Information recommendation method and device, equipment and storage medium

Abstract

The present disclosure relates to an information recommendation method and device, equipment and storage medium, and belongs to the technical field of computers. The method comprises: in response to an item recommendation instruction for a first account, obtaining a plurality of candidate items; determining the on-shelf rate and the acquisition rate of the plurality of candidate items; determining the recommendation probability of the plurality of candidate items based on the on-shelf rate and the acquisition rate of the plurality of candidate items; and recommending the plurality of candidate items to the first account based on the recommendation probability of the plurality of candidate items. Since the on-shelf rate indicates the probability of shelving a candidate item on an item supply platform of the first account in the case of recommending the candidate item to the first account, and the acquisition rate indicates the probability of detecting an acquisition operation on the shelved candidate item in the case of shelving the candidate item on the item supply platform, therefore, recommending the candidate item based on the on-shelf rate and the acquisition rate can improve the probability of shelving the candidate item on the item supply platform and being acquired by other users, and thus the accuracy of item recommendation can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to an information recommendation method, apparatus, device, and storage medium. Background Technology

[0002] With the development of computer technology, live-streaming shopping has rapidly become a mainstream shopping method. In this technology, servers recommend multiple items to merchants based on their preferences. Merchants then select items and list them on a supply platform, where users can then access these items. However, because merchants' preferences differ from users', the recommended items may not be what users like, resulting in low accuracy in item recommendations. Summary of the Invention

[0003] This disclosure provides an information recommendation method, apparatus, device, and storage medium, which can improve the accuracy of item recommendations. The technical solution of this disclosure is as follows: According to a first aspect of the present disclosure, an information recommendation method is provided, the method comprising: In response to the item recommendation command for the primary account, retrieve multiple alternative items; The listing rate and acquisition rate of the plurality of candidate items are determined. The listing rate indicates the probability that the candidate item is listed on the item supply platform of the first account when the candidate item is recommended to the first account. The acquisition rate indicates the probability that an acquisition operation of the listed candidate item is detected when the candidate item is listed on the item supply platform. Based on the listing rate and acquisition rate of the multiple candidate items, the recommendation probability of the multiple candidate items is determined, and the recommendation probability is positively correlated with the listing rate and the acquisition rate. Based on the recommendation probabilities of the multiple candidate items, the multiple candidate items are recommended to the first account.

[0004] In some embodiments, determining the acquisition rate of the plurality of candidate items includes: Obtain information about a first account, information about a second account, and information about the items of the multiple candidate items, wherein the second account is an account that follows the first account; For any candidate item, the acquisition rate prediction model is invoked to determine the acquisition rate of the candidate item based on the item information of the candidate item, the first account information, and the second account information.

[0005] In some embodiments, the training process of the acquisition rate prediction model includes: Obtain first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item by an account other than the first sample account is detected, or no acquisition operation of the sample item is detected. The acquisition rate prediction model is invoked to determine the first predicted label of the sample item based on the first sample account information, the second sample account information, and the sample item information; Based on the first sample label and the first predicted label, the acquisition rate prediction model is trained to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0006] In some embodiments, obtaining the first sample label includes: If an operation to acquire the sample item is detected within a first time period after the listing time of the sample item, the first sample tag corresponding to the sample item is determined as the first value, and the listing time is the time when the sample item is listed on the item supply platform of the first sample account. If no acquisition operation of the sample item is detected within the first time period after the listing time, the first sample label corresponding to the sample item is determined as the second value.

[0007] In some embodiments, determining the recommendation probability of the plurality of candidate items based on the listing rate and the acquisition rate of the plurality of candidate items includes: For any candidate item, the listing rate of the candidate item is multiplied by the acquisition rate to obtain the recommendation probability of the candidate item.

[0008] In some embodiments, determining the recommendation probability of the plurality of candidate items based on the listing rate and the acquisition rate of the plurality of candidate items includes: Based on the listing rate and acquisition rate of the multiple candidate items, a first item sequence and a second item sequence are determined. The first item sequence is an item sequence obtained by sorting the multiple candidate items in descending order of listing rate, and the second item sequence is an item sequence obtained by sorting the multiple candidate items in descending order of acquisition rate. For any candidate item, a first recommendation probability of the candidate item is determined based on its order in the first item sequence, and a second recommendation probability of the candidate item is determined based on its order in the second item sequence. The first recommendation probability and the second recommendation probability are weighted to obtain the recommendation probability of the candidate item.

[0009] In some embodiments, determining the availability rate of the plurality of candidate items includes: Obtain the first account information and the item information of the multiple candidate items; For any candidate item, the listing rate prediction model is invoked, and the listing rate of the candidate item is determined based on the item information of the candidate item and the first account information.

[0010] In some embodiments, the training process of the shelf availability prediction model includes: Obtain second sample information, which includes third sample account information, sample item information, and second sample tag. The second sample tag indicates whether the sample item is listed on the item supply platform of the third sample account after the sample item is recommended to the third sample account, or whether the sample item is not listed on the item supply platform. The product listing rate prediction model is invoked, and a second predicted tag for the sample item is determined based on the third sample account information and the sample item information; Based on the second sample label and the second predicted label, the shelf availability prediction model is trained to reduce the difference between the second predicted label obtained by the trained shelf availability prediction model and the second sample label.

[0011] In some embodiments, obtaining the second sample label includes: After recommending the sample item to the third sample account, and then listing the sample item on the item supply platform of the third sample account, the second sample tag corresponding to the sample item is determined as the third value. If the sample item is not listed on the item supply platform of the third sample account after recommending the sample item to the third sample account, the second sample tag corresponding to the sample item is determined as the fourth value.

[0012] In some embodiments, the step of obtaining multiple candidate items in response to an item recommendation instruction for a first account includes: In response to the item recommendation instruction, a second account is identified, which is an account that follows the first account; For any item category among multiple item categories, determine the number of second accounts that have acquired items within that item category from the identified second accounts; From the plurality of item categories, a first number of target item categories are determined, wherein the number of second accounts corresponding to the target item categories is the largest. Select the multiple candidate items from the target item category.

[0013] In some embodiments, before determining the listing rate and acquisition rate of the plurality of candidate items, the information recommendation method further includes: For any candidate item, obtain the number of third accounts corresponding to the candidate item, wherein the server detects the acquisition operation of the candidate item in the item supply platform of the third account within a second time period before the current time point; The multiple candidate items are sorted in descending order of the number of third accounts corresponding to the multiple candidate items; Get the second number of the candidate items that are ranked first.

