources, and the physical resources may be power resources, CPU resources, currency resources, etc. For example, when the target recommendation receiving object is the target user and the resource transfer request is a mobile phone charge recharge request, the computer device may respond to the mobile phone charge recharge request, and generate a recommendation request when the recharge is successful. Further, when the recommended object recommended to the recommendation receiving object in the response time period is determined from multiple candidate recommended objects, the computer device may generate a feedback result page, wherein the feedback result page carries the recommended object. For example, when the target user recharges successfully, the computer device may determine that the recommended object is clothes, thereby triggering the target user's terminal to display a feedback result page including a clothes advertisement. If the target user clicks on the clothes advertisement in the feedback page, it indicates that the recommended object is accepted.

[0137] In this embodiment, by recommending recommended objects to the target recommendation receiving object, the probability of the target recommendation receiving object searching for the required recommended object from a large number of recommended objects is reduced, thereby improving the object search efficiency and further saving resources such as electricity consumed during object search.

[0138] In one embodiment, reference Fig.11 , Fig.11 A schematic diagram of a process for determining a recommended object in a specific embodiment is shown:

[0139] Step 1102, the computer device receives a resource transfer request triggered by a target recommendation receiving object, and generates a recommendation request when performing a resource transfer operation based on the resource transfer request to obtain a resource transfer result; determines the target recommendation receiving object, multiple candidate recommended objects and a response time period according to the recommendation request.

[0140] Step 1104, for each of the multiple candidate recommended objects, without storing the acceptance probability value of the candidate recommended object being accepted by the target recommendation receiving object, the computer device obtains an acceptance rate estimation decision tree; each node in the acceptance rate estimation decision tree has a corresponding classification feature, and each leaf node has a corresponding probability value.

[0141] Step 1106, the computer device determines the first target leaf node to which the candidate recommended object is classified in the acceptance rate estimation decision tree according to the classification characteristics, receiving object information and recommended object information of each node corresponding to the acceptance rate estimation decision tree; and uses the probability value corresponding to the first target leaf node as the acceptance probability value of the candidate recommended object being accepted by the target recommended receiving object.

[0142] Step 1108, when no time preference value corresponding to the target recommended receiving object is stored, the computer device obtains the receiving object time preference decision tree corresponding to the response time period; each non-leaf node in the receiving object time preference decision tree corresponding to the response time period has a corresponding classification feature, and each leaf node has a corresponding probability value.

[0143] Step 1110, the computer device determines the second target leaf node to which the target recommended receiving object is classified in the receiving object time preference decision tree corresponding to the response time period based on the classification features corresponding to each non-leaf node in the receiving object time preference decision tree corresponding to the response time period and the receiving object information of the target recommended receiving object; and determines the time preference value corresponding to the target recommended receiving object in the response time period based on the probability value corresponding to the second target leaf node.

[0144] Step 1112, when the time preference value corresponding to the target candidate recommended object in the response time period is not stored, the computer device obtains the recommended object time preference decision tree corresponding to the response time period; each non-leaf node in the recommended object time preference decision tree has a corresponding classification feature, and each leaf node has a corresponding probability value.

[0145] Step 1114, the computer device determines the third target leaf node to which the target candidate recommended object should be classified in the recommended object time preference decision tree corresponding to the response time period based on the classification features corresponding to each non-leaf node in the recommended object time preference decision tree and the recommended object information of the target candidate recommended object; and determines the time preference value corresponding to the target candidate recommended object in the response time period based on the probability value corresponding to the third target leaf node.

[0146] Step 1116: The computer device determines the time correction factor corresponding to each of the multiple candidate recommended objects according to the time preference value corresponding to the target recommendation receiving object, the acceptance probability values ​​of the multiple candidate recommended objects, and the corresponding time preference values.

[0147] Step 1118: The computer device obtains the acceptance value corresponding to each of the multiple candidate recommended objects, and determines the recommendation score of each of the multiple candidate recommended objects based on the acceptance value corresponding to each of the multiple candidate recommended objects and the time correction factor.

