Content recommendation method and device, equipment and storage medium

Abstract

The invention discloses a content recommendation method and device, equipment and a storage medium, and relates to the technical field of Internet. The method comprises the steps that under the condition that a first payment terminal completes order payment, payment behavior data are acquired, the payment behavior data comprise transaction data of an order and identity data of the first payment terminal, and a content recommendation strategy is determined according to the payment behavior data, recommendation content corresponding to the recommendation strategy can be recommended to the first payment terminal through the settlement page of the order. Therefore, personalized recommendation can be carried out on the payment terminal through the settlement page of the order according to the order transaction data of the payment terminal and the identity data of the first payment terminal by the edge cloud server, the requirements of the user can be better met, and the accuracy of content recommendation is improved.

Description

Technical Field

[0001] The present application belongs to the field of Internet technology, and in particular, relates to a content recommendation method, device, equipment and storage medium. Background Art

[0002] With the continuous development of Internet technology, recommended content can be displayed in the order settlement interface after consumers pay for the order, such as resource exchange coupons, service release information, etc. However, these recommended contents are usually fixed content provided by the recommendation system, which usually does not meet user needs and reduces the accuracy of content recommendations. Summary of the invention

[0003] The embodiments of the present application provide a content recommendation method, apparatus, device and storage medium to at least solve the problem of low accuracy of content recommendation in the related art.

[0004] In a first aspect, an embodiment of the present application provides a content recommendation method, which is applied to an edge cloud server. The method may include:

[0005] When the first payment terminal completes the order payment, the payment behavior data is obtained, and the payment behavior data includes the transaction data of the order and the identity data of the first payment terminal;

[0006] Determine content recommendation strategy based on payment behavior data;

[0007] Recommended content corresponding to the recommendation strategy is recommended to the first payment terminal through the settlement page of the order.

[0008] In a second aspect, an embodiment of the present application provides a content recommendation device, which is applied to an edge cloud server, and the device may include:

[0009] An acquisition module, used to acquire payment behavior data when the first payment terminal completes order payment, the payment behavior data including transaction data of the order and identity data of the first payment terminal;

[0010] A determination module, used to determine content recommendation strategies based on payment behavior data;

[0011] The recommendation module is used to recommend the recommended content corresponding to the recommendation strategy to the first payment terminal through the settlement page of the order.

[0012] In a third aspect, an embodiment of the present application provides a computer device, the computer device comprising: a processor and a memory storing computer program instructions;

[0013] When the processor executes the computer program instructions, the content recommendation method shown in the first aspect is implemented.

[0014] In a fourth aspect, an embodiment of the present application provides a computer storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the content recommendation method as shown in the first aspect is implemented.

[0015] In a fifth aspect, an embodiment of the present application provides a chip, the chip including a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the content recommendation method shown in the first aspect.

[0016] In a sixth aspect, an embodiment of the present application provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the content recommendation method as shown in the first aspect.

[0017] The content recommendation method, device, equipment and storage medium of the embodiment of the present application can obtain payment behavior data when the first payment terminal completes the order payment. The payment behavior data includes the transaction data of the order and the identity data of the first payment terminal, and determine the content recommendation strategy based on the payment behavior data. The recommended content corresponding to the recommendation strategy can be recommended to the first payment terminal through the settlement page of the order. In this way, the edge cloud server can make personalized recommendations to the payment terminal through the settlement page of the order based on the order transaction data of the payment terminal and the identity data of the first payment terminal, which can better meet the needs of the payment terminal and improve the accuracy of content recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 A schematic diagram of the structure of a content recommendation system provided in an embodiment of the present application;

[0020] Figure 2 A schematic diagram of the structure of a content feedback model in a content recommendation system provided in an embodiment of the present application;

[0021] Figure 3 A flowchart of a content recommendation method provided in an embodiment of the present application;

[0022] Figure 4 is a structural diagram of a content recommendation device provided by an embodiment of the present application;

[0023] Figure 5 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0024] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0025] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0026] The acquisition, storage, use, and processing of data (including but not limited to the features and information in the text) in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0027] In order to solve the problems in the related technology, the following Figures 1 to 5 , the methods, devices, computer equipment and storage media of the embodiments of the present application are described in detail. It should be noted that these embodiments are not intended to limit the scope of the disclosure of the present application.

[0028] first, Figure 1 A schematic diagram of the structure of a content recommendation system provided in an embodiment of the present application.

[0029] like Figure 1 As shown, an embodiment of the present application provides a content recommendation system, which may include a three-layer interactive structure of a central cloud server, at least one cluster of edge cloud servers, and an electronic device at the payment end.

[0030] Among them, the central cloud server is used in the entire system to aggregate payment behavior data, reference recommendation content, feedback data on reference recommendation content, interactive behavior data and other data; as well as to train and update the global content feedback model and distribute the content feedback model to each cluster of edge cloud servers.

[0031] The edge cloud server is set in an area of ​​the electronic device close to the payment terminal, and is used to obtain the payment behavior data of the first payment terminal, and determine the content recommendation strategy according to the payment behavior data through the content feedback model issued by the central cloud server, and recommend the recommended content corresponding to the recommendation strategy to the first payment terminal through the settlement page of the order; and, interact with other edge cloud servers, and update the content feedback model based on the content feedback model provided by other edge cloud servers, obtain the updated content feedback model, and synchronize the updated content feedback model to other edge cloud servers; and interact with the central cloud server, and request the central cloud server to update the data based on the content feedback model, update the content feedback model, obtain the updated content feedback model, and update the content feedback model locally stored in the edge cloud server to the updated content feedback model issued by the central cloud server.

[0032] The electronic device at the payment end may be an electronic device with a payment function, which is used to upload aggregated payment behavior data, feedback data on reference recommended content, and other data to the edge cloud server. Among them, the electronic device includes but is not limited to a terminal, and may also be other devices other than a terminal. Exemplarily, the electronic device may be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmentedreality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personaldigital assistant, PDA), etc. It may also be a server, a network attached storage (Network AttachedStorage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiments of the present application are not specifically limited.