[0014] According to a second aspect of the present disclosure, a model training method is provided, the method comprising: Obtain first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item is detected, or no acquisition operation of the sample item is detected. The acquisition rate prediction model is invoked, and based on the first sample account information, the second sample account information, and the sample item information, the first predicted label of the sample item is determined; Based on the first sample label and the first predicted label, the acquisition rate prediction model is trained to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0015] According to a third aspect of the present disclosure, an information recommendation apparatus is provided, the apparatus comprising: The item acquisition unit is configured to execute an item recommendation command in response to the first account and acquire multiple candidate items; The first determining unit is configured to determine the listing rate and acquisition rate of the plurality of candidate items, wherein the listing rate indicates the probability that the candidate items are listed on the first account's item supply platform when the candidate items are recommended to the first account, and the acquisition rate indicates the probability that acquisition operations of other accounts besides the first account are detected when the candidate items are listed on the item supply platform. The second determining unit is configured to perform a determination of the recommendation probability of the plurality of candidate items based on the listing rate and the acquisition rate of the plurality of candidate items, wherein the recommendation probability is positively correlated with the listing rate and the acquisition rate; The item recommendation unit is configured to recommend the multiple candidate items to the first account based on the recommendation probability of the multiple candidate items.

[0016] In some embodiments, the first determining unit is configured to perform the following: obtaining first account information, second account information, and item information of the plurality of candidate items, wherein the second account is an account that follows the first account; for any candidate item, invoking an acquisition rate prediction model to determine the acquisition rate of the candidate item based on the item information of the candidate item, the first account information, and the second account information.

[0017] In some embodiments, the training process of the acquisition rate prediction model includes: Obtain first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item by an account other than the first sample account is detected, or no acquisition operation of the sample item is detected. The acquisition rate prediction model is invoked to determine the first predicted label of the sample item based on the first sample account information, the second sample account information, and the sample item information; Based on the first sample label and the first predicted label, the acquisition rate prediction model is trained to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0018] In some embodiments, obtaining the first sample label includes: If an operation to acquire the sample item is detected within a first time period after the listing time of the sample item, the first sample tag corresponding to the sample item is determined as the first value, and the listing time is the time when the sample item is listed on the item supply platform of the first sample account. If no acquisition operation of the sample item is detected within the first time period after the listing time, the first sample label corresponding to the sample item is determined as the second value.

[0019] In some embodiments, the second determining unit is configured to perform the following operation: for any candidate item, multiply the listing rate of the candidate item by the acquisition rate to obtain the recommendation probability of the candidate item.

[0020] In some embodiments, the second determining unit is configured to perform the following operations: determining a first item sequence and a second item sequence based on the listing rate and the acquisition rate of the plurality of candidate items; the first item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of listing rate; the second item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of acquisition rate; for any candidate item, determining a first recommendation probability of the candidate item based on its order in the first item sequence, and determining a second recommendation probability of the candidate item based on its order in the second item sequence; and weighting the first recommendation probability and the second recommendation probability to obtain the recommendation probability of the candidate item.

[0021] In some embodiments, the first determining unit is configured to perform the following actions: obtaining first account information and item information of the plurality of candidate items; for any candidate item, invoking an availability prediction model to determine the availability rate of the candidate item based on the item information of the candidate item and the first account information.

[0022] In some embodiments, the training process of the shelf availability prediction model includes: Obtain second sample information, which includes third sample account information, sample item information, and second sample tag. The second sample tag indicates whether the sample item is listed on the item supply platform of the third sample account after the sample item is recommended to the third sample account, or whether the sample item is not listed on the item supply platform. The product listing rate prediction model is invoked, and a second predicted tag for the sample item is determined based on the third sample account information and the sample item information; Based on the second sample label and the second predicted label, the shelf availability prediction model is trained to reduce the difference between the second predicted label obtained by the trained shelf availability prediction model and the second sample label.

[0023] In some embodiments, obtaining the second sample label includes: After recommending the sample item to the third sample account, and then listing the sample item on the item supply platform of the third sample account, the second sample tag corresponding to the sample item is determined as the third value. If the sample item is not listed on the item supply platform of the third sample account after recommending the sample item to the third sample account, the second sample tag corresponding to the sample item is determined as the fourth value.

[0024] In some embodiments, the item acquisition unit is configured to execute an action in response to the item recommendation instruction, determining a second account, wherein the second account is an account that follows the first account; for any item category among a plurality of item categories, determining the number of second accounts that have acquired items within the determined item category from the determined second accounts; determining a first number of target item categories from the plurality of item categories, wherein the number of second accounts corresponding to the target item categories is the largest; and selecting the plurality of candidate items from the target item categories.

[0025] In some embodiments, the item acquisition unit is further configured to perform the following: for any candidate item, acquire the number of third accounts corresponding to the candidate item, wherein the server detects an acquisition operation of the candidate item in the item supply platform of the third account within a second time period before the current time point; sort the multiple candidate items in descending order of the number of third accounts corresponding to the multiple candidate items; and acquire the second number of candidate items that are ranked first.

[0026] According to a fourth aspect of the present disclosure, a model training apparatus is provided, the apparatus comprising: The information acquisition unit is configured to acquire first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation on the listed sample item is detected, or no acquisition operation on the sample item is detected. The model invocation unit is configured to execute the invocation acquisition rate prediction model and determine the first prediction label of the sample item based on the first sample account information, the second sample account information and the sample item information; The model training unit is configured to train the acquisition rate prediction model based on the first sample label and the first predicted label, so as to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0027] According to a fifth aspect of the present disclosure, an electronic device is provided, the electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the information recommendation method or model training method as described above.

[0028] According to a sixth aspect provided in this disclosure, a computer-readable storage medium is provided that, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the information recommendation method or model training method as described above.

[0029] According to a seventh aspect of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the information recommendation method or model training method described above.

[0030] In this embodiment of the disclosure, since the listing rate of alternative items indicates the probability that alternative items will be listed on the first account's item supply platform when alternative items are recommended to the first account, and the acquisition rate of alternative items indicates the probability that an acquisition operation of the listed alternative items will be detected when alternative items are listed on the first account's item supply platform, the recommendation probability of alternative items is determined based on the listing rate and acquisition rate of alternative items. Recommending alternative items to the first account based on this recommendation probability can increase the probability that alternative items will be listed on the item supply platform and acquired by other users. Therefore, the accuracy of item recommendations can be improved.