[0148] Step 1120, the computer device filters out recommended objects to be recommended to the target recommendation receiving object within the response time period from the multiple candidate recommended objects according to the respective recommendation scores of the multiple candidate recommended objects; and feeds back a result page to the target recommendation receiving object; the result page carries the resource transfer result and the recommended objects filtered out from the multiple candidate recommended objects.

[0149] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0150] The present application also provides an application scenario, and the application scenario applies the above-mentioned method for determining a recommended object. Specifically, the application of the method for determining a recommended object in the application scenario is as follows:

[0151] refer to Fig.12 , the target recommendation receiving object may be a target user, and the multiple candidate recommended objects may be multiple different commodities. When the target user successfully recharges the mobile phone bill, the server may determine the user commodity preference score (also known as the acceptance probability value) corresponding to each commodity. Among them, the user commodity preference score is determined based on the commodity characteristics (also known as the recommended object information of the candidate recommended object) and the object characteristics (also known as the receiving object information of the target recommendation receiving object). The server may determine the time preference value corresponding to each commodity, and determine the time preference values ​​corresponding to multiple users, wherein the time preference value corresponding to each commodity is determined based on the commodity characteristics. The multiple users include the target user, and the time preference values ​​corresponding to the multiple users are determined based on the object characteristics. The server sorts the multiple commodities according to the time preference value corresponding to the target user, the user commodity preference score corresponding to each commodity and the commodity click value (also known as the accepted value), and the time preference value corresponding to each commodity, and uses the top N commodities as commodities to be recommended to the target user. The computer device may also use the time preference values ​​corresponding to the multiple users as propensity scores for grouped causal inference, and perform recommendation success rate improvement results based on the causal inference results. The recommendation success rate improvement result is used to optimize the recommended object determination algorithm. The above-mentioned object characteristics and product characteristics are collected after user authorization. Fig.12 A schematic diagram of the overall process of a method for determining a recommended object in an embodiment is shown.

[0152] The present application also provides an application scenario, which applies the above-mentioned method for determining a recommended object. Specifically, the application of the method for determining a recommended object in the application scenario is as follows:

[0153] The target recommendation recipient may specifically be a target user, and the multiple candidate recommended objects may specifically be multiple different news information. When the target user desires to view the news information, the computer device may determine the news information to be recommended from the multiple candidate news information according to the above-mentioned recommended object determination method, and recommend the news information to be recommended to the target user. Thus, when the target user clicks on the recommended news information to view the specific news content, it can be considered that the target user has accepted the recommended news information.

[0154] The above application scenarios are merely illustrative illustrations. It should be understood that the application of the recommended object determination method provided in each embodiment of the present application is not limited to the above scenarios.

[0155] Based on the same inventive concept, the embodiment of the present application also provides a recommended object determination device for implementing the recommended object determination method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more recommended object determination device embodiments provided below can refer to the limitations of the recommended object determination method above, and will not be repeated here.

[0156] In one embodiment, Fig.13 As shown, a device 1300 for determining a recommended object is provided, comprising: an acceptance probability determination module 1302, a time preference value determination module 1304 and a recommendation module 1306, wherein:

[0157] The acceptance probability determination module 1302 is used to determine the target recommendation receiving object, multiple candidate recommended objects and the response time period according to the recommendation request when a recommendation request is received; for each of the multiple candidate recommended objects, based on the receiving object information of the target recommendation receiving object and the recommended object information of the candidate recommended object, determine the acceptance probability value of the candidate recommended object being accepted by the target recommendation receiving object.

[0158] The time preference value determination module 1304 is used to determine the time preference value corresponding to each candidate recommended object or at least one of the target recommended receiving objects in the response time period; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or at least one of the target recommended receiving objects in the response time period.

