Recommendation information display method, device, equipment and storage medium

Abstract

The present application discloses a method, apparatus, device and storage device for displaying recommendation information, the method comprising: obtaining at least two candidate recommendation groups; each candidate recommendation group includes at least two pieces of recommendation information, and the order of arrangement of the recommendation information in different candidate recommendation groups is different; inputting the candidate recommendation groups into a deep learning model, and outputting the first number of views of each recommendation information in the candidate recommendation group; the deep learning model is used to predict the number of views of each recommendation information after being affected by adjacent recommendation information; based on the first number of views of each recommendation information in the candidate recommendation group, determining the expected cost of the candidate recommendation group; selecting a target recommendation group with the largest expected cost from at least two candidate recommendation groups and displaying the target recommendation group. In this way, when predicting the number of views of each recommendation information, the present application combines the influence of adjacent recommendation information on the number of views to predict the number of views of each recommendation information, thereby improving the prediction accuracy of the number of views, so that the expected cost of the recommendation group finally selected is the largest and the accuracy is high.

Description

Technical Field

[0001] The present application relates to data processing technology, and in particular to a method, device, equipment and storage medium for displaying recommendation information. Background Art

[0002] On e-commerce platforms, advertising and non-advertising content are often displayed together. Both types of content can influence user behavior. Advertising helps advertisers achieve their marketing goals and generates advertising revenue for the platform. Advertising enhances the user experience, which is key to a platform's long-term success. Therefore, properly allocating advertising and non-advertising content to maximize revenue for e-commerce platforms is a top priority.

[0003] To address this issue, existing methods typically rank ads based on the expected revenue they can generate for e-commerce platforms when allocating them to advertising and non-advertising content. However, this approach ignores the interplay between advertising and non-advertising content, resulting in suboptimal combinations and, consequently, inaccurate calculations of expected e-commerce platform revenue. Summary of the Invention

[0004] To solve the above technical problems, the present application provides a method, device, equipment and storage medium for displaying recommendation information.

[0005] The technical solution of this application is achieved as follows:

[0006] In a first aspect, a method for displaying recommendation information is provided, the method comprising:

[0007] Obtain at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order;

[0008] Inputting the candidate recommendation group into a deep learning model and outputting the first page views corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each recommendation information after being affected by adjacent recommendation information;

[0009] determining an expected cost of the candidate recommendation group based on a first number of views corresponding to each recommendation information in the candidate recommendation group;

[0010] A target recommendation group with the largest expected cost is selected from the at least two candidate recommendation groups, and the target recommendation group is displayed.

[0011] In the above scheme, the recommendation information includes paid advertising information; the expected cost of the candidate recommendation group is determined based on the first number of views corresponding to each recommendation information in the candidate recommendation group, including: determining the first total cost of all advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and the preset real charging standard; determining the second total cost of all advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and the preset first virtual charging standard; wherein the virtual charging standard is set according to the self-valuation of the recommendation platform; the expected cost is obtained based on the first total cost and the second total cost.

[0012] In the above scheme, the recommendation information also includes unpaid non-advertising information; the expected cost of the candidate recommendation group is determined based on the first number of views corresponding to each recommendation information in the candidate recommendation group, and also includes: based on the first number of views of each non-advertising information in the candidate recommendation group and a preset second virtual charging standard, determining the third total cost of all non-advertising information in the candidate recommendation group; and obtaining the expected cost based on the first total cost, the second total cost and the third total cost.

[0013] In the above scheme, the method also includes: creating a first preset function; wherein, the first preset function is a function for calculating the average value of the number of views of all advertising information in the target recommendation group; creating a second preset function; wherein, the second preset function is a function for calculating the average value of the actual cost of all advertising information in the target recommendation group; obtaining N historical recommendation groups; wherein, each historical recommendation group includes at least two sub-candidate recommendation groups; when the expected charge is the first total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average value of the number of views of all advertising information in the N target recommendation groups according to the first preset function to obtain a first average value; when the expected charge is the second total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group Select the target recommendation group with the largest second total cost from the corresponding at least two sub-candidate recommendation groups; calculate the average of the actual costs of all advertising information in the N target recommendation groups according to the second preset function to obtain a second average value; create new first preset functions corresponding to the N historical recommendation groups based on the first preset function, and average the N new first preset functions to obtain a third preset function; create new second preset functions corresponding to the N historical recommendation groups based on the second preset function, and average the N new second preset functions to obtain a fourth preset function; the first average value and the second average value constitute a first point, and the third preset function and the fourth preset function constitute a second point; calculate the minimum Euclidean distance between the two points to obtain the first virtual charge.

[0014] In the above scheme, the method also includes: creating a fifth preset function; wherein the fifth preset function is a function for calculating the average number of clicks on all non-advertising information in the target recommendation group; when the expected charge is the third total fee, selecting the target recommendation group with the largest third total fee from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average number of views of all non-advertising information in N target recommendation groups according to the fifth preset function to obtain a third average value; creating new fifth preset functions corresponding to the N historical recommendation groups based on the fifth preset function, and averaging the N new fifth preset functions to obtain a sixth preset function; the first average value, the second average value and the third average value constitute a third point, and the third preset function, the fourth preset function and the sixth preset function constitute a third point; calculating the minimum Euclidean distance between two points to obtain the first virtual charge and the second virtual charge.

[0015] In the above scheme, the deep learning model includes a first prediction network and a second prediction network; inputting the candidate recommendation group into the deep learning model and outputting the first page views corresponding to each recommendation information in the candidate recommendation group includes: inputting each recommendation information in the candidate recommendation group into the first prediction network, individually predicting the page views of each recommendation information, and outputting the second page views of each recommendation information; inputting the second page views of each recommendation information into the second prediction network, calibrating the second page views based on the impact of adjacent recommendation information on the page views, and outputting the first page views of each recommendation information.

[0016] In the above scheme, the at least two candidate recommendation groups include: obtaining M recommendation information; inputting each recommendation information into a first prediction network, and outputting the second page views corresponding to each recommendation information; wherein the first prediction network is used to individually predict the page views of each recommendation information; sorting the second page views of each recommendation information in descending order, and selecting the top K candidate recommendation information therefrom; wherein K is less than or equal to M, and K and M are positive integers; and permuting and combining the K candidate recommendation information to obtain the at least two candidate recommendation groups.