[0031] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0032] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0033] Figure 1 This is a schematic diagram illustrating an implementation environment according to an exemplary embodiment.

[0034] Figure 2 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment.

[0035] Figure 3 This is a flowchart illustrating another information recommendation method according to an exemplary embodiment.

[0036] Figure 4 This is a flowchart illustrating a model training method according to an exemplary embodiment.

[0037] Figure 5 This is a flowchart illustrating another model training method according to an exemplary embodiment.

[0038] Figure 6 This is a flowchart illustrating yet another information recommendation method according to an exemplary embodiment.

[0039] Figure 7 This is a flowchart illustrating another information recommendation method according to an exemplary embodiment.

[0040] Figure 8 This is a schematic diagram illustrating an item recommendation process according to an exemplary embodiment.

[0041] Figure 9 This is a block diagram illustrating an information recommendation device according to an exemplary embodiment.

[0042] Figure 10 This is a block diagram illustrating a model training apparatus according to an exemplary embodiment.

[0043] Figure 11 This is a block diagram illustrating a terminal according to an exemplary embodiment.

[0044] Figure 12 This is a block diagram illustrating a server according to an exemplary embodiment. Detailed Implementation

[0045] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0046] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0047] It should be noted that the user information involved in this disclosure (including but not limited to user device information, user personal information, etc.) may be information authorized by the user or fully authorized by all parties.

[0048] As used in this disclosure, the terms "at least one", "multiple", "each", and "any" mean that there is one, two, or more than one; multiple means two or more; each refers to each of the corresponding multiples; and any means any one of the multiples. For example, multiple accounts include three accounts, where each means each of the three accounts, and any means any one of the three accounts, which could be the first, the second, or the third.

[0049] Figure 1 This is a schematic diagram of an implementation environment provided by an embodiment of this disclosure. See also... Figure 1 The implementation environment includes a terminal 101 and a server 102. The terminal 101 and the server 102 are connected via a wireless or wired network. For example, the terminal 101 may be a laptop, mobile phone, tablet, or other terminal. For example, the server 102 may be an application backend server or a cloud server providing services such as cloud computing and cloud storage.

[0050] Terminal 101 is used to log in to an account and send an item recommendation request to server 102. The account is registered on server 102, and the item recommendation request requests item recommendations for that account. Server 102 receives the item recommendation request, generates an item recommendation instruction for the account, determines at least one alternative item as a recommendation for the first account using the information recommendation method provided in this disclosure, and sends the at least one alternative item to terminal 101. Terminal 101 receives and displays the at least one alternative item. In this way, the user corresponding to the terminal can select an item from the at least one alternative item and list the selected item on the account's item supply platform. Other users can then access the items on the item supply platform. For example, after selecting an alternative item from the displayed at least one alternative item, the user sends an item listing request to server 102 through terminal 101. The item listing request carries the user-selected alternative item. Server 102 adds the user-selected alternative item to the first account's item listing set, and the items in the item listing set are displayed on the first account's item supply platform.

[0051] For example, terminal 101 has an application installed. Terminal 101 logs into its account in the application, sends an item recommendation request to server 102 through the application, receives at least one alternative item from server 102, and displays at least one alternative item. Terminal 101 also uses the application to list user-selected items on an item supply platform. For example, the application can be an application in the operating system of terminal 101 or a third-party application. For example, the application can be a live streaming application, a shopping application, a game application, etc., and this disclosure embodiment does not limit this.

[0052] The method provided in this disclosure can be applied to scenarios where food recommendations are made to users. For example, a user sends a food recommendation request to a server via a terminal. The request includes the terminal's currently logged-in first account. After receiving the request, the server identifies multiple candidate foods and determines their recommendation probabilities using the method provided in this disclosure. Based on these probabilities, the server selects the candidate foods with the highest recommendation probabilities and sends them to the terminal. The terminal receives and displays these candidate foods. The user can then choose a food from these options and upload it to the account's item supply platform. Other users can then access the food from this platform.

[0053] The method provided in this disclosure can be applied to scenarios where cosmetics are recommended to users. For example, a user sends a cosmetic recommendation request to a server via a terminal. The request includes the terminal's currently logged-in first account. Upon receiving the request, the server identifies multiple candidate cosmetics and determines their recommendation probabilities using the method provided in this disclosure. Based on these probabilities, the server selects the candidate cosmetics with higher recommendation probabilities and sends them to the terminal. The terminal receives and displays these candidate cosmetics. The user can then choose a cosmetic from these options and list it on the account's product supply platform. Other users can then access the cosmetics on this platform.

[0054] The method provided in this disclosure can be applied to scenarios involving recommending electronic products to users. For example, a user sends an electronic product recommendation request to a server via a terminal. The request includes the terminal's currently logged-in first account. Upon receiving the request, the server identifies multiple candidate electronic products and determines their recommendation probabilities using the method provided in this disclosure. Based on these probabilities, the server selects the candidate electronic products with higher recommendation probabilities and sends them to the terminal. The terminal receives and displays these candidate electronic products. The user can then select products from these options and list them on the account's product supply platform. Other users can then access these products on the product supply platform.

[0055] The information recommendation method provided in this disclosure can also be applied to scenarios where other types of items are recommended to users, such as sporting goods, school supplies, etc. This disclosure does not limit this application.

[0056] Figure 2 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment, such as... Figure 2 As shown, the method includes the following steps.

[0057] 201. In response to the item recommendation instruction for the first account, the server retrieves multiple alternative items.

[0058] The item recommendation instruction for the first account specifies that the first account should recommend items. For example, after receiving an item recommendation request from the terminal carrying the first account information, the server generates an item recommendation instruction for the first account. For example, the terminal sends the item recommendation request to the interface corresponding to the item recommendation service, and the server receives the request through that interface. The first account can be any account registered on the server, such as an account registered by a merchant.

[0059] For example, the types of candidate items can be any, such as cosmetics, food, electronic products, office supplies, etc. For example, the server stores a set of candidate items, which is used to store multiple candidate items. Accordingly, the server retrieves multiple candidate items from the set of candidate items.

[0060] For example, the item recommendation request also carries item recommendation criteria. Accordingly, the server selects multiple candidate items from the candidate item set that meet the item recommendation criteria. For example, the item recommendation criteria include the resource range in which the resource quantity corresponding to the item falls, the acquisition quantity range that the item acquisition quantity needs to meet, the type of the item, and the method of selecting candidate items from the candidate item set, etc.