[0033] Based on the above content recommendation system, the embodiment of the present application provides the following content recommendation method.

[0034] Step 1: The central cloud server pre-processes the payment behavior data, reference recommended content, feedback data on the reference recommended content, interactive behavior data, etc. of each cluster of edge cloud servers. The central cloud server extracts at least one of the following data from the aforementioned data: transaction data of orders such as payment amount, payment period, and payment location; identity data that can be used to characterize user age, gender, etc.; recommended content such as recommended content tags and content classifications; feedback data such as the number of times the recommended content is recommended and the number of times the recommended content is visited, and performs data cleaning and desensitization on the aforementioned data.

[0035] Step 2: The central cloud server can use a neural network model. The structure of the neural network model can be as follows: Figure 2 As shown, the payment behavior data and the reference recommended content of each type of reference recommended content are learned, and the complex feature interaction relationship is mined to capture the relationship between the recommended content and the payment behavior. Based on this, the following formula (1) is shown:

[0036] H0=X

[0037] H l =f(W l ·H l-1 +b l ),l=1,2...,L

[0038]

[0039] Among them, W l is the weight matrix of the first layer of the neural network model, H l is the output of the lth hidden layer, X is the input feature such as payment behavior data, reference recommendation content and other data features, f is the activation function (such as ReLU), b l is the bias term, It is the predicted value of the neural network model, which is the expected feedback data of the payment end for the L-type reference recommended content, such as positive feedback data or negative feedback data, wherein the positive feedback data is used to represent the data of the users of the payment end accessing the L-type reference recommended content, and the negative feedback data is used to represent the data of the users of the payment end not accessing the L-type reference recommended content.

[0040] Then, the weighted cross entropy loss is used to optimize the neural network model, specifically combined with the following formula (2):

[0041]

[0042] Among them, y iis the true value, i.e., the real feedback data of the users on the payment end to the reference recommended content; N is the number of samples, i.e., the number of payment behavior data, reference recommended content, feedback data on the reference recommended content, etc. on the payment end. When the difference between the true value and the predicted value of the neural network model is less than or equal to the preset difference value, the trained global content feedback model of the multi-cluster edge cloud server is obtained, which is also called the global content feedback model.

[0043] Step 3: The central cloud server distributes the trained global content feedback model to the edge cloud servers of each cluster and provides resource replacement data P corresponding to the feedback data of the reference recommended content. i and the number of times the reference recommended content is recommended R i .

[0044] Step 4: Complete the order payment through the payment system installed on the electronic device of the payment end, trigger the order payment callback, and the payment behavior data related to the payment callback is sent by the payment system to the edge cloud server, where the payment behavior data includes the transaction data of the order and the identity data of the payment end.

[0045] Step 5: The edge cloud server receives the payment behavior data and determines the content recommendation strategy. The edge cloud server can input the payment behavior data into the global content feedback model issued by the central cloud server, and determine the first feedback data expected by the payment end for each type of reference recommended content in the N types of reference recommended content based on the payment behavior data through the global content feedback model. In order to ensure the balance between personalization and diversification of recommended content, the upper confidence bound algorithm (UCB) is introduced in the embodiment of the present application to determine the content recommendation score of the payment end for the i-type reference recommended content based on UCB and multi-factor interaction behavior data, that is, It can be expressed by the following formula (3):

[0046]

[0047] Among them, R is the total number of times N types of reference recommended content are recommended, R i is the number of times the reference recommended content i is recommended, D i is the diversity score of the reference recommendation content i, w y 、w p and w d is the weight coefficient, α is the threshold parameter that controls the degree of exploration, which is used to control the balance between exploring the reference recommendation content and using the reference recommendation content, J is the number of categories of the reference recommendation content i feature, c(j) is the number of times feature j appears in the edge recommendation set, C is the sum of the number of times each reference recommendation content feature is recommended in the edge recommendation set, and Di The higher the value, the less likely the reference recommendation content appears in the user's interest distribution.

[0048] Based on the above formula (3), the content recommendation scores of N types of reference recommended content can be calculated, and the reference recommended content with the highest content recommendation score can be selected for recommendation to the payment terminal. Based on this, the content recommendation score with the highest content recommendation score can be selected through the following formula (4):

[0049]

[0050] Among them, {i1,i2,…,i k} is the index set of k reference recommended contents with the largest content recommendation scores, It is a collection of all reference recommended contents. represents the content recommendation score of the jth reference recommended content after sorting, and

[0051] Therefore, the k reference recommended contents with the largest content recommendation scores can be obtained, and based on the content distribution of the settlement page of the order, the content recommendation strategy of the k reference recommended contents on the settlement page can be determined. The content recommendation strategy includes but is not limited to the display duration and mode of the recommended contents.

[0052] Step 6: The electronic device at the payment end displays a settlement page, which includes k reference recommended contents with the largest content recommendation scores. The recommended contents include, but are not limited to, resource exchange vouchers, service release information, and the like. Based on this, the electronic device can determine the feedback data to be sent to the edge cloud server based on the user's access to the recommended contents on the settlement page, and put it on the edge cloud server so that the edge cloud server can update the local content feedback model of the edge cloud server based on the feedback data, payment behavior data, and the recommended contents displayed on the settlement page.

[0053] Therefore, the edge cloud server in the embodiment of the present application can combine the support of the content feedback model of the central cloud server to make personalized recommendations for recommended content based on the payment behavior data of the payment terminal, which can better meet the needs of the payment terminal. At the same time, the global content feedback model of the central cloud server and the local UCB algorithm of the edge cloud server can be combined to achieve distributed and local adaptive recommendations, thereby improving the accuracy of content recommendations. In addition, multi-factor interactive behavior data is added to the UCB algorithm, and the feedback data, resource replacement data, and diversity scores of the reference recommended content expected by the payment terminal for various reference recommended content are taken as targets, so that the recommended content can achieve a dynamic balance between resource replacement and diversity, thereby meeting the service requirements of the recommendation system.

[0054] It should be noted that the content recommendation method provided in the embodiment of the present application can be applied to any payment scenario where an order settlement page can be displayed.

[0055] Figure 2 A flowchart of a content recommendation method provided in an embodiment of the present application.