[0017] In the above solution, the browsing volume includes at least one of the following: the number of clicks and browsing time of the recommended information.

[0018] In a second aspect, a device for displaying recommendation information is provided, the device comprising:

[0019] an acquiring unit, configured to acquire at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order;

[0020] a processing unit, configured to input the candidate recommendation group into a deep learning model and output a first pageview count corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is configured to predict the pageview count of each recommendation information after being affected by adjacent recommendation information;

[0021] a determining unit, configured to determine an expected cost of the candidate recommendation group based on a first number of views corresponding to each piece of recommendation information in the candidate recommendation group;

[0022] A selection unit is configured to select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

[0023] According to a third aspect, an electronic device is provided, comprising: a processor and a memory configured to store a computer program that can be run on the processor, wherein the processor is configured to execute the steps of the aforementioned method when running the computer program.

[0024] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program implements the steps of the aforementioned method when executed by a processor.

[0025] The present application provides a method for displaying recommendation information, the method comprising: obtaining at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the arrangement order of the recommendation information in different candidate recommendation groups is different; inputting the candidate recommendation groups into a deep learning model, and outputting the first page views corresponding to each piece of recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each piece of recommendation information after being affected by adjacent recommendation information; based on the first page views corresponding to each piece of recommendation information in the candidate recommendation group, determining the expected cost of the candidate recommendation group; selecting a target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and displaying the target recommendation group. In this way, when predicting the page views of each piece of recommendation information, the present application combines the influence of adjacent recommendation information on the page views to predict the page views of each piece of recommendation information, thereby improving the prediction accuracy of the page views, so that the expected cost of the recommendation group finally selected is the largest and the accuracy is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a schematic diagram of the first flow chart of the method for displaying recommendation information in an embodiment of the present application;

[0027] Figure 2 This is a display result diagram of the recommended information in the embodiment of this application;

[0028] Figure 3 This is a second flow chart of the method for displaying recommendation information in an embodiment of the present application;

[0029] Figure 4This is a schematic diagram of the third flow chart of the method for displaying recommendation information in an embodiment of the present application;

[0030] Figure 5 This is a schematic diagram of the structure of the recommended information display device in the embodiment of the present application;

[0031] Figure 6 This is a schematic diagram of the structure of the electronic device in the embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application is described in detail below with reference to the accompanying drawings. The attached drawings are for reference only and are not used to limit the embodiments of the present application.

[0033] Example 1

[0034] The present application embodiment provides a method for displaying recommendation information. Figure 1 This is a first flow chart of the method for displaying recommended information in an embodiment of the present application. Figure 1 As shown, the specific steps of the recommendation information display method may include:

[0035] Step 101: Obtain at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the order of arrangement of the recommendation information in different candidate recommendation groups is different;

[0036] It should be noted that the recommendation information display method of this application can be applied to both website platforms and e-commerce platforms. Here, the website platform will recommend related link information (i.e., recommended information) based on the content of the page currently browsed by the user, and the e-commerce platform will recommend related product information (i.e., recommended information) based on the product currently browsed by the user.

[0037] In practical applications, for e-commerce platforms, when a user browses a product, a group of products will be recommended based on the product. These products are then permuted and combined to form at least two candidate product groups (i.e., recommendation groups). Since the at least two candidate product groups are the result of permutations and combinations, different candidate product groups contain at least two identical pieces of product information, but in different order. Product information includes product code information, product type information, product brand information, product-related text, and product-related images.

[0038] In some embodiments, the at least two candidate recommendation groups include: obtaining M recommendation information; inputting each recommendation information into a first prediction network, and outputting the second page views corresponding to each recommendation information; wherein the first prediction network is used to individually predict the page views of each recommendation information; sorting the second page views of each recommendation information in descending order, and selecting the top K candidate recommendation information therefrom; wherein K is less than or equal to M, and K and M are positive integers; and permuting and combining the K candidate recommendation information to obtain the at least two candidate recommendation groups.

[0039] Here, in order to reduce the amount of calculation, K recommendation information is selected from M recommendation information, and the K recommendation information is arranged and combined to obtain all candidate recommendation groups. Specifically, the first prediction network is used to separately predict the second page views corresponding to each recommendation information, and each recommendation information is sorted according to the size of the second page views of each recommendation information. The top K candidate recommendation information is selected and arranged and combined to obtain at least two candidate recommendation groups (i.e. ). The first prediction network may be a pointwise network.

[0040] In practical applications, the recommended information mentioned in the above embodiments refers to recommended information that helps website platforms or e-commerce platforms achieve marketing goals and generate revenue, that is, paid recommended information. Typically, to improve user experience and maintain long-term public visibility, website platforms or e-commerce platforms will also independently select a batch of non-paid recommended information, which has a fixed position. The combination of the two constitutes at least two candidate recommendation groups, that is, each candidate recommendation group includes paid recommended information and unpaid recommended information.

[0041] Step 102: Input the candidate recommendation group into a deep learning model, and output the first page views corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each recommendation information after being affected by adjacent recommendation information;

[0042] It should be noted that the deep learning model is used to predict the number of views of each recommendation information after being affected by adjacent recommendation information. Combined with the fact that each candidate recommendation group mentioned above includes paid recommendation information and unpaid recommendation information, the adjacent recommendation information here can be paid recommendation information adjacent to unpaid recommendation information, or paid recommendation information adjacent to paid recommendation information, or unpaid recommendation information adjacent to unpaid recommendation information. Here, since this application fully considers the mutual influence between adjacent recommendation information, compared with the existing method that ignores the mutual influence between paid recommendation information and unpaid recommendation information, the recommendation group finally selected is the best, which improves the user experience.

[0043] It should be noted that the first page views refer to the amount of information generated during the browsing of recommended information. For example, if the recommended information is the recommended product information when browsing a product on an e-commerce platform, the corresponding first page views may include the number of clicks and browsing time to enter the product details page when browsing the product information.