[0061] 202. The server determines the availability and acquisition rate of multiple candidate items.

[0062] The listing rate indicates the probability that, when recommending alternative items to the first account, those items will be listed on the first account's item supply platform. A higher listing rate indicates a greater probability that the user of the first account likes the alternative item. After the server recommends alternative items to the first account, the user of the first account can decide whether to list the alternative item on their own item supply platform. Therefore, by determining the listing rate of alternative items when recommending items to the first account, and recommending items based on this listing rate, the server can increase the probability that the recommended item will be listed on the user's item supply platform, thereby improving the accuracy of item recommendations.

[0063] The acquisition rate indicates the probability that, when a candidate item is listed on the item supply platform, accounts other than the primary account are detected acquiring that candidate item. A higher acquisition rate indicates a greater probability that other users like the candidate item. After the user belonging to the primary account lists an item on the item supply platform, other users can decide whether to acquire the item from that platform. Therefore, when recommending items to the primary account, the server determines the acquisition rate of the candidate item. Recommending items based on this acquisition rate to the primary account increases the probability that the recommended item will be acquired by other users after being listed on the primary account's item supply platform, thereby improving the accuracy of item recommendations.

[0064] The goods supply platform can display items. After an item is listed on the goods supply platform, users can see the items listed on the platform and then purchase them. For example, the goods supply platform includes live streaming rooms, online stores, etc.

[0065] 203. The server determines the recommendation probability of multiple candidate items based on their listing rate and acquisition rate.

[0066] The recommendation probability is positively correlated with the listing rate and the acquisition rate. For example, the server multiplies the listing rate and acquisition rate of candidate items to obtain the recommendation probability of the candidate item. A higher recommendation probability indicates a greater probability that the candidate item will be successfully listed on the first account's item supply platform and successfully acquired by other users; that is, a greater probability that the candidate item is liked by users. Of course, the recommendation probability of candidate items can also be determined in other ways, such as by weighted summing of the listing rate and acquisition rate of candidate items. This embodiment of the disclosure does not limit this approach.

[0067] 204. The server recommends multiple alternative items to the first account based on the recommendation probability of multiple alternative items.

[0068] For example, the server sorts multiple candidate items in descending order of recommendation probability and recommends the top-ranked candidate items (up to a target number) to the first account. The target number is arbitrary, such as 20. Alternatively, the server recommends candidate items with a recommendation probability greater than a probability threshold to the first account. The probability threshold is arbitrary, such as 0.8. This method ensures that candidate items recommended to the first account have a high recommendation probability. A higher recommendation probability indicates a greater probability that the candidate item will be successfully listed on the first account's product supply platform and successfully acquired by other users; in other words, a greater probability that the candidate item will be liked by users, thus improving the accuracy of item recommendations.

[0069] It's important to note that if only the listing rate is used to recommend candidate items for the first account, the acquisition volume of these items after they are listed on the supply platform by the user of the first account is often low. This results in a lower GMV (Gross Merchandise Volume) for the first account. However, in this embodiment, recommending candidate items for the first account by combining both the listing rate and the acquisition rate not only increases the probability that the recommended candidate items will be listed on the supply platform by the user of the first account, but also increases the probability that these candidate items will be acquired by other users. This, in turn, increases the GMV for the first account and consequently, the GMV for the live streaming application.

[0070] Another point to note is that after recommending items for an account using the information recommendation method provided in this embodiment, the account's item supply platform's item exposure count on the item recommendation day, the number of times items in the item supply platform were obtained on the item recommendation day, the number of accounts that obtained items in the item supply platform on the item recommendation day, and the item supply platform's GMV on the item recommendation day all experienced a leap in growth.

[0071] Figure 3 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment. In this embodiment, the acquisition rate and listing rate of candidate items are determined using an acquisition rate prediction model and a listing rate prediction model. Figure 3 As shown, the method includes the following steps.

[0072] 301. In response to the item recommendation instruction for the first account, the server retrieves multiple alternative items.

[0073] The implementation method for this step is the same as step 201 above, and will not be repeated here.

[0074] 302. The server calls the acquisition rate prediction model to determine the acquisition rate of multiple candidate items.

[0075] In some embodiments, this step is implemented as follows: the server obtains first account information, second account information, and item information of multiple candidate items; for any candidate item, it calls the acquisition rate prediction model to determine the acquisition rate of the candidate item based on the item information of the candidate item, the first account information, and the second account information. The second account is an account that follows the first account.

[0076] For example, the first account information includes the attribute information of the user to whom the first account belongs, information about the items listed by the first account before the current time, and information about items listed by the first account that have been acquired by other users. The second account information includes the attribute information of the user to whom the second account belongs, and information about the items acquired by the second account. The user's attribute information includes the user's age, gender, and region. The item information includes the type of item and the quantity of resources corresponding to the item.

[0077] For example, the server inputs first account information, second account information, and item information into the acquisition rate prediction model. The acquisition rate prediction model makes a prediction based on the input first account information, second account information, and item information, obtaining a predicted label for the item, and then outputs the predicted label. For example, this predicted label is a value between 0 and 1, representing the acquisition rate of the candidate item predicted by the acquisition rate prediction model. The training process of the acquisition rate prediction model is described below. Figure 4 The embodiments shown are not described in detail here.

[0078] In this embodiment of the disclosure, since the acquisition rate prediction model has high accuracy and high prediction efficiency in predicting the acquisition rate, determining the acquisition rate of candidate items through the acquisition rate prediction model can improve prediction efficiency and ensure the accuracy of the acquisition rate of candidate items. Using the acquisition rate for item recommendation can improve the efficiency and accuracy of item recommendation.

[0079] 303. The server calls the listing rate prediction model to determine the listing rate of multiple candidate items.

[0080] In some embodiments, the server obtains first account information and item information of multiple candidate items; for any candidate item, the server calls the listing rate prediction model to determine the listing rate of the candidate item based on the item information of the candidate item and the first account information.

[0081] For example, the first account information includes the attribute information of the user to whom the first account belongs, and information about the items that the first account has listed before the current time. The user's attribute information includes the user's age, gender, and region. The item information includes the item type and the quantity of resources corresponding to the item.