[0159] The recommendation module 1306 is used to determine the recommendation scores of multiple candidate recommended objects according to the corresponding time preference values ​​and the acceptance probability values ​​of the multiple candidate recommended objects; and to filter out the recommended objects that are recommended to the target recommendation receiving object within the response time period from the multiple candidate recommended objects according to the recommendation scores of the multiple candidate recommended objects.

[0160] In one of the embodiments, the acceptance probability determination module 1302 is also used to obtain an acceptance rate estimation decision tree when the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object is not stored; each node in the acceptance rate estimation decision tree has a corresponding classification feature, and each leaf node has a corresponding probability value; according to the classification features corresponding to each node in the acceptance rate estimation decision tree, the receiving object information and the recommended object information of the targeted candidate recommended object, the first target leaf node to which the targeted candidate recommended object in the acceptance rate estimation decision tree is classified is determined; and the probability value corresponding to the first target leaf node is used as the acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object.

[0161] In one of the embodiments, the acceptance probability determination module 1302 is also used to splice the receiving object information and the recommended object information of the candidate recommended object to obtain spliced ​​information; starting from the root node of the acceptance rate estimation decision tree, determine the current node, and based on the splicing information and the classification characteristics of the child nodes connected to the current node, filter out the target child node to which the splicing information is classified from the child nodes connected to the current node; use the target child node as the new current node, return to the step of filtering out the target child node to which the splicing information is classified from the child nodes connected to the current node based on the splicing information and the classification characteristics of the child nodes connected to the current node, and continue to execute until a leaf node in the acceptance rate estimation decision tree is reached, and the leaf node reached is used as the first target leaf node.

[0162] In one embodiment, the acceptance probability determination module 1302 is also used to obtain a first training set; the first training set includes multiple first training subsets; each first training subset includes multiple first attributes; the multiple first attributes include attributes of the recommended receiving object dimension and attributes of the recommended object dimension; based on the multiple first attributes, the first training subset in the first training set is classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the classification feature corresponding to the corresponding node is selected from the multiple first attributes, until the first preset classification end condition is reached, and a first intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the first intermediate decision tree, the probability value corresponding to each leaf node in the first intermediate decision tree is determined, and the acceptance rate estimation decision tree is obtained by combining the first intermediate decision tree and the probability value corresponding to each leaf node in the first intermediate decision tree.

[0163] In one embodiment, the acceptance probability determination module 1302 is also used to determine, for each leaf node in the first intermediate decision tree, the total number of first training subsets classified to the targeted leaf node, and determine the number of positive samples in the first training subset classified to the targeted leaf node; the positive sample is a training subset whose attribute value of the target attribute is a preset value; the target attribute is an attribute that characterizes whether the corresponding recommended object is accepted; and the ratio of the number of positive samples to the total number is used as the probability value corresponding to the targeted leaf node.

[0164] In one of the embodiments, the time preference value determination module 1304 is also used to determine the time preference value corresponding to the target recommendation receiving object in the response time period based on the receiving object information of the target recommendation receiving object; and to determine the time preference value corresponding to each of the multiple candidate recommended objects in the response time period based on the recommended object information of each of the multiple candidate recommended objects.

[0165] In one of the embodiments, the time preference value determination module 1304 is also used to obtain a receiving object time preference decision tree corresponding to a response time period when no time preference value corresponding to the target recommended receiving object is stored; each non-leaf node in the receiving object time preference decision tree corresponding to the response time period has a corresponding classification feature, and each leaf node has a corresponding probability value; based on the classification features corresponding to each non-leaf node in the receiving object time preference decision tree corresponding to the response time period and the receiving object information of the target recommended receiving object, determine the second target leaf node to which the target recommended receiving object is classified in the receiving object time preference decision tree corresponding to the response time period; based on the probability value corresponding to the second target leaf node, determine the time preference value corresponding to the target recommended receiving object in the response time period.