[0056] like Figure 2 As shown, the content recommendation method can be applied to Figure 1 For any edge cloud server shown in , the content recommendation method may specifically include the following steps:

[0057] Step 210, when the first payment terminal completes the order payment, the payment behavior data is obtained, and the payment behavior data includes the transaction data of the order and the identity data of the first payment terminal; Step 220, based on the payment behavior data, a content recommendation strategy is determined; Step 230, recommended content corresponding to the recommendation strategy is recommended to the first payment terminal through the settlement page of the order.

[0058] In this way, the edge cloud server can make personalized recommendations to the payment end through the order settlement page based on the payment end's order transaction data and the identity data of the first payment end, which can better meet the needs of the payment end and improve the accuracy of content recommendations.

[0059] The above steps are described in detail below, as shown below.

[0060] First, involving step 210, in some embodiments of the present application, the edge cloud server may obtain payment behavior data sent by the first payment terminal, wherein the payment behavior data may include transaction data of the order and identity data of the first payment terminal.

[0061] Specifically, the transaction data includes but is not limited to at least one of the following: payment amount, payment period, and payment location. The identity data includes but is not limited to at least one of the following: user age and user gender.

[0062] Secondly, regarding step 220, in some embodiments of the present application, step 220 may specifically include step 2201 and step 2202.

[0063] Step 2201: Determine, based on the payment behavior data, first feedback data of the first payment terminal for each type of reference recommended content in N types of reference recommended content, where N is a positive integer.

[0064] Exemplarily, the first feedback data expected by the first payment end for each type of reference recommended content can be determined based on the payment behavior data of the first payment end, and the first feedback data includes positive feedback data or negative feedback data. Among them, the positive feedback data is used to characterize the data of the user of the first payment end accessing each type of reference recommended content in N categories, such as the data of the user of the first payment end clicking on each type of reference recommended content on the electronic device of the first payment end is recorded as positive feedback data. The negative feedback data is used to characterize the data of the user of the payment end not accessing each type of reference recommended content in N categories, such as the data of the user of the first payment end clicking on each type of reference recommended content on the electronic device of the first payment end within a preset time period is recorded as negative feedback data, or the data of the user of the first payment end not clicking on each type of reference recommended content on the electronic device of the first payment end is recorded as negative feedback data.

[0065] Step 2202, determining a content recommendation strategy based on the first feedback data and interactive behavior data of each type of reference recommended content; wherein the interactive behavior data includes at least one of the following: first resource replacement data corresponding to the first feedback data, a diversity score of each type of reference recommended content, and a recommendation value score of each type of reference recommended content.

[0066] The first resource replacement data corresponding to the first feedback data may include at least one of the following: a resource replacement amount for recommending each type of reference recommended content, and a resource replacement amount for feedback data. The resource replacement amount for positive feedback data is greater than the resource replacement amount for negative feedback data. The amount of data for the first resource replacement amount in the embodiment of the present application may be preset.

[0067] It should be noted that the specific determination process of the diversity score and the recommendation value score involved in step 2202 in the embodiment of the present application can be as follows.

[0068] In some embodiments of the present application, the diversity score is used to characterize the probability of each type of reference recommended content appearing in the distribution of recommended content interacted with by the first payment terminal. Based on this, before step 220, the content recommendation method may further include:

[0069] The diversity score of each category of reference recommendation content is determined according to the number of categories of features in each category of reference recommendation content, the number of times each category of reference recommendation content is recommended, and the total number of times N categories of reference recommendation content are recommended.

[0070] For example, the diversity score D of the i-th reference recommendation content in N categories is i It can be calculated by the following formula (4).

[0071]

[0072] Where J is the number of categories of the feature of reference recommendation content i, c(j) is the number of times feature j appears in the edge recommendation set, C is the sum of the number of times each reference recommendation content feature is recommended in the edge recommendation set, and D i The higher the value, the less likely the reference recommendation content appears in the user's interest distribution.

[0073] In some embodiments of the present application, the recommendation value score is used to characterize the probability that the feedback data of the first payment terminal for each type of reference recommended content is positive feedback data and the amount of resource replacement data corresponding to the feedback data of each type of reference recommended content is greater than or equal to a preset amount of data. Based on this, before step 220, the content recommendation method may further include:

[0074] The recommendation value score of each type of reference recommended content is determined according to the number of times each type of reference recommended content is recommended and the second resource replacement data with each type of reference recommended content.

[0075] For example, the recommendation value score U of the i-th reference recommendation content in N categories is i It can be calculated by the following formula (6).

[0076]

[0077] Among them, R is the total number of times N types of reference recommended content are recommended, R i is the number of times the i-th type of reference recommendation content is recommended, and α is the threshold parameter that controls the degree of exploration, which is used to control the balance between exploring the reference recommendation content and utilizing the reference recommendation content.

[0078] Therefore, the content recommendation method provided in the embodiment of the present application can introduce multi-factor interactive behavior data into the UCB algorithm, and use the expected feedback data of the reference recommended content, resource replacement data, the diversity score of each type of reference recommended content, and the recommendation value score of each type of reference recommended content as a comprehensive method to determine the content recommendation score of the reference recommended content. In this way, the diversity of the recommended content can be taken into account, and the problem of content homogeneity can be avoided. It can better meet the needs of the payment end and improve the accuracy of content recommendations.

[0079] In some embodiments of the present application, the above step 2201 may specifically include:

[0080] Through the content feedback model, according to the payment behavior data, the first feedback data of the first payment end for each type of reference recommended content is determined; wherein the content feedback model is trained by the reference payment behavior data, each type of reference recommended content and the second feedback data of each type of reference recommended content.

[0081] The content feedback model can be one of the following models: a global content feedback model issued by the central cloud server, an edge cloud server based on the payment behavior data of the payment terminal, such as Figure 2 The neural network model shown is obtained by self-training, the edge cloud server updates the content feedback model based on the content feedback model update data, and the global content feedback model is updated by the central cloud server.

[0082] Here, the content feedback model can be trained by the above formula (1) and formula (2).