[0082] For example, the server inputs the first account information and item information into the listing rate prediction model. Based on the input first account information and item information, the listing rate prediction model makes predictions to obtain predicted labels for candidate items, and then outputs these predicted labels. For example, the predicted label is a value between 0 and 1, representing the listing rate of the candidate items predicted by the listing rate prediction model. The training process of the listing rate prediction model is described below. Figure 5 The embodiments shown are not described in detail here.

[0083] In this embodiment of the disclosure, since the listing rate prediction model has high accuracy and high prediction efficiency in predicting the listing rate, determining the listing rate of candidate items through the listing rate prediction model can improve prediction efficiency and ensure the accuracy of the listing rate of candidate items. Using the listing rate for item recommendation can improve the efficiency and accuracy of item recommendation.

[0084] It should be noted that the present invention does not limit the order of steps 302 and 303; that is, step 303 can be executed first, followed by step 302.

[0085] 304. The server determines the recommendation probability of multiple candidate items based on their listing rate and acquisition rate.

[0086] Among them, the recommendation probability is positively correlated with the listing rate and the acquisition rate.

[0087] In some embodiments, the server determines the recommendation probability of multiple candidate items based on their listing rate and acquisition rate. This includes: for any candidate item, the server multiplies the listing rate by the acquisition rate to obtain the recommendation probability. The recommendation probability indicates the probability that the candidate item will be successfully listed on the first account's item supply platform and successfully acquired by other users. A higher recommendation probability indicates a greater probability of the candidate item being successfully listed on the first account's item supply platform and successfully acquired by other users, meaning a greater probability that the candidate item will be liked by users. Recommending items to users based on this recommendation probability increases the probability that the recommended items will be liked by users, thereby improving recommendation accuracy.

[0088] In some embodiments, the server determines the recommendation probability of multiple candidate items based on the listing rate and acquisition rate of multiple candidate items, including the following steps (1)-(3).

[0089] (1) The server determines the first item sequence and the second item sequence based on the listing rate and acquisition rate of multiple candidate items. The first item sequence is the item sequence obtained by sorting multiple candidate items in descending order of listing rate, and the second item sequence is the item sequence obtained by sorting multiple candidate items in descending order of acquisition rate.

[0090] (2) For any candidate item, the server determines the first recommendation probability of the candidate item based on its order in the first item sequence, and determines the second recommendation probability of the candidate item based on its order in the second item sequence.

[0091] The earlier an item appears in the first item sequence, the higher its probability of being the first recommended item. Conversely, the later an item appears in the first item sequence, the lower its probability of being the first recommended item. Similarly, the earlier an item appears in the second item sequence, the higher its probability of being the second recommended item. Conversely, the later an item appears in the second item sequence, the lower its probability of being the second recommended item.

[0092] For example, the server determines the ratio of the order of the candidate item in the first item sequence to the number of candidate items in the first item sequence, and determines the difference between 1 and the ratio as the first recommendation probability.

[0093] For example, the server determines the ratio of the order of the candidate item in the second item sequence to the number of candidate items in the second item sequence, and determines the difference between 1 and the ratio as the second recommendation probability.

[0094] (3) The server performs weighted processing on the first recommendation probability and the second recommendation probability to obtain the recommendation probability of the candidate item.

[0095] For example, the server performs a weighted sum of the first recommendation probability and the second recommendation probability to obtain the recommendation probability of the candidate item. The weights of the first recommendation probability and the second recommendation probability can be set to any value, for example, the weight of the first recommendation probability is 0.6 and the weight of the second recommendation probability is 0.4. This embodiment of the disclosure does not limit this.

[0096] In this embodiment, the first recommendation probability and the second recommendation probability are determined by the order of the candidate items in the first item sequence and the order of the candidate items in the second item sequence. Then, the first recommendation probability and the second recommendation probability are weighted to obtain the final recommendation probability of the candidate items, instead of directly determining the recommendation probability of the candidate items by the absolute values ​​of the acquisition rate and the listing rate of the candidate items. This makes the difference between the determined recommendation probabilities of each candidate item smaller, that is, the distribution of the recommendation probabilities of multiple candidate items is not too scattered, which can improve the efficiency of subsequent data processing of recommendation probabilities, thereby improving the efficiency of item recommendation.

[0097] 305. The server recommends multiple alternative items to the first account based on the recommendation probability of multiple alternative items.

[0098] The implementation method for this step is the same as step 204 above, and will not be repeated here.

[0099] In this embodiment of the disclosure, the acquisition rate and listing rate of candidate items are determined by the acquisition rate prediction model and the listing rate prediction model, which ensures the acquisition efficiency and accuracy of the acquisition rate and listing rate of candidate items, thereby ensuring the accuracy of the recommendation probability of candidate items and the recommendation efficiency of candidate items.

[0100] Figure 4 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment. This embodiment describes the training process of an acquisition rate prediction model used to determine the acquisition rate of candidate items. Figure 4 As shown, the method includes the following steps.

[0101] 401. The server obtains the first sample information, which includes the first sample account information, the second sample account information, the sample item information, and the first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item by an account other than the first sample account is detected, or no acquisition operation of the sample item is detected.

[0102] For example, the server randomly selects multiple accounts from the registered accounts as first sample accounts. Then, other accounts that follow the first sample accounts are designated as second sample accounts. Items previously listed by the first sample accounts on the supply platform are identified as sample items. For example, the first sample account information includes the attribute information of the user to which the first sample account belongs, information about items listed by the first sample account before the current time, and information about items listed by the first sample account that have been acquired by other users. The second sample account information includes the attribute information of the user to which the second sample account belongs, and information about items acquired by the second sample account. The user's attribute information includes the user's age, gender, and region. The item information includes the item type and the quantity of resources corresponding to the item.

[0103] In some embodiments, the server obtains the first sample tag by: if the server detects a retrieval operation on the sample item within a first time period after the sample item's listing time, determining the first sample tag corresponding to the sample item as a first value; and if the server does not detect a retrieval operation on the sample item within the first time period after the listing time, determining the first sample tag corresponding to the sample item as a second value. The listing time is the time when the sample item is listed on the item supply platform of the first sample account. The first value indicates that the sample item was retrieved by other users within the first time period after listing. The second value indicates that the sample item was not retrieved by other users within a second time period after listing. For example, the first value is 1, and the second value is 0. For example, the first time period can be any duration, such as 3 days. For example, if the first sample tag is the first value, the sample corresponding to the first sample tag is called a positive sample. If the first sample tag is the second value, the sample corresponding to the first sample tag is called a negative sample.