[0166] In one embodiment, the time preference value determination module 1304 is also used to obtain a second training set corresponding to the response time period; the second training set includes multiple second training subsets; each second training subset is information related to the response time period collected for the corresponding recommendation object and includes multiple second attributes; based on the multiple second attributes, the second training subsets in the second training set are classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the classification feature corresponding to the corresponding node is selected from the multiple second attributes until the second preset classification end condition is reached, and a second intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the second intermediate decision tree, the probability value corresponding to each leaf node in the second intermediate decision tree is determined, and the probability value corresponding to each leaf node in the second intermediate decision tree is combined to obtain the receiving object time preference decision tree corresponding to the response time period.

[0167] In one of the embodiments, each of the multiple candidate recommended objects is used as the target candidate recommended object, and the time preference value determination module 1304 is also used to obtain a recommended object time preference decision tree corresponding to the response time period when the time preference value corresponding to the target candidate recommended object in the response time period is not stored; each non-leaf node in the recommended object time preference decision tree has a corresponding classification feature, and each leaf node has a corresponding probability value; according to the classification features corresponding to each non-leaf node in the recommended object time preference decision tree and the recommended object information of the target candidate recommended object, determine the third target leaf node to which the target candidate recommended object should be classified in the recommended object time preference decision tree corresponding to the response time period; according to the probability value corresponding to the third target leaf node, determine the time preference value corresponding to the target candidate recommended object in the response time period.

[0168] In one embodiment, the time preference value determination module 1304 is also used to obtain a third training set corresponding to the response time period; the third training set includes multiple third training subsets; each third training subset is information related to the response time period collected for the corresponding recommended object, and includes multiple third attributes; based on the multiple third attributes, the third training subsets in the third training set are classified layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, the classification feature corresponding to the corresponding node is selected from the multiple third attributes, until the third preset classification end condition is reached, and a third intermediate decision tree is obtained; according to the classification results corresponding to each leaf node in the third intermediate decision tree, the probability value corresponding to each leaf node in the third intermediate decision tree is determined, and the third intermediate decision tree and the probability value corresponding to each leaf node in the third intermediate decision tree are combined to obtain the recommended object time preference decision tree corresponding to the response time period.

[0169] In one of the embodiments, the recommendation module 1306 is also used to determine the time correction factors corresponding to each of the multiple candidate recommended objects based on the time preference value corresponding to the target recommendation receiving object, the acceptance probability values ​​of the multiple candidate recommended objects and the corresponding time preference values; obtain the acceptance values ​​corresponding to each of the multiple candidate recommended objects, and determine the recommendation scores of each of the multiple candidate recommended objects based on the acceptance values ​​and the time correction factors corresponding to each of the multiple candidate recommended objects.

[0170] In one embodiment, the recommended object determination device 1300 also includes an optimization module, which is used to obtain multiple sample receiving object time preference decision trees generated based on sample object information of multiple sample recommendation receiving objects; each sample receiving object time preference decision tree corresponds to a preset time period; according to the multiple sample receiving object time preference decision trees, the time propensity scores corresponding to the multiple sample recommendation receiving objects in each preset time period are determined; according to the time propensity scores corresponding to the multiple sample recommendation receiving objects in each preset time period, the multiple sample recommendation receiving objects are grouped to obtain sample recommendation receiving object groups corresponding to each preset time period; based on the sample recommendation receiving object groups corresponding to each preset time period, the recommendation success rate improvement result of the preset sample recommended objects in each preset time period is determined; the recommendation success rate improvement result is used to optimize the recommended object determination method.

[0171] In one of the embodiments, each of the multiple sample recommendation receiving objects is used as the target sample recommendation receiving object, and the optimization module is also used to determine, for each preset time period in the multiple preset time periods, a fourth target leaf node to which the target sample recommendation receiving object is classified in the sample receiving object time preference decision tree corresponding to the preset time period; and determine the time propensity score of the target sample recommendation receiving object in the preset time period based on the probability value corresponding to the fourth target leaf node.

[0172] In one of the embodiments, the optimization module is also used to cluster multiple sample recommendation receiving objects according to the time tendency score corresponding to the preset time period for each preset time period in multiple preset time periods, so as to obtain a sample recommendation receiving object group corresponding to the preset time period.