[0083] Based on this, the embodiments of the present application provide at least two methods to update the content feedback model, as shown below.

[0084] In some embodiments of the present application, after step 2201 , the content recommendation method may further include steps 3101 and 3102 .

[0085] Step 3101, obtaining feedback data of the first payment end on the settlement page, the feedback data including positive feedback data of the first payment end on the recommended content or negative feedback data of the first payment end on the recommended content.

[0086] Exemplarily, the feedback data may be positive feedback data or negative feedback data. Among them, the positive feedback data is used to characterize the data of the user of the first payment end accessing the recommended content, such as the data of the user of the first payment end clicking the recommended content on the electronic device of the first payment end is recorded as positive feedback data. The negative feedback data is used to characterize the data of the user of the payment end not accessing the recommended content, such as the data of the user of the first payment end clicking the recommended content on the electronic device of the first payment end within a preset time period is recorded as negative feedback data, or the data of the user of the first payment end not clicking the recommended content on the electronic device of the first payment end is recorded as negative feedback data.

[0087] Step 3202, update the content feedback model based on the feedback data, payment behavior data and recommended content displayed on the settlement page.

[0088] Exemplarily, the content feedback model can be updated through the following steps, as shown below. The edge cloud server collects incremental data such as new feedback data, payment behavior data, and recommended content displayed on the settlement page. t , and periodically perform local training. For each time window t, the edge cloud server uses the current incremental data to train the local content feedback model. Specifically, the parameters of the updated content feedback model can be calculated by referring to the following formula (7):

[0089]

[0090] Among them, θ t+1is the parameter of the content feedback model after the t+1th round of incremental training, θ t is the parameter of the content feedback model after the tth round of incremental training, η is the learning rate, is the incremental data X in round t, based on feedback data, payment behavior data, and recommended content displayed on the checkout page t The calculated gradient.

[0091] It should be noted that the updated content feedback model can be trained through a set of data including newly added feedback data, payment behavior data and recommended content displayed on the settlement page in one time window, or it can be trained through at least two sets of data including newly added feedback data, payment behavior data and recommended content displayed on the settlement page in at least two time windows.

[0092] In some embodiments of the present application, in addition to incremental data consisting of feedback data, payment behavior data, and recommended content displayed on the settlement page received by the edge cloud server itself, the edge cloud server can also receive content feedback model update data sent by other edge cloud servers with which it has a communication relationship to update its own content feedback model. Based on this, after step 2201, the content recommendation method can also include steps 3103 to 3105.

[0093] Step 3103, receiving content feedback model update data sent by M edge cloud servers in a cluster, the content feedback model update data including the payment behavior data of the second payment terminal and the feedback data of the second payment terminal for each type of reference recommended content, where M is a positive integer.

[0094] Among them, the content feedback model update data may include incremental data consisting of feedback data related to the payment end obtained by a cluster of edge cloud servers, payment behavior data, and recommended content displayed on the settlement page.

[0095] Step 3104, updating the data according to the content feedback model, updating the content feedback model, and obtaining an updated content feedback model.

[0096] Exemplarily, step 3103 and step 3104 can be described in detail by the following content. After each period T1, the edge cloud server receives content feedback model update data uploaded by N edge cloud servers in a cluster, and the edge cloud server evaluates the model quality of each edge cloud server in a cluster. Specifically, the content feedback model can be evaluated by the quality evaluation data involved in the embodiment of the present application, and the uploaded incremental data can be aggregated according to the quality evaluation data by the following formula (8) to obtain the updated content feedback model θ region , and after the update is completed, the updated content feedback model is sent to each cluster of edge cloud servers.

[0097]

[0098] Where M is the number of edge cloud servers in a cluster, θ i Update data for the content feedback model uploaded by the i-th edge cloud server, Q i is the quality evaluation data of the i-th edge cloud server, D i is the number of training data samples processed by the ith edge cloud server. The combination of the two represents the data weight of the server. m The model evaluation quality of the edge cloud server representative is the sum of the data weights of all edge cloud servers, which can be expressed as follows:

[0099]

[0100] And, it should be noted that the specific process of updating the content feedback model can be updated through the above step 3202 and the content involved in formula (7), and the updated content feedback model is obtained, which will not be repeated here. Specifically, the content feedback model update data is used as X, and the content feedback model is retrained.

[0101] Step 3105: Send the updated content feedback model to each cluster of edge cloud servers.

[0102] Therefore, the embodiment of the present application can first update the content feedback model stored in itself based on the incremental data of the first payment end obtained by the edge cloud server, and further update the content feedback model in combination with the incremental data obtained by other clusters of edge servers, so as to share the updated content feedback model with other clusters of edge cloud servers. There is no need for each cluster of edge cloud servers to update the content feedback model separately, which significantly reduces the amount of resources required by a cluster of edge cloud servers to update the content feedback model. At the same time, it can achieve more accurate content recommendations based on regional characteristics, thereby improving the flexibility and adaptability of the recommendation system.

[0103] In some embodiments of the present application, before step 3105, the updated content feedback model is also evaluated. If the evaluation passes, the updated content feedback model can be sent to each cluster of edge cloud servers. Conversely, if the evaluation fails, the content feedback model can be updated in the following manner. Based on this, before step 3105, the content recommendation method can also include steps 3106 to 3108.

[0104] Step 3106, determining the model quality evaluation value of the content feedback model according to the quality evaluation data of the content feedback model update data, wherein the quality evaluation data includes at least one of the following: accuracy quality evaluation data, integrity quality evaluation data, and representative quality evaluation data. The accuracy quality evaluation data is used to characterize the accuracy of the content feedback model update data, the integrity quality evaluation data is used to characterize the completeness of the content feedback model update data, and the representative quality evaluation data is used to characterize the diversity of the types of features in the content feedback model update data.

[0105] For example, after each cycle T2, the edge cloud server as the edge cloud server representative will upload the content feedback model update data obtained locally to the central server. Similarly, the central server aggregates the incremental information uploaded by each cluster edge cloud server, such as the edge cloud server of cluster 1 and the representative edge cloud server of cluster 2, and updates the global model θ global , and after the update is completed, the model is sent back to each edge cloud server.