[0104] In this embodiment of the disclosure, if a retrieval operation is detected within the first time period after the sample item is listed, the first sample tag corresponding to the sample item is determined as a first value. If no retrieval operation is detected within the first time period after the listing, the first sample tag corresponding to the sample item is determined as a second value. Then, a retrieval rate prediction model is trained based on the first sample tag, so that the predicted tag obtained by the retrieval rate prediction model can reflect the retrieval rate of the item, that is, the probability that the item is retrieved by other users within the first time period after listing. This enables the recommendation of items based on the retrieval rate, thereby improving the accuracy of item recommendation.

[0105] 402. The server call acquisition rate prediction model determines the first predicted label of the sample item based on the first sample account information, the second sample account information, and the sample item information.

[0106] For example, the server inputs the first sample account information, the second sample account information, and the sample item information into the acquisition rate prediction model. The acquisition rate prediction model makes a prediction based on the input first sample account information, second sample account information, and sample item information to obtain the first predicted label of the sample item, and then outputs the first predicted label.

[0107] For example, the first prediction label is a value between 0 and 1, representing the acquisition rate of the sample item predicted by the acquisition rate prediction model. For instance, a first prediction label of 0.8 means that after the sample item is listed on the item supply platform of the first sample account, there is an 80% probability that it will be acquired by other users within the first time period.

[0108] 403. The server trains the acquisition rate prediction model based on the first sample label and the first predicted label.

[0109] The training objective of the acquisition rate prediction model is to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model. Accordingly, the server trains the acquisition rate prediction model based on the first sample label and the first predicted label to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0110] For example, the server determines the loss value of the acquisition rate prediction model based on the first sample label and the first predicted label. The loss value indicates the difference between the first predicted label and the first sample label. The server adjusts the model parameters based on the loss value so that the loss value obtained by the adjusted acquisition rate prediction model is reduced until the loss value is less than the loss threshold.

[0111] In this embodiment, considering that the first sample account information, the second sample account information, and the sample item information are all important factors that can influence the successful acquisition of sample items by other users in the item supply platform of the first sample account, an acquisition rate prediction model is trained based on these three types of information. This allows the model to learn the relationship between these three types of information and the first sample tag, i.e., the relationship between these three types of information and whether the sample items in the item supply platform of the first sample account are successfully acquired by other users. Therefore, based on the information of any account, the information of other accounts following that account, and the information of any item, it can accurately determine the probability that an item in the item supply platform of an account is acquired by other users. Subsequently, based on this probability, the recommendation probability of the item is determined, and items are recommended based on this recommendation probability. This increases the probability that the recommended items will be liked by users, thus improving the accuracy of item recommendations.

[0112] For example, the acquisition rate prediction model can also be trained using other information besides the three types mentioned above, and this disclosure does not limit this. For example, the acquisition rate prediction model can also be trained by other electronic devices besides the server, and this disclosure does not limit this either.

[0113] Figure 5 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment. This embodiment describes the training process of a shelf-rate prediction model used to determine the shelf-rate of candidate items. Figure 5 As shown, the method includes the following steps.

[0114] 501. The server obtains the second sample information, which includes the third sample account information, sample item information, and the second sample tag. The second sample tag indicates whether the sample item is listed on the third sample account's item supply platform after recommending the sample item to the third sample account, or whether the sample item is not listed on the item supply platform.

[0115] For example, the server randomly selects multiple accounts from the already registered accounts as third-sample accounts, and then identifies the items recommended to these third-sample accounts as sample items. For example, the third-sample account information includes the attribute information of the user to whom the third-sample account belongs, and information about the items listed by the third-sample account before the current time. The user's attribute information includes the user's age, gender, and region. The item information includes the item type and the quantity of resources corresponding to the item.

[0116] In some embodiments, the server obtains the second sample tag by: if the server recommends a sample item to a third sample account and then lists the sample item on the third sample account's item supply platform, the second sample tag corresponding to the sample item is determined to be a third value; if the server recommends a sample item to a third sample account but does not list the sample item on the third sample account's item supply platform, the second sample tag corresponding to the sample item is determined to be a fourth value. The third value indicates that after recommending the sample item to the third sample account, the user of the third sample account listed the sample item on the third sample account's item supply platform. The fourth value indicates that after recommending the sample item to the third sample account, the user of the third sample account did not list the sample item on the third sample account's item supply platform. For example, the third value is 1, and the fourth value is 0. For example, if the second sample tag is the third value, the sample corresponding to the second sample tag is called a positive sample. If the second sample tag is the fourth value, the sample corresponding to the first sample tag is called a negative sample.

[0117] For example, considering that after recommending a sample item to a third sample account, the user of that third sample account needs some time to decide whether to list the sample item on the product supply platform, to avoid the delay in the user's listing behavior causing the second sample tag corresponding to the sample item to be incorrectly marked, the server delays the marking time of the second sample tag corresponding to the sample item. That is, within the target time period after recommending the sample item to the third sample account, the server does not mark the second sample tag corresponding to the sample item. Instead, after the target time interval from the time point when recommending the sample item to the third sample account, if the sample item is listed on the third sample account's product supply platform, the server determines the second sample tag corresponding to the sample item as the third value; if the sample item is not listed on the third sample account's product supply platform, the server determines the second sample tag corresponding to the sample item as the fourth value. Here, the target time period can be any length, such as 30 minutes, 1 hour, etc.

[0118] In this embodiment of the disclosure, after recommending a sample item to a third sample account, if the sample item is listed on the third sample account's item supply platform, the second sample tag corresponding to the sample item is determined as a third value. If the sample item is not listed on the third sample account's item supply platform, the second sample tag corresponding to the sample item is determined as a fourth value. Then, a listing rate prediction model is trained based on the second sample tag, so that the predicted tag obtained by the listing rate prediction model can reflect the listing rate of the item, that is, the probability that the item will be listed on the item supply platform by the user after recommending the item to the user. This enables the recommendation of items based on the listing rate, thereby improving the accuracy of item recommendation.

[0119] 502. The server calls the listing rate prediction model and determines the second prediction label of the sample item based on the third sample account information and sample item information.

[0120] For example, the server inputs the third sample account information and sample item information into the listing rate prediction model. The listing rate prediction model makes a prediction based on the input third sample account information and sample item information to obtain the second predicted label of the sample item, and then outputs the second predicted label.