[0044] In some embodiments, the deep learning model includes a first prediction network and a second prediction network; inputting the candidate recommendation group into the deep learning model and outputting the first number of views corresponding to each recommendation information in the candidate recommendation group includes: inputting each recommendation information in the candidate recommendation group into the first prediction network, individually predicting the number of views of each recommendation information, and outputting the second number of views of each recommendation information; inputting the second number of views of each recommendation information into the second prediction network, calibrating the second number of views based on the impact of adjacent recommendation information on the number of views, and outputting the first number of views of each recommendation information.

[0045] In other words, the first prediction network is used to predict the pageviews of each recommendation individually, obtaining a rough first pageview count. The second prediction network is then used to predict the pageviews of each recommendation again, taking into account the impact of adjacent recommendations on the current recommendation, to obtain a more accurate second pageview count. In other words, the first pageview count is more accurate than the second pageview count.

[0046] In practical applications, when using the second prediction network to predict the number of views of each recommended information again, the multi-head self-attention mechanism is used in the self-attention layer to learn the mutual influence between adjacent recommended information, and then the second number of views of each recommended information is output through the Dense layer.

[0047] For example, if the recommended information is a first ballpoint pen, and adjacent to it is a similar second ballpoint pen, compared with the case where the second ballpoint pen does not exist, the page views obtained by the first prediction network for predicting the first ballpoint pen are relatively low; compared with the case where the second ballpoint pen exists, the page views obtained by the second prediction network for predicting the first ballpoint pen again are relatively high; that is, the existence of the second ballpoint pen has an impact on the page views of the first ballpoint pen.

[0048] Step 103: determining the expected cost of the candidate recommendation group based on the first page views corresponding to each recommendation information in the candidate recommendation group;

[0049] In some embodiments, the recommendation information includes paid advertising information; this step specifically includes: determining the first total cost of all advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and a preset real charging standard; determining the second total cost of all advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and a preset first virtual charging standard; wherein the virtual charging standard is set according to the recommendation platform's own valuation; and obtaining the expected cost based on the first total cost and the second total cost.

[0050] In some embodiments, the recommendation information also includes unpaid non-advertising information; this step specifically also includes: determining the third total cost of all non-advertising information in the candidate recommendation group based on the first number of views of each non-advertising information in the candidate recommendation group and a preset second virtual charging standard; and obtaining the expected cost based on the first total cost, the second total cost and the third total cost.

[0051] It should be noted that the virtual charges mentioned in the above two embodiments are because the website platform or e-commerce platform, in addition to the actual advertising charges, also values the number of page views of advertisements and non-advertisements. The higher the page views, the more users use the platform, which increases the platform's own valuation, that is, the page views bring additional income to the platform (which can be understood as virtual income). For the virtual charging standards involved, a reasonable virtual charging standard can be calculated based on historical data. The specific calculation method is specifically described in the following embodiments.

[0052] Step 104: Select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

[0053] That is to say, the expected cost corresponding to each candidate recommendation group is calculated, and the candidate recommendation group with the largest expected cost is selected as the target recommendation group, and the recommended information in the target recommendation group is displayed on the website platform or e-commerce platform in the order of arrangement.

[0054] Based on the above embodiment, illustratively, the paid recommendation information includes advertising information, and the unpaid recommendation information includes non-advertising information. Figure 2 This is a display result diagram of the recommended information in the embodiment of this application. Figure 2As shown, 4 (K) advertising messages are selected from the 6 (i.e., M) advertising messages in 201 and permuted and combined to obtain 12 advertising groups in 202. These are then combined with the two independently selected non-advertising messages to obtain 12 recommendation groups (i.e., the at least two candidate recommendation groups mentioned in step 101 above). After steps 102, 103, and 104, the display result in 203 (i.e., the target recommendation group) is obtained, namely, Advertisement 1, Non-Advertisement 2, Advertisement 3, Non-Advertisement 4, Advertisement 5, and Advertisement 6. The expected cost corresponding to the display result in 203 is the highest.

[0055] Here, the execution entity of steps 101 to 104 may be a processor of the electronic device.

[0056] By adopting the above technical solution, when predicting the number of views of each recommended information, this application combines the impact of adjacent recommended information on the number of views to predict the number of views of each recommended information, thereby improving the prediction accuracy of the number of views, so that the expected cost of the recommended group finally selected is maximized and the accuracy is high.

[0057] Example 2

[0058] Based on the above embodiment, the embodiment of the present application takes the example of recommendation information including paid advertising information and unpaid non-advertising information, and provides a method for displaying recommendation information for calculating the recommendation group corresponding to the maximum expected cost of advertising information. Figure 3 This is a second flow chart of the method for displaying recommended information in an embodiment of the present application. Figure 3 As shown, the specific steps of the recommendation information display method may include:

[0059] Step 301: Obtain at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the order of arrangement of the recommendation information in different candidate recommendation groups is different;

[0060] It should be noted that the recommendation information display method of this application can be applied to both website platforms and e-commerce platforms. Here, the website platform will recommend related link information (i.e., recommended information) based on the content of the page currently browsed by the user, and the e-commerce platform will recommend related product information (i.e., recommended information) based on the product currently browsed by the user.

[0061] In practice, the recommended information provided by websites or e-commerce platforms includes both paid and unpaid recommendations. Paid recommendations help the platform achieve its marketing goals and generate revenue, while unpaid recommendations help the platform improve user experience and maintain long-term public visibility. In other words, recommended information includes both paid recommendations (including advertising information) and unpaid recommendations (including non-advertising information).

[0062] In some embodiments, the at least two candidate recommendation groups include: obtaining M recommendation information; inputting each recommendation information into a first prediction network, and outputting the second page views corresponding to each recommendation information; wherein the first prediction network is used to individually predict the page views of each recommendation information; sorting the second page views of each recommendation information in descending order, and selecting the top K candidate recommendation information therefrom; wherein K is less than or equal to M, and K and M are positive integers; and permuting and combining the K candidate recommendation information to obtain the at least two candidate recommendation groups.