[0173] In one of the embodiments, each preset time period is used as a target preset time period; the optimization module is also used to divide the sample recommendation receiving object group corresponding to the target preset time period into an experimental group and a control group; the sample recommendation receiving object in the experimental group is the object that receives the sample recommended object within the target preset time period; the sample recommendation receiving object in the control group is the object that does not receive the sample recommended object within the target preset time period; according to the acceptance result of each sample recommendation receiving object in the experimental group accepting the sample recommended object in the target preset time period, the first success probability value of the sample recommended object being successfully recommended within the target preset time period corresponding to the sample recommendation receiving object group is determined; according to the acceptance result of each sample recommendation receiving object in the control group accepting the sample recommended object in the non-target preset time period, the second success probability value of the sample recommended object being successfully recommended within the non-target preset time period corresponding to the sample recommendation receiving object group is determined; according to the first success probability value and the second success probability value corresponding to each sample recommendation receiving object group corresponding to the target preset time period, the recommendation success rate improvement result of the preset sample recommended object in the target preset time period is determined.

[0174] In one of the embodiments, the acceptance probability determination module 1302 is also used to receive a resource transfer request triggered by a target recommendation receiving object, and generate a recommendation request when performing a resource transfer operation based on the resource transfer request to obtain a resource transfer result; the recommendation module 1306 is also used to feed back a result page to the target recommendation receiving object; the result page carries the resource transfer result and a recommended object selected from multiple candidate recommended objects.

[0175] Each module in the above-mentioned recommended object determination device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0176] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.14 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store recommended object determination data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for determining a recommended object is implemented.

[0177] Those skilled in the art will understand that Fig.14 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0178] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0179] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0180] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0181] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and 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-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0182] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0183] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining a recommended object, characterized in that: The method comprises: When a recommendation request is received, determining a target recommendation recipient, multiple candidate recommended objects, and a response time period according to the recommendation request; For each of the plurality of candidate recommended objects, determining an acceptance probability value of the targeted candidate recommended object being accepted by the target recommended receiving object according to the receiving object information of the target recommended receiving object and the recommended object information of the targeted candidate recommended object; Determine a time preference value corresponding to the response time period for at least one of each of the candidate recommended objects or the target recommended receiving object; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or at least one of the target recommended receiving object in the response time period; Determining the recommendation scores of the multiple candidate recommended objects according to the corresponding time preference values ​​and the acceptance probability values ​​of the multiple candidate recommended objects; According to the recommendation scores of the multiple candidate recommended objects, the recommended objects to be recommended to the target recommendation receiving object in the response time period are screened out from the multiple candidate recommended objects.

2. The method according to claim 1, characterized in that The determining, based on the receiving object information of the target recommendation receiving object and the recommended object information of the targeted candidate recommended object, the acceptance probability value of the targeted candidate recommended object being accepted by the target recommendation receiving object comprises: In the case where there is no stored acceptance probability value of the candidate recommended object being accepted by the target recommended receiving object, obtaining an acceptance rate estimation decision tree; each node in the acceptance rate estimation decision tree has a corresponding first classification feature, and each leaf node has a corresponding probability value; Determine a first target leaf node to which the candidate recommended object is classified in the acceptance rate estimation decision tree according to the first classification feature corresponding to each node in the acceptance rate estimation decision tree, the receiving object information, and the recommended object information of the candidate recommended object; The probability value corresponding to the first target leaf node is used as the acceptance probability value of the candidate recommended object being accepted by the target recommendation receiving object.