[0121] For example, the second prediction label is a value between 0 and 1, representing the listing rate prediction of the sample item by the listing rate prediction model. For instance, a second prediction label of 0.8 means that after recommending the sample item to the third sample account, the third sample account has an 80% probability of choosing to list the sample item on the item supply platform.

[0122] 503. The server trains the shelf availability prediction model based on the second sample label and the second predicted label.

[0123] The training objective of the shelf availability prediction model is to reduce the difference between the second predicted label and the second sample label obtained by the trained model. Accordingly, the server trains the shelf availability prediction model based on the second sample label and the second predicted label to reduce the difference between the second predicted label and the second sample label obtained by the trained model.

[0124] For example, the server determines the loss value of the shelf-rate prediction model based on the second sample label and the second predicted label. The loss value indicates the difference between the second predicted label and the second sample label. The server adjusts the model parameters based on the loss value so that the loss value obtained by the adjusted shelf-rate prediction model is reduced until the loss value is less than the loss threshold.

[0125] In this embodiment, considering that both the third sample account information and the sample item information are important factors influencing whether the third sample account will list the sample item on the item supply platform after recommending it, a listing rate prediction model is trained based on these two types of information. This allows the model to learn the relationship between these two types of information and the second sample tag—that is, the relationship between these two types of information and whether the third sample account chooses to list the sample item on the item supply platform after recommending it. This enables the model to accurately determine the probability that an account will list an item on the item supply platform after recommending it, based on the information of any account and any item. Subsequently, the recommendation probability of an item is determined based on this probability, and recommending items based on this probability increases the probability that the recommended items will be liked by users, thus improving the accuracy of item recommendations.

[0126] For example, the product availability prediction model can also be trained using other information besides the two types of information mentioned above, such as information about other accounts that follow the third sample account. This embodiment of the disclosure does not limit this. For example, the product availability prediction model can also be trained by other electronic devices other than the server. This embodiment of the disclosure does not limit this either.

[0127] Figure 6 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment, which describes a method for selecting candidate items. Figure 6 As shown, the method includes the following steps.

[0128] 601. In response to the item recommendation command, the server identifies the second account, which is the account that follows the first account.

[0129] 602. For any item category among multiple item categories, the server determines the number of second accounts that have acquired items within that item category from the identified second accounts.

[0130] For example, these multiple item categories are all third-level item categories. Item categories include multi-level item categories, and lower-level item categories are subcategories of higher-level item categories. For example, the item categories include: sports-related items - outdoor sports-related items - water sports-related items. Among them, water sports-related items are third-level item categories. In this embodiment of the disclosure, selecting candidate items from multiple third-level item categories can yield a more diverse range of candidate items.

[0131] 603. The server determines the first number of target item categories from multiple item categories, and the second account corresponding to the target item category has the largest number.

[0132] For example, the server sorts multiple item categories in descending order according to the number of corresponding second accounts, and determines the first number of item categories in the sorted list as the target item category. The first number can be any number, such as 3, 4, etc., and this embodiment of the disclosure does not impose any limitation on it.

[0133] 604. The server selects multiple candidate items from the target item category.

[0134] In this embodiment, by utilizing items acquired by users of the second account that follow the first account, the target item categories most favored by these users are determined. Candidate items are then selected from these target item categories, increasing the likelihood that the candidate items recommended to the first account will be liked by these users, thereby improving the accuracy of item recommendations. Furthermore, it eliminates the need to utilize historical item information from the first account's item supply platform, allowing for accurate item recommendations even if the first account has never listed any items before. This avoids the "cold start" problem caused by the lack of historical item information when recommending items to the first account.

[0135] For example, the server randomly selects a target number of items from each target item category. This target number can be any number, such as 5 or 6. For example, the server selects multiple candidate items from the target item category based on information such as the acquisition volume of the item, the allocated resources corresponding to the item, and the number of third-party accounts corresponding to the item. Here, the acquisition volume of an item refers to the total amount of data acquired for that item across all item supply platforms. The allocated resources corresponding to an item refer to the amount of resources allocated by the server to the account belonging to that item supply platform after other users acquire the item from it. Furthermore, for any given item, if the server detects an acquisition operation of that item on any item supply platform within the second time period prior to the current time point, then the account belonging to that item supply platform is considered the third-party account corresponding to that item.

[0136] For example, the server selects multiple candidate items from the target item category based on information such as the quantity of items acquired, the amount of resources allocated to the items, and the number of third-party accounts associated with the items, including the following steps (A)-(E): (A) For any target item category, the server sorts the items in the target item category in descending order of acquisition quantity to obtain a third item sequence. Based on the order of each item in the third item sequence, the first selection probability of each item is determined.

[0137] The earlier an item appears in the third item sequence, the greater its probability of being selected as the first item. Conversely, the later an item appears in the third item sequence, the lower its probability of being selected as the first item.

[0138] For example, the server determines the ratio of the order of an item in the third item sequence to the number of items in the third item sequence, and determines the difference between 1 and the ratio as the first selection probability.

[0139] (B) The server sorts the items in the target item category in descending order of the number of allocated resources to obtain the fourth item sequence. Based on the order of each item in the fourth item sequence, the second selection probability of each item is determined.

[0140] The earlier an item appears in the fourth item sequence, the greater its probability of being selected as the second item. Conversely, the later an item appears in the fourth item sequence, the lower its probability of being selected as the second item.

[0141] For example, the server determines the ratio of the order of an item in the fourth item sequence to the number of candidate items in the fourth item sequence, and determines the difference between 1 and the ratio as the second selection probability.

[0142] (C) The server sorts the items in the target item category in descending order of the number of corresponding third accounts to obtain the fifth item sequence. Based on the order of each item in the fifth item sequence, the third selection probability of each item is determined.

[0143] The earlier an item appears in the fifth item sequence, the greater its probability of being selected as the third item. Conversely, the later an item appears in the fifth item sequence, the lower its probability of being selected as the third item.

[0144] For example, the server determines the ratio of the order of an item in the fifth item sequence to the number of candidate items in the fifth item sequence, and determines the difference between 1 and the ratio as the third selection probability.

[0145] (D) The server performs weighted processing on the first, second, and third selection probabilities of each item to obtain the final selection probability of each item.

[0146] For example, the server performs a weighted sum of the first, second, and third selection probabilities for each item to obtain the final selection probability for each item. For example, the weights of the first, second, and third selection probabilities can be arbitrary, such as 0.3 for the first selection probability, 0.3 for the second selection probability, and 0.4 for the third selection probability.