[0106] The edge cloud server can measure the model quality evaluation value Q of the updated content feedback model by determining the accuracy, completeness, and representativeness of the content feedback model update data. i , to determine whether the central cloud server needs to participate in updating the content feedback model, where the model quality evaluation value can be determined by the following formula (10), where α, β and γ are weight factors, and formula (10) is specifically as follows:

[0107]

[0108] in, Represents the accuracy quality evaluation data, which is used to characterize the accuracy of the content feedback model update data, that is, to reflect the user's actual behavior preferences. i represents the total number of recommended contents on edge cloud server i, Represents the total number of positive feedback data, that is, the total number of clicked content. It can be determined by the following formula (11):

[0109]

[0110] in, Integrity quality assessment data is used to characterize the completeness of the content feedback model update data, that is, to reflect whether the data record is comprehensive and whether there are any missing key fields. F represents the field feature number, represents the missing rate of the vth feature on edge cloud server i, then It can be determined by the following formula (12):

[0111]

[0112] in, Represents representative quality assessment data, which is used to characterize the diversity of the types of features in the content feedback model update data, that is, to reflect whether the data can fully cover different data features and avoid bias. The embodiment of the present application reflects representativeness by calculating the entropy of the features of each recommended content. The higher the feature entropy, the more uniform the category distribution of the dimensional features is, and the more extensive the behaviors covered are. v represents the number of categories of feature v, p j represents the probability of category j appearing in the data, It can be determined by the following formula (13):

[0113]

[0114] Step 3107, when the model quality evaluation value is less than or equal to the preset threshold, a new request is sent to the central cloud server, the new request carries the content feedback model update data, and the new request is used to request the central cloud server to update the content feedback model in the central cloud server according to the content feedback model update data and the content feedback model update data sent by P two-cluster edge cloud servers, and obtain an updated content feedback model, where P is a positive integer.

[0115] Step 3108: upon receiving the updated content feedback model sent by the central cloud server, the updated content feedback model is synchronized to each cluster of edge cloud servers, and the content feedback model is replaced with the updated content feedback model.

[0116] For example, the above steps 3107 and 3108 can be illustrated. When the incremental model training is performed locally on the edge cloud server, due to the limitation of computing resources, there may be problems such as insufficient training or local overfitting, resulting in deviations in the model parameters transmitted back to the central server, thereby affecting the accuracy of the global model. Within a specific time window, the central server can use part of the global data for periodic retraining to ensure that the model does not deviate from the global optimum, and can be repeatedly executed as follows Figure 1 The content shown involves the training and distribution of the content feedback model shown in steps 1 to 3.

[0117] Therefore, through the consistency and hierarchical update mechanism of the content feedback model in multiple edge cloud servers, regional aggregation can be performed between edge cloud servers to generate local models, and then uploaded to the central server step by step for global aggregation. This mechanism significantly reduces bandwidth usage and can achieve more accurate recommendations based on regional characteristics, improving the flexibility and adaptability of the recommendation system. In addition, by evaluating the quality of the content feedback model based on accuracy quality assessment data, completeness quality assessment data, and representative quality assessment data, the interference of low-quality data can be effectively reduced, and the accuracy and stability of the global content feedback model can be improved.

[0118] In some embodiments of the present application, the above-mentioned step 2202 may specifically include steps 22021 to 22022.

[0119] Step 22021, determining the content recommendation score of the first payment end for each type of reference recommended content based on the first feedback data and the interactive behavior data of each type of reference recommended content.

[0120] Step 22022, based on the content recommendation score of each category of reference recommended content, determine the target reference recommended content from the N categories of reference recommended content, the target reference recommended content having the largest content recommendation score.

[0121] Exemplarily, the above formula (4) may be used to filter the content recommendation score with the largest content recommendation score.

[0122] Step 22023: determine the reference content recommendation strategy associated with the target reference recommended content as the content recommendation strategy.

[0123] Specifically, step 22023 may include: based on the association relationship between the reference recommended content and the reference content recommendation strategy, determining the reference content recommendation strategy associated with the target reference recommended content as the content recommendation strategy.

[0124] Exemplarily, if there are k reference recommended contents with the largest content recommendation scores, the content recommendation strategy of the k reference recommended contents on the settlement page can be determined based on the content distribution of the settlement page of the order. The content recommendation strategy includes but is not limited to the display duration and mode of the recommended contents.

[0125] Therefore, the edge cloud server in the embodiment of the present application can combine the support of the content feedback model of the central cloud server to make personalized recommendations for recommended content based on the payment behavior data of the payment end, which can better meet the needs of the payment end. At the same time, it can also combine the global content feedback model of the central cloud server and the local UCB algorithm of the edge cloud server, so as to achieve distributed and local adaptive recommendations, thereby improving the accuracy of content recommendations.

[0126] Based on this, in some embodiments, the above step 22021 may specifically include:

[0127] Through the confidence interval upper bound algorithm, according to the first feedback data and the interactive behavior data of each type of reference recommended content, the content recommendation score of the first payment terminal for each type of reference recommended content is determined.

[0128] Specifically, in some embodiments of the present application, the interactive behavior data in the embodiments of the present application includes at least one of the following: first resource replacement data corresponding to the first feedback data, and a diversity score of each type of reference recommended content. Based on this, the above-mentioned step of determining the content recommendation score of each type of reference recommended content by the first payment terminal through the confidence interval upper bound algorithm according to the first feedback data and interactive behavior data of each type of reference recommended content may specifically include:

[0129] According to a first weight coefficient corresponding to the first feedback data and a second weight coefficient corresponding to the interactive behavior data, a weighted sum is performed on the first feedback data and the interactive behavior data to obtain a content recommendation score of the first payment end for each type of reference recommended content.

[0130] For example, taking the interactive behavior data including the first resource replacement data corresponding to the first feedback data and the diversity score of each type of reference recommended content as an example, the second weight coefficient includes w p and w d Based on this, the content recommendation score of each type of reference recommended content can be determined by the following formula (14):

[0131]

[0132] Among them, the first feedback data The first weight coefficient w y .