3. The method according to claim 2, characterized in that The step of determining the first target leaf node to which the candidate recommended object is classified in the acceptance rate estimation decision tree according to the first classification feature corresponding to each node in the acceptance rate estimation decision tree, the receiving object information, and the recommended object information of the candidate recommended object, comprises: splicing the received object information and the recommended object information of the targeted candidate recommended object to obtain spliced ​​information; Starting from the root node of the acceptance rate estimation decision tree, a current node is determined, and according to the splicing information, the first classification feature of the child node connected to the current node and the classification feature values ​​corresponding to each child node connected to the current node, a target child node to which the splicing information is classified is screened from the child nodes connected to the current node; The target child node is taken as the new current node, and the step of filtering out the target child node to which the splicing information is classified from the child nodes connected to the current node according to the splicing information, the first classification feature of the child node connected to the current node, and the classification feature values ​​corresponding to each child node connected to the current node is returned, and the step of continuing to execute until a leaf node in the reception rate estimation decision tree is reached, and the leaf node reached is taken as the first target leaf node.

4. The method according to claim 2, characterized in that: The steps of generating the acceptance rate estimation decision tree include: Acquire a first training set; the first training set includes a plurality of first training subsets; each of the first training subsets includes a plurality of first attributes; the plurality of first attributes include attributes of a dimension of a recommended receiving object and attributes of a dimension of a recommended object; Based on the multiple first attributes, classify the first training subset in the first training set layer by layer, and select a first classification feature corresponding to the corresponding node from the multiple first attributes based on the classification result corresponding to the corresponding node in the corresponding layer, until the first preset classification end condition is reached, so as to obtain a first intermediate decision tree; According to the classification results corresponding to each leaf node in the first intermediate decision tree, the probability value corresponding to each leaf node in the first intermediate decision tree is determined, and the acceptance rate estimation decision tree is obtained by combining the first intermediate decision tree and the probability value corresponding to each leaf node in the first intermediate decision tree.

5. The method according to claim 4, characterized in that The determining, according to the classification results corresponding to each leaf node in the first intermediate decision tree, a probability value corresponding to each leaf node in the first intermediate decision tree, comprises: For each leaf node in the first intermediate decision tree, determine the total number of first training subsets classified to the targeted leaf node, and determine the number of positive samples in the first training subset classified to the targeted leaf node; the positive sample is the first training subset whose attribute value of the target attribute is a preset value; the target attribute is an attribute representing whether to accept the corresponding recommended object; The ratio of the number of the positive samples to the total number is used as the probability value corresponding to the targeted leaf node.

6. The method according to claim 1, characterized in that The determining of the time preference value corresponding to the response time period of at least one of each of the candidate recommended objects or the target recommended receiving object comprises: Determine, according to the recipient information of the target recommendation recipient, a time preference value corresponding to the target recommendation recipient in the response time period; According to the recommended object information of each of the multiple candidate recommended objects, the time preference value corresponding to each of the multiple candidate recommended objects in the response time period is determined.

7. The method according to claim 6, characterized in that The determining, according to the receiving object information of the target recommendation receiving object, the time preference value corresponding to the target recommendation receiving object in the response time period includes: In the case where no time preference value corresponding to the target recommended receiving object is stored, obtaining a receiving object time preference decision tree corresponding to the response time period; each non-leaf node in the receiving object time preference decision tree corresponding to the response time period has a corresponding second classification feature, and each leaf node has a corresponding probability value; Determine the second target leaf node to which the target recommended receiving object is classified in the receiving object time preference decision tree corresponding to the response time period according to the second classification feature corresponding to each non-leaf node in the receiving object time preference decision tree corresponding to the response time period and the receiving object information of the target recommended receiving object; According to the probability value corresponding to the second target leaf node, the time preference value corresponding to the target recommendation receiving object in the response time period is determined.

8. The method according to claim 7, characterized in that The step of generating a receiving object time preference decision tree corresponding to the response time period comprises: Acquire a second training set corresponding to the response time period; the second training set includes a plurality of second training subsets; each of the second training subsets is information related to the response time period collected for a corresponding recommendation object, and includes a plurality of second attributes; Based on the multiple second attributes, classify the second training subset in the second training set layer by layer, and select a second classification feature corresponding to the corresponding node from the multiple second attributes based on the classification result corresponding to the corresponding node in the corresponding layer, until the second preset classification end condition is reached, so as to obtain a second intermediate decision tree; According to the classification results corresponding to each leaf node in the second intermediate decision tree, the probability value corresponding to each leaf node in the second intermediate decision tree is determined, and the second intermediate decision tree and the probability value corresponding to each leaf node in the second intermediate decision tree are combined to obtain the receiving object time preference decision tree corresponding to the response time period.