[0063] Step 302: Input the candidate recommendation group into a deep learning model, and output the first page views corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each recommendation information after being affected by adjacent recommendation information;

[0064] In some embodiments, the deep learning model includes a first prediction network and a second prediction network; inputting the candidate recommendation group into the deep learning model and outputting the first number of views corresponding to each recommendation information in the candidate recommendation group includes: inputting each recommendation information in the candidate recommendation group into the first prediction network, individually predicting the number of views of each recommendation information, and outputting the second number of views of each recommendation information; inputting the second number of views of each recommendation information into the second prediction network, calibrating the second number of views based on the impact of adjacent recommendation information on the number of views, and outputting the first number of views of each recommendation information.

[0065] Step 303: determining a first total fee for all the advertising information in the candidate recommendation group based on the first page views of each advertisement information in the candidate recommendation group and a preset actual charging standard;

[0066] It should be noted that the number of views includes at least one of the following: the number of clicks on the recommended information and the browsing time.

[0067] Here, the preset actual charging standard may be the actual charging standard corresponding to each click on an advertisement, or the actual charging standard corresponding to a preset browsing time.

[0068] That is, the number of clicks on each advertisement information in the candidate recommendation group is multiplied by the actual charging standard corresponding to each click on the advertisement information and then added together to obtain the first total fee of all advertisement information in the candidate recommendation group.

[0069] For example, represents the click rate of advertisement information j in candidate recommendation group w, b j represents the actual charging standard for advertising information j, then the first total cost is

[0070] Step 304: Determine a second total fee for all advertisements in the candidate recommendation group based on the first number of views of each advertisement in the candidate recommendation group and a preset first virtual charging standard; wherein the virtual charging standard is set according to the recommendation platform's own valuation;

[0071] It should be noted that in addition to the actual charges for advertising information, website platforms or e-commerce platforms also value the number of views of advertising information. The higher the number of views, the more users use the platform, and the platform's own valuation increases, that is, the number of views brings additional income to the platform (which can be understood as virtual income).

[0072] Here, the preset virtual charging standard may be a virtual charging standard corresponding to each click on an advertisement, or a virtual charging standard corresponding to a preset browsing time.

[0073] That is, the number of clicks on each advertisement information in the candidate recommendation group is multiplied by the virtual charging standard corresponding to each click on the advertisement information and then added together to obtain the second total fee for all advertisement information in the candidate recommendation group.

[0074] For example, represents the click rate of advertisement information j in candidate recommendation group w, v a represents the virtual charging standard of advertisement information j (i.e., the first virtual charging standard), then the second total fee is

[0075] Step 305: Obtain the expected cost based on the first total cost and the second total cost;

[0076] Based on the above example, the expected cost for each candidate recommendation group is:

[0077]

[0078] Step 306: Select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

[0079] That is, the expected cost corresponding to each candidate recommendation group is calculated according to formula (1) in step 305, and the candidate recommendation group with the largest expected cost is selected as the target recommendation group, and the recommended information in the target recommendation group is displayed on the website platform or e-commerce platform in the order of arrangement.

[0080] Based on the above example, we actually want to explain how to balance the two goals of revenue brought to the platform by advertising information and click-through rate of advertising information based on the first virtual charge, so as to maximize the expected revenue brought to the platform.

[0081] Regarding the setting of the first virtual charging standard for each advertising information mentioned in step 304, in some embodiments, the method further includes: creating a first preset function; wherein the first preset function is a function for calculating the average value of the number of views of all advertising information in the target recommendation group; creating a second preset function; wherein the second preset function is a function for calculating the average value of the actual cost of all advertising information in the target recommendation group; obtaining N historical recommendation groups; wherein each historical recommendation group includes at least two sub-candidate recommendation groups; when the expected charge is the first total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average value of the number of views of all advertising information in the N target recommendation groups according to the first preset function to obtain a first average value; the expected charge is When the second total fee is calculated, the target recommendation group with the largest second total fee is selected from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; the average of the actual fees of all advertising information in the N target recommendation groups is calculated according to the second preset function to obtain a second average value; new first preset functions corresponding to the N historical recommendation groups are created based on the first preset function, and the N new first preset functions are averaged to obtain a third preset function; new second preset functions corresponding to the N historical recommendation groups are created based on the second preset function, and the N new second preset functions are averaged to obtain a fourth preset function; the first average value and the second average value constitute a first point, and the third preset function and the fourth preset function constitute a second point; the minimum Euclidean distance between the two points is calculated to obtain the first virtual charge.

[0082] For example, create a first virtual charging v a The first default function under the standard is:

[0083]

[0084] Created in the first virtual charging v a The second default function under the standard is:

[0085]

[0086] In formula (2) and formula (3), “*” represents the first virtual charge v a The corresponding target recommendation group under the standard.

[0087] Here, the first virtual charge v a=0, indicating that the expected charge is the first total charge, that is, no virtual charge for advertising information is made. The maximization problem of formula (1) actually becomes the maximization of the real cost of advertising. According to formula (1), the first total cost of each of the at least two sub-candidate recommendation groups corresponding to each historical recommendation group is calculated, and the target recommendation group with the largest first total cost is selected. Then, the second preset function of formula (3) is combined to calculate the average of the page views of all advertising information in the N target recommendation groups to obtain the first average value, that is:

[0088] First Virtual Charge v a =∞, that is, the first virtual charge is infinite, and the real cost of advertising information can be ignored. The maximization problem of formula (1) actually becomes the maximization of the virtual cost of advertising. According to formula (1), the second total cost of each of the at least two sub-candidate recommendation groups corresponding to each historical recommendation group is calculated, and the target recommendation group with the largest second total cost is selected. Then, the first preset function of formula (2) is combined to calculate the average of the page views of all advertising information in the N target recommendation groups to obtain the second average value, that is:

[0089] Here, the first average and the second average correspond to maximizing the real cost of advertising and maximizing the virtual cost of advertising, respectively. The first average and the second average constitute the first point (It can be called the ideal point).