[0133] In some other embodiments of the present application, the interactive behavior data in the embodiments of the present application includes a recommendation value score of each type of reference recommended content and at least one of the following objects: first resource replacement data corresponding to the first feedback data, and a diversity score of each type of reference recommended content. Based on this, the above-mentioned step of determining the content recommendation score of each type of reference recommended content by the first payment terminal through the confidence interval upper bound algorithm according to the first feedback data and interactive behavior data of each type of reference recommended content may specifically include:

[0134] According to a first weight coefficient corresponding to the first feedback data and a third weight coefficient corresponding to the object, weighted summation is performed on the first feedback data and the object to obtain an initial content recommendation score;

[0135] The initial content recommendation score is adjusted by the recommendation value score to obtain the content recommendation score of the first payment end for each type of reference recommended content.

[0136] For example, the interactive behavior data includes the first resource replacement data corresponding to the first feedback data, the diversity score of each type of reference recommended content, and the recommendation value score of each type of reference recommended content. The second weight coefficient includes w p and w d Based on this, the content recommendation score of each type of reference recommended content can be determined by the following formula (15):

[0137]

[0138] Therefore, the content recommendation method provided in the embodiment of the present application can use the accuracy, completeness and representativeness of the newly added data to measure the quality of the edge model before each update, and use the quality as a weight for multiple model aggregation. In addition, multi-factor interaction behavior data is added to the UCB algorithm, and the expected feedback data, resource replacement data and diversity scores of the reference recommended content from the payment end are taken as the target, so that the recommended content can achieve a dynamic balance between resource replacement and diversity, thereby meeting the service requirements of the recommendation system.

[0139] Then, involving step 230, in some embodiments of the present application, data for displaying a settlement page may be sent to the first payment terminal so that the settlement page is displayed through the electronic device of the payment terminal, and the settlement page includes recommended content determined by the edge server, such as the k reference recommended content with the largest content recommendation score. Among them, the recommended content includes, but is not limited to, resource redemption vouchers, service release information and other content. Based on this, the electronic device can determine the feedback data to be sent to the edge cloud server based on the user's access to the recommended content in the settlement page, and put it on the edge cloud server, so that the edge cloud server can update the local content feedback model of the edge cloud server based on the feedback data, payment behavior data and recommended content displayed on the settlement page.

[0140] Therefore, the embodiment of the present application can combine the support of the content feedback model of the central cloud server, and make personalized recommendations for recommended content based on the payment behavior data of the payment end, which can better meet the needs of the payment end. At the same time, it can also combine the global content feedback model of the central cloud server and the local UCB algorithm of the edge cloud server, so as to achieve distributed and local adaptive recommendations, and improve the accuracy of content recommendations. In addition, multi-factor interactive behavior data is added to the UCB algorithm, and the feedback data, resource replacement data and diversity scores of the reference recommended content expected by the payment end for various reference recommended content are taken as targets, so that the recommended content can achieve a dynamic balance between resource replacement and diversity, thereby meeting the service requirements of the recommendation system, and combining the settlement page of the order to make personalized recommendations to the payment end, which can better meet the needs of the payment end and improve the accuracy of content recommendations.

[0141] The present application also provides a content recommendation device, specifically in combination with Figure 4 Provide detailed explanation.

[0142] Figure 4 It is a structural diagram of a content recommendation device provided by an embodiment of the present application.

[0143] In some embodiments of the present application, Figure 4 The content recommendation device shown can be set in the computer device provided in the embodiment of the present application.

[0144] like Figure 4 As shown, the content recommendation device 40 may specifically include:

[0145] An acquisition module 401 is used to acquire payment behavior data when the first payment terminal completes order payment, where the payment behavior data includes transaction data of the order and identity data of the first payment terminal;

[0146] A determination module 402, for determining a content recommendation strategy based on the payment behavior data;

[0147] The recommendation module 403 is used to recommend the recommended content corresponding to the recommendation strategy to the first payment terminal through the settlement page of the order.

[0148] In this way, the content recommendation device in the embodiment of the present application can make personalized recommendations to the payment end through the settlement page of the order based on the edge cloud server according to the order transaction data of the payment end and the identity data of the first payment end, which can better meet the needs of the payment end and improve the accuracy of content recommendations.

[0149] The content recommendation device 40 in the embodiment of the present application is described in detail below.

[0150] In some embodiments of the present application, the determination module 402 may be specifically configured to determine, based on the payment behavior data, first feedback data of the first payment terminal for each type of reference recommended content in N types of reference recommended content, where N is a positive integer;

[0151] A content recommendation strategy is determined based on the first feedback data and interactive behavior data of each type of reference recommended content; wherein the interactive behavior data includes at least one of the following: first resource replacement data corresponding to the first feedback data, a diversity score of each type of reference recommended content, and a recommendation value score of each type of reference recommended content.

[0152] In some embodiments of the present application, the determination module 402 may be specifically configured to determine, through a content feedback model and based on payment behavior data, first feedback data of the first payment terminal for each type of reference recommended content;

[0153] The content feedback model is trained by reference payment behavior data, each type of reference recommended content, and second feedback data of each type of reference recommended content.

[0154] In some embodiments of the present application, the acquisition module 401 may also be used to acquire feedback data of the first payment end on the settlement page, the feedback data including positive feedback data of the first payment end on the recommended content or negative feedback data of the first payment end on the recommended content;

[0155] The content recommendation device 40 in the embodiment of the present application may further include an updating module, which is used to update the content feedback model according to the feedback data, the payment behavior data and the recommended content displayed on the settlement page.

[0156] In some embodiments of the present application, the content recommendation device 40 in the embodiment of the present application may further include a receiving module for receiving content feedback model update data sent by M edge cloud servers in a cluster, wherein the content feedback model update data includes payment behavior data of the second payment terminal and feedback data of the second payment terminal for each type of reference recommended content, where M is a positive integer;

[0157] The content recommendation device 40 in the embodiment of the present application may further include an updating module, which is used to update data according to the content feedback model, update the content feedback model, and obtain an updated content feedback model.