9. The method according to claim 6, characterized in that Taking each of the plurality of candidate recommended objects as a target candidate recommended object, the step of determining the time preference value of the target candidate recommended object corresponding to the response time period includes: In the case where the time preference value of the target candidate recommended object corresponding to the response time period is not stored, obtaining a recommended object time preference decision tree corresponding to the response time period; each non-leaf node in the recommended object time preference decision tree has a corresponding third classification feature, and each leaf node has a corresponding probability value; Determine, according to the third classification feature corresponding to each non-leaf node in the recommended object time preference decision tree corresponding to the response time period, and the recommended object information of the target candidate recommended object, the third target leaf node to which the target candidate recommended object should be classified in the recommended object time preference decision tree corresponding to the response time period; According to the probability value corresponding to the third target leaf node, the time preference value corresponding to the target candidate recommended object in the response time period is determined.

10. The method according to claim 9, characterized in that The step of generating the recommended object time preference decision tree corresponding to the response time period includes: Acquire a third training set corresponding to the response time period; the third training set includes a plurality of third training subsets; each of the third training subsets is information related to the response time period collected for a corresponding recommended object, and includes a plurality of third attributes; Based on the multiple third attributes, classify the third training subset in the third training set layer by layer, and corresponding to each classification result in each layer, based on the classification result corresponding to the corresponding node in the corresponding layer, select a third classification feature corresponding to the corresponding node from the multiple third attributes, until a third preset classification end condition is reached, to obtain a third intermediate decision tree; According to the classification results corresponding to each leaf node in the third intermediate decision tree, the probability value corresponding to each leaf node in the third intermediate decision tree is determined, and the third intermediate decision tree and the probability value corresponding to each leaf node in the third intermediate decision tree are combined to obtain a recommended object time preference decision tree corresponding to the response time period.

11. The method according to claim 1, characterized in that: The determining, according to the corresponding time preference values ​​and the acceptance probability values ​​of the plurality of candidate recommended objects, the recommendation scores of the plurality of candidate recommended objects respectively comprises: Determine the time correction factor corresponding to each of the multiple candidate recommended objects according to the time preference value corresponding to the target recommendation receiving object, the acceptance probability values ​​corresponding to each of the multiple candidate recommended objects, and the time preference values ​​corresponding to each of the multiple candidate recommended objects; The acceptance values ​​corresponding to the multiple candidate recommended objects are obtained, and the recommendation scores of the multiple candidate recommended objects are determined according to the acceptance values ​​corresponding to the multiple candidate recommended objects and the time correction factor.

12. The method according to claim 1, characterized in that The method further comprises: Acquire multiple sample receiving object time preference decision trees generated based on sample object information of multiple sample recommendation receiving objects; each of the sample receiving object time preference decision trees corresponds to a preset time period; Determine, according to the time preference decision trees of the multiple sample receiving objects, the time tendency scores corresponding to each of the multiple sample recommendation receiving objects in each of the preset time periods; According to the time tendency scores corresponding to the multiple sample recommendation receiving objects in each of the preset time periods, the multiple sample recommendation receiving objects are grouped to obtain sample recommendation receiving object groups corresponding to each of the preset time periods; Based on the sample recommendation receiving object group corresponding to each of the preset time periods, the recommendation success rate improvement result of the preset sample recommended objects in each of the preset time periods is determined; the recommendation success rate improvement result is used to optimize the recommended object determination method.