[0090] Then, according to the first preset function of formula (2), new first preset functions corresponding to N historical recommendation groups are created respectively, and the N new first preset functions are averaged to obtain the third preset function:

[0091]

[0092] Then, according to the second preset function of formula (3), new second preset functions corresponding to N historical recommendation groups are created respectively, and the N new second preset functions are averaged to obtain the third preset function:

[0093]

[0094] Here, the third preset function and the fourth preset function constitute the second point In order to balance the two goals of revenue brought to the platform by advertising information and the click-through rate of advertising information, so as to maximize the expected revenue brought to the platform, the Euclidean distance between the first point and the second point is calculated, that is:

[0095]

[0096]

[0097] The minimum value of formula (6) is obtained, that is, the virtual fee corresponding to the minimum Euclidean distance is taken as the first virtual fee.

[0098] By adopting the above technical solution, when predicting the number of views of each recommended information, this application combines the impact of adjacent recommended information on the number of views to predict the number of views of each recommended information, thereby improving the prediction accuracy of the number of views, so that the expected cost of the recommended group finally selected is maximized and the accuracy is high.

[0099] Example 3

[0100] Based on the above embodiment, the present embodiment takes the example of recommendation information including paid advertising information and unpaid non-advertising information, and provides a method for displaying recommended information for calculating the recommended group corresponding to the maximum expected cost of advertising information and non-advertising information. Figure 4 This is a third flow chart of the method for displaying recommended information in an embodiment of the present application. Figure 4 As shown, the specific steps of the recommendation information display method may include:

[0101] Step 401: Obtain at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order;

[0102] It should be noted that the recommendation information display method of this application can be applied to both website platforms and e-commerce platforms. Here, the website platform will recommend related link information (i.e., recommended information) based on the content of the page currently browsed by the user, and the e-commerce platform will recommend related product information (i.e., recommended information) based on the product currently browsed by the user.

[0103] In practice, the recommended information provided by websites or e-commerce platforms includes both paid and unpaid recommendations. Paid recommendations help the platform achieve its marketing goals and generate revenue, while unpaid recommendations help the platform improve user experience and maintain long-term public visibility. In other words, recommended information includes both paid recommendations (including advertising information) and unpaid recommendations (including non-advertising information).

[0104] In some embodiments, the at least two candidate recommendation groups include: obtaining M recommendation information; inputting each recommendation information into a first prediction network, and outputting the second page views corresponding to each recommendation information; wherein the first prediction network is used to individually predict the page views of each recommendation information; sorting the second page views of each recommendation information in descending order, and selecting the top K candidate recommendation information therefrom; wherein K is less than or equal to M, and K and M are positive integers; and permuting and combining the K candidate recommendation information to obtain the at least two candidate recommendation groups.

[0105] Step 402: Input the candidate recommendation group into a deep learning model, and output the first page views corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each recommendation information after being affected by adjacent recommendation information;

[0106] In some embodiments, the deep learning model includes a first prediction network and a second prediction network; inputting the candidate recommendation group into the deep learning model and outputting the first number of views corresponding to each recommendation information in the candidate recommendation group includes: inputting each recommendation information in the candidate recommendation group into the first prediction network, individually predicting the number of views of each recommendation information, and outputting the second number of views of each recommendation information; inputting the second number of views of each recommendation information into the second prediction network, calibrating the second number of views based on the impact of adjacent recommendation information on the number of views, and outputting the first number of views of each recommendation information.

[0107] Step 403: determining a first total fee for all the advertising information in the candidate recommendation group based on the first page views of each advertising information in the candidate recommendation group and a preset actual charging standard;

[0108] It should be noted that the number of views includes at least one of the following: the number of clicks on the recommended information and the browsing time.

[0109] Here, the preset actual charging standard may be the actual charging standard corresponding to each click on an advertisement, or the actual charging standard corresponding to a preset browsing time.

[0110] That is, the number of clicks on each advertisement information in the candidate recommendation group is multiplied by the actual charging standard corresponding to each click on the advertisement information and then added together to obtain the first total fee of all advertisement information in the candidate recommendation group.

[0111] For example, represents the click rate of advertisement information j in candidate recommendation group w, b j represents the actual charging standard for advertising information j, then the first total cost is

[0112] Step 404: Determine a second total fee for all advertisements in the candidate recommendation group based on the first number of views of each advertisement in the candidate recommendation group and a preset first virtual charging standard; wherein the virtual charging standard is set according to the recommendation platform's own valuation;

[0113] It should be noted that in addition to the actual charges for advertising information, website platforms or e-commerce platforms also value the number of views of advertising information. The higher the number of views, the more users use the platform, and the platform's own valuation increases, that is, the number of views brings additional income to the platform (which can be understood as virtual income).

[0114] Here, the preset first virtual charging standard may be a virtual charging standard corresponding to each click on an advertisement, or may be a virtual charging standard corresponding to a preset browsing time.

[0115] That is, the number of clicks on each advertisement information in the candidate recommendation group is multiplied by the first virtual charging standard corresponding to each click on the advertisement information and then added together to obtain the second total fee for all advertisement information in the candidate recommendation group.

[0116] For example, represents the click rate of advertisement information j in candidate recommendation group w, v a represents the virtual charging standard of advertisement information j (i.e., the first virtual charging standard), then the second total fee is

[0117] Step 405: The recommendation information further includes unpaid non-advertisement information; the recommendation information further includes unpaid non-advertisement information; based on the first number of views of each non-advertisement information in the candidate recommendation group and a preset second virtual charging standard, determining a third total fee for all non-advertisement information in the candidate recommendation group;

[0118] It should be noted that in addition to the real charges and virtual fees for advertising information, website platforms or e-commerce platforms also value the number of views of non-advertising information. The higher the number of views, the more users use the platform, and the platform's own valuation increases, that is, the number of views brings additional income to the platform (which can be understood as virtual income).

[0119] Here, the preset second virtual charging standard may be a virtual charging standard corresponding to each click on non-advertisement information, or a virtual charging standard corresponding to a preset browsing time.

[0120] That is, the number of clicks on each non-advertisement information in the candidate recommendation group is multiplied by the second virtual charging standard corresponding to each click on a non-advertisement information and then added together to obtain the third total fee for all non-advertisement information in the candidate recommendation group.