[0158] The content recommendation device 40 in the embodiment of the present application may further include a sending module, which is used to send an updated content feedback model to each cluster of edge cloud servers.

[0159] In some embodiments of the present application, the determination module 402 may be specifically used to determine the model quality assessment value of the content feedback model according to the quality assessment data of the content feedback model update data, wherein the quality assessment data includes at least one of the following: accuracy quality assessment data, integrity quality assessment data, and representative quality assessment data;

[0160] The content recommendation device 40 in the embodiment of the present application may further include a sending module, which is used to send a new request to the central cloud server when the model quality evaluation value is less than or equal to a preset threshold, and the new request carries content feedback model update data. The new request is used to request the central cloud server to update the content feedback model in the central cloud server according to the content feedback model update data and the content feedback model update data sent by the P two-cluster edge cloud servers, and obtain an updated content feedback model, where P is a positive integer;

[0161] The sending module can also be used to, upon receiving the updated content feedback model sent by the central cloud server, synchronize the updated content feedback model to each cluster of edge cloud servers and replace the content feedback model with the updated content feedback model.

[0162] In some embodiments of the present application, the determination module 402 may be specifically configured to determine, based on the first feedback data and the interaction behavior data of each type of reference recommended content, a content recommendation score of the first payment terminal for each type of reference recommended content;

[0163] Based on the content recommendation score of each type of reference recommendation content, determine the target reference recommendation content from the N types of reference recommendation content, and the target reference recommendation content has the largest content recommendation score;

[0164] A reference content recommendation strategy associated with the target reference recommended content is determined as a content recommendation strategy.

[0165] In some embodiments of the present application, the determination module 402 can be specifically used to determine the content recommendation score of the first payment end for each type of reference recommended content based on the first feedback data and interaction behavior data of each type of reference recommended content through a confidence interval upper bound algorithm.

[0166] In some embodiments of the present application, the determination module 402 can be specifically used to, when the interactive behavior data includes at least one of the following: first resource replacement data corresponding to the first feedback data, and a diversity score for each type of reference recommended content, perform weighted summation of the first feedback data and the interactive behavior data according to a first weight coefficient corresponding to the first feedback data and a second weight coefficient corresponding to the interactive behavior data to obtain a content recommendation score for each type of reference recommended content by the first payment end.

[0167] In some embodiments of the present application, the determination module 402 may be specifically configured to, when the interaction behavior data includes a recommendation value score of each type of reference recommended content and at least one of the following objects: first resource replacement data corresponding to the first feedback data, and a diversity score of each type of reference recommended content, perform weighted summation of the first feedback data and the object according to a first weight coefficient corresponding to the first feedback data and a third weight coefficient corresponding to the object to obtain an initial content recommendation score;

[0168] The initial content recommendation score is adjusted by the recommendation value score to obtain the content recommendation score of the first payment end for each type of reference recommended content.

[0169] In some embodiments of the present application, the determination module 402 can be specifically used to determine the diversity score of each category of reference recommended content based on the number of categories of features in each category of reference recommended content, the number of times each category of reference recommended content is recommended, and the total number of times N categories of reference recommended content are recommended; wherein the diversity score is used to characterize the probability of each category of reference recommended content appearing in the distribution of recommended content interacted with the first payment end.

[0170] In some embodiments of the present application, the determination module 402 can be specifically used to determine the recommendation value score of each type of reference recommended content according to the number of times each type of reference recommended content is recommended and the second resource replacement data with each type of reference recommended content. The recommendation value score is used to characterize the probability that the feedback data of the first payment terminal for each type of reference recommended content is positive feedback data and the data amount of the resource replacement data corresponding to the feedback data of each type of reference recommended content is greater than or equal to the preset data amount.

[0171] Based on the same inventive concept, the present application also provides a computer device. Figure 5 Provide detailed explanation.

[0172] Figure 5 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application.

[0173] like Figure 5 As shown, the computer device may include at least one of the following involved in the embodiments of the present application: an edge cloud server, an electronic device at the payment end. The computer device may include a processor 501 and a memory 502 storing computer program instructions.

[0174] Specifically, the processor 501 may include a central processing unit (CPU), or an application specific integrated circuit (Application Specific Integrated Circuit (ASTC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0175] The memory 502 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 502 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 502 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 502 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 502 is a non-volatile solid-state memory. In a specific embodiment, the memory 502 includes a solid-state storage (ROM). In appropriate cases, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM) or a flash memory or a combination of two or more of these.

[0176] The processor 501 implements any one of the content recommendation methods in the above embodiments by reading and executing computer program instructions stored in the memory 502 .

[0177] In one example, the computer device may further include a communication interface 503 and a bus 510. Figure 5 As shown, the processor 501, the memory 502, and the communication interface 503 are connected via a bus 510 and communicate with each other.

[0178] The communication interface 503 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0179] Bus 510 includes hardware, software or both, and the parts of flow control device are coupled to each other.For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industry standard system (ETSA) bus, front side bus (FSB), hypertransport (HT) interconnection, industry standard system (TSA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel system (MCA) bus, peripheral component interconnection (PCT) bus, PCT-Express (PCT-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 510 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0180] The content recommendation device can execute the content recommendation method in the embodiment of the present application, thereby realizing the combination Figures 1 to 4 Described is a content recommendation method and apparatus.

[0181] In addition, in combination with the content recommendation method in the above embodiments, the present application embodiment can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any one of the content recommendation methods in the above embodiments is implemented.

[0182] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0183] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0184] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0185] The above are only specific implementation methods of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A content recommendation method, applied to an edge cloud server, comprising: When the first payment terminal completes the order payment, obtaining payment behavior data, the payment behavior data including the transaction data of the order and the identity data of the first payment terminal; Determining a content recommendation strategy based on the payment behavior data; Recommended content corresponding to the recommendation strategy is recommended to the first payment terminal through the settlement page of the order.

2. The method according to claim 1, wherein: Determining the content recommendation strategy according to the payment behavior data includes: determining first feedback data of the first payment terminal for each type of the reference recommended content in N types of reference recommended content according to the payment behavior data, where N is a positive integer; The content recommendation strategy is determined based on the first feedback data and interaction behavior data of each category of the reference recommended content; wherein the interaction behavior data includes at least one of the following: first resource replacement data corresponding to the first feedback data, a diversity score of each category of the reference recommended content, and a recommendation value score of each category of the reference recommended content.