13. The method according to claim 12, characterized in that taking each of the plurality of sample recommendation receiving objects as a target sample recommendation receiving object; The step of determining the time propensity score corresponding to each of the preset time periods of the target sample recommendation recipients includes: For each preset time period in a plurality of preset time periods, determining a fourth target leaf node to which the target sample recommendation receiving object is classified in a sample receiving object time preference decision tree corresponding to the preset time period; According to the probability value corresponding to the fourth target leaf node, the time propensity score of the target sample recommendation recipient in the preset time period is determined.

14. The method according to claim 12, characterized in that The step of grouping the plurality of sample recommendation receiving objects according to the time tendency scores corresponding to the plurality of sample recommendation receiving objects in each of the preset time periods to obtain a sample recommendation receiving object group corresponding to each of the preset time periods comprises: For each preset time period in the plurality of preset time periods, the plurality of sample recommendation receiving objects are clustered according to the time tendency score corresponding to the preset time period, so as to obtain a sample recommendation receiving object group corresponding to the preset time period.

15. The method according to claim 12, characterized in that Taking each of the preset time periods as a target preset time period; The step of determining the recommendation success rate improvement result of the preset sample recommended objects in the target preset time period includes: For each sample recommendation receiving object group corresponding to the target preset time period, the sample recommendation receiving object group is divided into an experimental group and a control group; the sample recommendation receiving objects in the experimental group are objects that have received the sample recommended objects within the target preset time period; the sample recommendation receiving objects in the control group are objects that have not received the sample recommended objects within the target preset time period; Determine, according to the acceptance result of each sample recommendation receiving object in the experimental group accepting the sample recommended object in the target preset time period, a first success probability value of successful recommendation of the sample recommended object within the target preset time period corresponding to the sample recommendation receiving object group; Determine, according to the acceptance result of each sample recommendation receiving object in the control group accepting the sample recommended object in the non-target preset time period, a second success probability value of the sample recommended object being successfully recommended within the non-target preset time period corresponding to the sample recommendation receiving object group; According to the first success probability value and the second success probability value respectively corresponding to each sample recommendation receiving object group corresponding to the target preset time period, the recommendation success rate improvement result of the sample recommended objects in the target preset time period is determined.

16. The method according to any one of claims 1 to 15, characterized in that When a recommendation request is received, before determining a target recommendation recipient, a plurality of candidate recommended objects and a response time period according to the recommendation request, the method further includes: receiving a resource transfer request triggered by the target recommendation receiving object, and generating a recommendation request when performing a resource transfer operation based on the resource transfer request to obtain a resource transfer result; After selecting the recommended objects recommended to the target recommendation receiving object in the response time period from the multiple candidate recommended objects according to the recommendation scores of the multiple candidate recommended objects, the method further includes: Feedback a result page to the target recommendation receiving object; the result page carries the resource transfer result and the recommended object screened from the multiple candidate recommended objects.

17. A device for determining a recommended object, characterized in that: The device comprises: The acceptance probability determination module is used to determine, when receiving a recommendation request, a target recommendation receiving object, multiple candidate recommended objects and a response time period according to the recommendation request; for each of the multiple candidate recommended objects, determine an acceptance probability value of the candidate recommended object being accepted by the target recommendation receiving object according to the receiving object information of the target recommendation receiving object and the recommended object information of the candidate recommended object; A time preference value determination module is used to determine the time preference value corresponding to the response time period for at least one of the candidate recommended objects or the target recommended receiving objects; the time preference value represents the probability of successful recommendation of the corresponding candidate recommended object or at least one of the target recommended receiving objects in the response time period; The recommendation module is used to determine the recommendation score of each of the multiple candidate recommended objects based on the corresponding time preference value and the acceptance probability value of each of the multiple candidate recommended objects; and to filter out the recommended objects to be recommended to the target recommendation receiving object within the response time period from the multiple candidate recommended objects based on the recommendation scores of each of the multiple candidate recommended objects.

18. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 16 are implemented.

19. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 16 are implemented.

20. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 16 are implemented.

Images