[0121] For example, represents the click rate of non-advertising information j in candidate recommendation group w, v b represents the virtual charging standard of advertisement information j (i.e., the first virtual charging standard), then the second total fee is

[0122] Step 406: Obtain the expected cost based on the first total cost, the second total cost, and the third total cost;

[0123] Based on the above example, the expected cost for each candidate recommendation group is:

[0124]

[0125] Step 407: Select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

[0126] That is, the expected cost corresponding to each candidate recommendation group is calculated according to formula (7) in step 406, and the candidate recommendation group with the largest expected cost is selected as the target recommendation group, and the recommended information in the target recommendation group is displayed on the website platform or e-commerce platform in the order of arrangement.

[0127] Based on the above example, we actually want to explain how to balance the three goals of revenue brought to the platform by advertising information, click-through rate of advertising information, and click-through rate of non-advertising information based on the first virtual charge, so as to maximize the expected revenue brought to the platform.

[0128] Regarding the setting of the first virtual charging standard for each advertising information mentioned in step 404, and the setting of the second virtual charging standard for each advertising information mentioned in step 405, in some embodiments, the method further includes: creating a first preset function; wherein, the first preset function is a function for calculating the average value of the page views of all advertising information in the target recommendation group; creating a second preset function; wherein, the second preset function is a function for calculating the average value of the real costs of all advertising information in the target recommendation group; creating a fifth preset function; wherein, the fifth preset function is a function for calculating the average value of the click volume of all non-advertising information in the target recommendation group; obtaining N historical recommendation groups; wherein, each historical recommendation group includes at least two sub-candidate recommendation groups; when the expected charge is the first total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average value of the page views of all advertising information in the N target recommendation groups according to the first preset function to obtain a first average value; when the expected charge is the second total cost, selecting the target recommendation group with the largest second total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; according to the The second preset function calculates the average of the actual costs of all advertising information in the N target recommendation groups to obtain a second average value; when the expected charge is the third total cost, the target recommendation group with the largest third total cost is selected from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; the average of the browsing volume of all non-advertising information in the N target recommendation groups is calculated according to the fifth preset function to obtain a third average value; based on the first preset function, new first preset functions corresponding to the N historical recommendation groups are created respectively, and the N new first preset functions are averaged to obtain a third preset function; based on the second preset function, new second preset functions corresponding to the N historical recommendation groups are created respectively, and the N new second preset functions are averaged to obtain a fourth preset function; based on the fifth preset function, new fifth preset functions corresponding to the N historical recommendation groups are created respectively, and the N new fifth preset functions are averaged to obtain a sixth preset function; the first average value, the second average value and the third average value constitute a third point, and the third preset function, the fourth preset function and the sixth preset function constitute a third point; the minimum Euclidean distance between two points is calculated to obtain the first virtual charge and the second virtual charge.

[0129] For example, based on the example in the above embodiment, a second virtual charging v b The fifth preset function under the standard is:

[0130]

[0131] In formula (8), * represents the second virtual charge vb The corresponding target recommendation group under the standard.

[0132] Here, the second virtual charge v b =∞, that is, the second virtual charge is infinite, and the real cost of non-advertising information can be ignored. The maximization problem of formula (7) actually becomes maximizing the virtual cost of non-advertising. According to formula (7), the third total cost of each sub-candidate recommendation group in at least two sub-candidate recommendation groups corresponding to each historical recommendation group is calculated, and the target recommendation group with the largest third total cost is selected. Then, combined with the fifth preset function of formula (8), the average of the views of all non-advertising information in the N target recommendation groups is calculated to obtain the third average value, that is:

[0133] Here, the first average, the second average, and the third average correspond to maximizing the real cost of advertising, maximizing the virtual cost of advertising, and maximizing the virtual cost of non-advertising, respectively. The first average, the second average, and the third average constitute the third point (It can be called the ideal point).

[0134] Then, according to the fifth preset function of formula (8), new fifth preset functions corresponding to N historical recommendation groups are created respectively, and the N new fifth preset functions are averaged to obtain the sixth preset function:

[0135]

[0136] Here, the third preset function, the fourth preset function and the sixth preset function constitute the fourth point In order to balance the three goals of revenue brought to the platform by advertising information and the click-through rate of advertising information and the click-through rate of non-advertising information, so as to maximize the expected revenue brought to the platform, the Euclidean distance between the third and fourth points is calculated, that is:

[0137]

[0138]

[0139] Here, v represents (v a ,v b ) vector, find the minimum value of formula (10) and obtain the first virtual charge and the second virtual charge.

[0140] By adopting the above technical solution, when predicting the number of views of each recommended information, this application combines the impact of adjacent recommended information on the number of views to predict the number of views of each recommended information, thereby improving the prediction accuracy of the number of views, so that the expected cost of the recommended group finally selected is maximized and the accuracy is high.

[0141] Example 4

[0142] In order to implement the method of the embodiment of the present application, based on the same inventive concept, the embodiment of the present application further provides a recommendation information display device. Figure 5 This is a schematic diagram of the structure of the recommended information display device in the embodiment of the present application.

[0143] like Figure 5 As shown, the recommendation information display device includes:

[0144] An acquiring unit 501 is configured to acquire at least two candidate recommendation groups; each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order;

[0145] Processing unit 502 is configured to input the candidate recommendation group into a deep learning model and output a first pageview count corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is configured to predict the pageview count of each recommendation information after being affected by adjacent recommendation information;

[0146] A determining unit 503 is configured to determine an expected cost of the candidate recommendation group based on the first page views corresponding to each recommendation information in the candidate recommendation group;

[0147] The selection unit 504 is configured to select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

[0148] In some embodiments, the recommendation information includes paid advertising information; the device includes: a determination unit 503, specifically used to determine the first total cost of all advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and a preset real charging standard; based on the first number of views of each advertising information in the candidate recommendation group and a preset first virtual charging standard, determine the second total cost of all advertising information in the candidate recommendation group; wherein, the virtual charging standard is set according to the self-valuation of the recommendation platform; the expected cost is obtained based on the first total cost and the second total cost.