3. The method according to claim 2, wherein: The step of determining, based on the payment behavior data, first feedback data for each type of reference recommended content in the N types of reference recommended content includes: Determining, by means of a content feedback model, first feedback data of the first payment terminal for each type of the reference recommended content according to the payment behavior data; The content feedback model is trained by reference payment behavior data, each type of reference recommended content and second feedback data of each type of reference recommended content.

4. The method according to claim 3, wherein: The method further comprises: Acquire feedback data of the first payment end on the settlement page, wherein the feedback data includes positive feedback data of the first payment end on the recommended content or negative feedback data of the first payment end on the recommended content; The content feedback model is updated according to the feedback data, the payment behavior data and the recommended content displayed on the settlement page.

5. The method according to claim 3, wherein: The method further comprises: Receiving content feedback model update data sent by M edge cloud servers in a cluster, wherein the content feedback model update data includes payment behavior data of the second payment terminal and feedback data of the second payment terminal for each type of reference recommended content, where M is a positive integer; According to the content feedback model update data, the content feedback model is updated to obtain an updated content feedback model; The updated content feedback model is sent to each of the cluster of edge cloud servers.

6. The method according to claim 5, wherein: Before sending the updated content feedback model to each edge cloud server, the method further includes: Determine a model quality assessment value of the content feedback model according to quality assessment data of the content feedback model update data, wherein the quality assessment data includes at least one of the following: accuracy quality assessment data, integrity quality assessment data, and representative quality assessment data; When the model quality evaluation value is less than or equal to a preset threshold, a new request is sent to the central cloud server, where the new request carries the content feedback model update data, and the new request is used to request the central cloud server to update the content feedback model in the central cloud server according to the content feedback model update data and the content feedback model update data sent by P two-cluster edge cloud servers, and obtain an updated content feedback model, where P is a positive integer; When receiving the updated content feedback model sent by the central cloud server, the updated content feedback model is synchronized to each of the cluster of edge cloud servers, and the content feedback model is replaced by the updated content feedback model.

7. The method according to claim 2, wherein: The determining the content recommendation strategy according to the first feedback data and the interaction behavior data of each type of the reference recommended content includes: Determining, according to the first feedback data and the interactive behavior data of each type of the reference recommended content, a content recommendation score of each type of the reference recommended content by the first payment terminal; Based on the content recommendation score of each category of the reference recommended content, determining a target reference recommended content from the N categories of reference recommended content, wherein the target reference recommended content has the largest content recommendation score; A reference content recommendation strategy associated with the target reference recommended content is determined as the content recommendation strategy.

8. The method according to claim 7, wherein: Determining, according to the first feedback data and the interaction behavior data of each type of the reference recommended content, a content recommendation score of each type of the reference recommended content by the first payment terminal includes: By using a confidence interval upper bound algorithm, according to the first feedback data and the interaction behavior data of each type of the reference recommended content, a content recommendation score of the first payment terminal for each type of the reference recommended content is determined.

9. The method according to claim 8, wherein: The interactive behavior data includes at least one of the following: first resource replacement data corresponding to the first feedback data, and a diversity score of each type of the reference recommended content; Determining, by the confidence interval upper bound algorithm, the content recommendation score of each type of the reference recommended content by the first payment terminal according to the first feedback data and the interaction behavior data of each type of the reference recommended content, includes: According to a first weight coefficient corresponding to the first feedback data and a second weight coefficient corresponding to the interactive behavior data, a weighted sum is performed on the first feedback data and the interactive behavior data to obtain a content recommendation score of the first payment terminal for each type of the reference recommended content.

10. The method according to claim 8, wherein: The interactive behavior data includes a recommendation value score of each type of the reference recommended content and at least one of the following objects: first resource replacement data corresponding to the first feedback data, and a diversity score of each type of the reference recommended content; Determining, by the confidence interval upper bound algorithm, the content recommendation score of each type of the reference recommended content by the first payment terminal according to the first feedback data and the interaction behavior data of each type of the reference recommended content, includes: performing a weighted summation of the first feedback data and the object according to a first weight coefficient corresponding to the first feedback data and a third weight coefficient corresponding to the object to obtain an initial content recommendation score; The initial content recommendation score is adjusted according to the recommendation value score to obtain a content recommendation score of each type of the reference recommended content by the first payment terminal.

11. The method according to claim 2, wherein: The diversity score is used to characterize the probability of each type of the reference recommended content appearing in the distribution of recommended content interacted with by the first payment terminal; The method further comprises: The diversity score of each category of the reference recommended content is determined according to the number of categories of features in each category of the reference recommended content, the number of times each category of the reference recommended content is recommended, and the total number of times the N categories of reference recommended content are recommended.

12. The method according to claim 2, wherein: The recommendation value score is used to represent the probability that the feedback data of the first payment terminal for each type of the reference recommended content is positive feedback data and the amount of resource replacement data corresponding to the feedback data of each type of the reference recommended content is greater than or equal to a preset amount of data; the method further includes: The recommendation value score of each type of the reference recommended content is determined according to the number of times each type of the reference recommended content is recommended and the second resource replacement data with each type of the reference recommended content.

13. A content recommendation device, comprising: An acquisition module, configured to acquire payment behavior data when the first payment terminal completes order payment, wherein the payment behavior data includes transaction data of the order and identity data of the first payment terminal; A determination module, used to determine a content recommendation strategy based on the payment behavior data; A recommendation module is used to recommend recommended content corresponding to the recommendation strategy to the first payment terminal through the settlement page of the order.

14. An electronic device, comprising: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the steps of the content recommendation method according to any one of claims 1 to 12 are implemented.

15. A storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement the steps of the content recommendation method according to any one of claims 1 to 12.

16. A computer program product, characterized in that The program product is stored in a storage medium, and the program product is executed by at least one processor to implement the steps of the content recommendation method according to any one of claims 1 to 12.

Images