[0149] In some embodiments, the recommendation information also includes unpaid non-advertising information; the device includes: a determination unit 503, specifically used to determine the third total cost of all non-advertising information in the candidate recommendation group based on the first number of views of each non-advertising information in the candidate recommendation group and a preset second virtual charging standard; and obtain the expected cost based on the first total cost, the second total cost and the third total cost.

[0150] In some embodiments, the method further includes: creating a first preset function; wherein the first preset function is a function for calculating the average value of the number of views of all advertising information in the target recommendation group; creating a second preset function; wherein the second preset function is a function for calculating the average value of the actual cost of all advertising information in the target recommendation group; obtaining N historical recommendation groups; wherein each historical recommendation group includes at least two sub-candidate recommendation groups; when the expected charge is the first total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average value of the number of views of all advertising information in the N target recommendation groups according to the first preset function to obtain a first average value; when the expected charge is the second total cost, selecting the target recommendation group with the largest first total cost from at least two sub-candidate recommendation groups corresponding to each historical recommendation group A target recommendation group with the largest second total cost is selected from at least two sub-candidate recommendation groups corresponding to the group; the average of the actual costs of all advertising information in the N target recommendation groups is calculated according to the second preset function to obtain a second average value; new first preset functions corresponding to the N historical recommendation groups are created based on the first preset function, and the N new first preset functions are averaged to obtain a third preset function; new second preset functions corresponding to the N historical recommendation groups are created based on the second preset function, and the N new second preset functions are averaged to obtain a fourth preset function; the first average value and the second average value constitute a first point, and the third preset function and the fourth preset function constitute a second point; the minimum Euclidean distance between the two points is calculated to obtain the first virtual charge.

[0151] In some embodiments, the method also includes: creating a fifth preset function; wherein the fifth preset function is a function for calculating the average number of clicks on all non-advertising information in the target recommendation group; when the expected charge is the third total fee, selecting the target recommendation group with the largest third total fee from at least two sub-candidate recommendation groups corresponding to each historical recommendation group; calculating the average number of views of all non-advertising information in N target recommendation groups according to the fifth preset function to obtain a third average value; creating new fifth preset functions corresponding to the N historical recommendation groups based on the fifth preset function, and averaging the N new fifth preset functions to obtain a sixth preset function; the first average value, the second average value and the third average value constitute a third point, and the third preset function, the fourth preset function and the sixth preset function constitute a third point; calculating the minimum Euclidean distance between two points to obtain the first virtual charge and the second virtual charge.

[0152] In some embodiments, the deep learning model includes a first prediction network and a second prediction network; the device includes: a processing unit 502, specifically used to input each recommendation information in the candidate recommendation group into the first prediction network, separately predict the browsing volume of each recommendation information, and output the second browsing volume of each recommendation information; input the second browsing volume of each recommendation information into the second prediction network, calibrate the second browsing volume based on the impact of adjacent recommendation information on the browsing volume, and output the first browsing volume of each recommendation information.

[0153] In some embodiments, the page views include at least one of the following: the number of clicks and the page view duration of the recommended information.

[0154] By adopting the above technical solution, when predicting the number of views of each recommended information, this application combines the impact of adjacent recommended information on the number of views to predict the number of views of each recommended information, thereby improving the prediction accuracy of the number of views, so that the expected cost of the recommended group finally selected is maximized and the accuracy is high.

[0155] The present application also provides another electronic device. Figure 6 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application.

[0156] like Figure 6 As shown, the electronic device includes: a processor 601 and a memory 602 configured to store a computer program that can be run on the processor;

[0157] The processor 601 is configured to execute the method steps in the aforementioned embodiment when running the computer program.

[0158] Of course, in actual application, Figure 6 As shown, the various components in the electronic device are coupled together via a bus system 603. It is understood that the bus system 603 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 603 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, Figure 6 Various buses are labeled as bus system 603.

[0159] In practical applications, the processor may be at least one of an application-specific integrated circuit (ASIC), a digital signal processing device (DSPD), a programmable logic device (PLD), a field-programmable gate array (FPGA), a controller, a microcontroller, and a microprocessor. It is understood that for different devices, the electronic device used to implement the functions of the processor may also be other, and the embodiments of the present application do not specifically limit this.

[0160] The above-mentioned memory can be a volatile memory (volatile memory), such as a random-access memory (RAM); or a non-volatile memory (non-volatile memory), such as a read-only memory (ROM), a flash memory, a hard disk (HDD) or a solid-state drive (SSD); or a combination of the above types of memory, and provides instructions and data to the processor.

[0161] In an exemplary embodiment, the present application also provides a computer-readable storage medium for storing a computer program.

[0162] Optionally, the computer-readable storage medium can be applied to any one of the methods in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the processor in each method in the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0164] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0165] In addition, the functional units in the embodiments of the present invention can all be integrated into one processing module, or each unit can be a separate unit, or two or more units can be integrated into one unit; the above-mentioned integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units. It can be understood by those skilled in the art that all or part of the steps of the above-mentioned method embodiments can be completed by hardware related to program instructions, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiments; and the above-mentioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks or optical disks.

[0166] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0167] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0168] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0169] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for displaying recommended information, characterized in that: The method comprises: Obtain at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order; wherein each candidate recommendation group includes paid recommendation information and unpaid recommendation information; Inputting the candidate recommendation group into a deep learning model and outputting the first page views corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is used to predict the page views of each recommendation information after being affected by adjacent recommendation information; Based on the first number of views corresponding to each recommendation information in the candidate recommendation group, the expected fee of the candidate recommendation group is determined; when the recommendation information includes paid advertising information, the expected fee includes a first total fee and a second total fee; when the recommendation information also includes unpaid non-advertising information, the expected fee includes the first total fee, the second total fee and the third total fee; wherein, the first total fee is determined based on the first number of views of each advertising information in the candidate recommendation group and a preset real charging standard; the second total fee is determined based on the first number of views of each advertising information in the candidate recommendation group and a preset first virtual charging standard; the third total fee is determined based on the first number of views of each non-advertising information in the candidate recommendation group and a preset second virtual charging standard; A target recommendation group with the largest expected cost is selected from the at least two candidate recommendation groups, and the target recommendation group is displayed.

2. The method according to claim 1, characterized in that The recommended information includes paid advertising information; The determining the expected cost of the candidate recommendation group based on the first page views corresponding to each recommendation information in the candidate recommendation group includes: determining the first total fee of all the advertising information in the candidate recommendation group based on the first page views of each piece of advertising information in the candidate recommendation group and the preset actual charging standard; Determining the second total fee for all the advertising information in the candidate recommendation group based on the first number of views of each advertising information in the candidate recommendation group and the preset first virtual charging standard; wherein the virtual charging standard is set according to the recommendation platform's own valuation; The expected cost is obtained according to the first total cost and the second total cost.

3. The method according to claim 2, characterized in that The recommended information also includes unpaid non-advertising information; The determining of the expected cost of the candidate recommendation group based on the first page views corresponding to each recommendation information in the candidate recommendation group further includes: determining the third total fee for all non-advertisement information in the candidate recommendation group based on the first page views of each non-advertisement information in the candidate recommendation group and the preset second virtual charging standard; The expected cost is obtained according to the first total cost, the second total cost and the third total cost.

4. The method according to claim 3, characterized in that The method further comprises: Creating a first preset function; wherein the first preset function is a function for calculating the average of the number of views of all advertisement information in the target recommendation group; Creating a second preset function; wherein the second preset function is a function for calculating the average value of the actual costs of all the advertising information in the target recommendation group; Obtain N historical recommendation groups; wherein each historical recommendation group includes at least two sub-candidate recommendation groups; When the expected cost is the first total cost, selecting a target recommendation group with the largest first total cost from at least two candidate sub-recommendation groups corresponding to each historical recommendation group; Calculating the average of the number of views of all advertisement information in the N target recommendation groups according to the first preset function to obtain a first average value; When the expected cost is the second total cost, selecting a target recommendation group with the largest second total cost from at least two candidate sub-recommendation groups corresponding to each historical recommendation group; Calculating the average of the actual costs of all the advertising information in the N target recommendation groups according to the second preset function to obtain a second average value; Creating new first preset functions corresponding to the N historical recommendation groups based on the first preset function, and averaging the N new first preset functions to obtain a third preset function; creating new second preset functions corresponding to the N historical recommendation groups based on the second preset function, and averaging the N new second preset functions to obtain a fourth preset function; The first average value and the second average value constitute a first point, and the third preset function and the fourth preset function constitute a second point; Calculate the minimum Euclidean distance between the two points to obtain the first virtual charge.

5. The method according to claim 4, characterized in that The method further comprises: Creating a fifth preset function; wherein the fifth preset function is a function for calculating the average of click counts of all non-advertisement information in the target recommendation group; When the expected cost is the third total cost, selecting the target recommendation group with the largest third total cost from the at least two candidate sub-recommendation groups corresponding to each historical recommendation group; Calculating the average of the views of all non-advertisement information in the N target recommendation groups according to the fifth preset function to obtain a third average value; creating new fifth preset functions corresponding to the N historical recommendation groups based on the fifth preset function, and averaging the N new fifth preset functions to obtain a sixth preset function; The first average value, the second average value, and the third average value constitute a third point, and the third preset function, the fourth preset function, and the sixth preset function constitute a third point; The minimum Euclidean distance between the two points is calculated to obtain the first virtual charge and the second virtual charge.

6. The method according to claim 1, characterized in that The deep learning model includes a first prediction network and a second prediction network; Inputting the candidate recommendation group into the deep learning model and outputting the first page views corresponding to each recommendation information in the candidate recommendation group includes: Input each recommendation information in the candidate recommendation group into the first prediction network, perform a separate prediction on the page views of each recommendation information, and output a second page view count of each recommendation information; The second page views of each recommendation information are input into the second prediction network, the second page views are calibrated in combination with the influence of adjacent recommendation information on the page views, and the first page views of each recommendation information are output.

7. The method according to claim 1, characterized in that The at least two candidate recommendation groups include: Get M recommended information; Input each recommendation information into a first prediction network, and output a second pageview count corresponding to each recommendation information; wherein the first prediction network is used to individually predict the pageview count of each recommendation information; Sort the second views of each recommended information in descending order, and select the top K candidate recommended information; where K is less than or equal to M, and K and M are positive integers; The K candidate recommendation information are arranged and combined to obtain the at least two candidate recommendation groups.

8. The method according to any one of claims 1 to 7, characterized in that The page views include at least one of the following: the number of clicks on the recommended information and the page view duration.

9. A recommendation information display device, characterized in that: The device comprises: an acquisition unit, configured to acquire at least two candidate recommendation groups; wherein each candidate recommendation group includes at least two pieces of recommendation information, and the recommendation information in different candidate recommendation groups is arranged in a different order; wherein each candidate recommendation group includes paid recommendation information and unpaid recommendation information; a processing unit, configured to input the candidate recommendation group into a deep learning model and output a first pageview count corresponding to each recommendation information in the candidate recommendation group; wherein the deep learning model is configured to predict the pageview count of each recommendation information after being affected by adjacent recommendation information; A determination unit is configured to determine an expected fee for the candidate recommendation group based on the first number of views corresponding to each recommendation information in the candidate recommendation group; when the recommendation information includes paid advertising information, the expected fee includes a first total fee and a second total fee; when the recommendation information also includes unpaid non-advertising information, the expected fee includes the first total fee, the second total fee and a third total fee; wherein the first total fee is determined based on the first number of views of each advertising information in the candidate recommendation group and a preset real charging standard; the second total fee is determined based on the first number of views of each advertising information in the candidate recommendation group and a preset first virtual charging standard; and the third total fee is determined based on the first number of views of each non-advertising information in the candidate recommendation group and a preset second virtual charging standard; A selection unit is configured to select the target recommendation group with the largest expected cost from the at least two candidate recommendation groups, and display the target recommendation group.

10. An electronic device, characterized in that: The electronic device comprises: a processor and a memory configured to store a computer program capable of running on the processor, Wherein, the processor is configured to execute the steps of the method according to any one of claims 1 to 8 when running the computer program.

11. 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 8 are implemented.

Images