[0147] (E) The server selects candidate items from the target item category according to the selection probability of multiple items in the target item category.

[0148] For example, the server sorts multiple items in descending order of selection probability and selects the top-ranked items (with a target quantity) as candidate items. The target quantity can be any number, such as 5 or 6. Alternatively, the server selects items with a selection probability greater than a probability threshold as candidate items. The probability threshold can be any value, such as 0.8. This method ensures that the selected candidate items have a high selection probability. A higher selection probability indicates a greater allocation of resources to the candidate item and a higher probability of it being acquired by the user; in other words, it means the candidate item is more likely to be liked by the user, thus improving the accuracy of item recommendations.

[0149] It should be noted that the server can also determine the selection probability of an item based on any one or two of the following: the acquisition quantity of the item, the amount of allocated resources corresponding to the item, and the number of third accounts corresponding to the item. For example, the third selection probability determined based on the number of third accounts corresponding to the item can be determined as the final selection probability of the item. This embodiment of the disclosure does not limit this.

[0150] 605. The server determines the availability and acquisition rate of multiple candidate items.

[0151] The listing rate indicates the probability that, when recommending alternative items to the first account, the alternative items will be listed on the first account's item supply platform. The acquisition rate indicates the probability that, when alternative items are listed on the item supply platform, other accounts besides the first account will be detected acquiring the listed alternative items.

[0152] 606. The server determines the recommendation probability of multiple candidate items based on their listing rate and acquisition rate.

[0153] Among them, the recommendation probability is positively correlated with the listing rate and the acquisition rate.

[0154] 607. The server recommends multiple alternative items to the first account based on the recommendation probability of multiple alternative items.

[0155] For the implementation of steps 605-607, please refer to steps 302-305 above, which will not be repeated here.

[0156] For example, after determining multiple candidate items based on their recommendation probabilities, the server reorders these candidate items according to their respective item categories, resulting in a candidate item sequence. In this sequence, adjacent candidate items do not belong to the same item category. This allows the terminal to display candidate items according to this sequence, avoiding over-concentration of items belonging to the same category. Alternatively, the server can reorder the candidate items according to their third-level item categories, again resulting in a candidate item sequence where adjacent candidate items do not belong to the same third-level item category. This method of reordering candidate items based on their third-level item categories is called diversity-based dispersion based on third-level item categories.

[0157] It should be noted that steps 601-604 above actually utilize the preferences of followers of the primary account to select candidate items. Of course, other methods can also be used to select candidate items, such as the following: The first method is historical item recall. For any item category among multiple item categories, the server determines the acquisition quantity corresponding to that item category. This acquisition quantity is the acquisition quantity of items in that item category on the first account's item supply platform. For example, if items A and B in this item category are listed on the first account's item supply platform, and item A has an acquisition quantity of 50 and item B has an acquisition quantity of 40, then the acquisition quantity corresponding to this item category is 90. Then, the server determines a first number of target item categories from multiple item categories, with the target item category having the highest acquisition quantity. Next, candidate items are selected from the target item categories. The implementation method for determining the first number of target item categories from multiple item categories and selecting candidate items from the target item categories is the same as the implementation method of steps 603-604 above, and will not be repeated here. The aforementioned method for recalling historical items is based on the acquisition volume of items in the first account's item supply platform. It determines the target item category that is most popular with users among the categories of items that the first account has listed, and selects alternative items from the target item category. This increases the likelihood that the alternative items recommended to the first account will be liked by users, thereby improving the accuracy of item recommendations.

[0158] The second method is graph recall. The behavioral data of multiple accounts registered on the server is represented using a bipartite graph. For example, the behavioral data includes multiple tuples, each tuple (u, i) representing that account u has interacted with item i. In the bipartite graph, the nodes corresponding to accounts belonging to the same tuple are directly connected to the nodes corresponding to items. Then, based on the relevance of the nodes of items not directly connected to the first account in the bipartite graph to the nodes of the first account, the multiple item nodes with the highest relevance are selected. The items corresponding to these multiple item nodes are the candidate items. For example, the relevance between the nodes of the first account and the item nodes is determined by at least one of the following: the number of paths between the two nodes, the path length between the two nodes (i.e., the number of nodes traversed by the path between the two nodes), and the out-degree of the nodes traversed by the path between the two nodes, where the number of outgoing edges of a node is called the out-degree of that node. This graph recall method, utilizing the structural relationship between account nodes and item nodes in the graph, can accurately determine the relevance between accounts and items, thereby selecting items with a high relevance to the first account as candidate items, thus improving the accuracy of item recommendations.

[0159] The third method is list-based recall. This involves selecting candidate items from a search volume ranking list. This list includes multiple items with high search volumes, sorted in descending order of search volume. Users can search for items on a product supply platform or retrieve items from the platform. The server counts the number of searches for each item to determine its search volume, creating a search volume ranking list. For example, the server selects candidate items from the search volume ranking list in the same way as selecting candidate items from the target product category, and will not be elaborated further. For instance, the server selects a target number of items ranked high in the search volume ranking list as candidate items. Since search volume indicates user interest, recommending high-search-volume candidate items based on the search volume ranking list increases the likelihood that users will like the recommended items, thus improving the accuracy of item recommendations.

[0160] It's important to note that the server can also select candidate items based on other rankings, such as the acquisition volume ranking. This ranking includes multiple items with high acquisition volumes, sorted in descending order of acquisition volume. Alternatively, the server can select candidate items based on the resource allocation quantity ranking. This ranking includes multiple items with high resource allocation quantities, sorted in descending order of resource allocation quantity. The method for selecting candidate items from other rankings is similar to that from the search volume ranking, and will not be elaborated upon here.

[0161] The fourth method is collaborative filtering. Collaborative filtering includes item-based collaborative filtering, which, when recommending items to the first account, identifies items liked by the user associated with that account and determines similar alternative items. It also includes user-based collaborative filtering, which, when recommending items to the first account, identifies similar accounts and determines alternative items liked by the users of those similar accounts. Collaborative filtering can identify alternative items liked by the user associated with the first account, thereby improving the accuracy of item recommendations.

[0162] The fifth method is the dual-tower model recall. This model includes a user feature extraction network and an item feature extraction network, used to acquire user features and item features respectively. For any given user, based on the distance between the user features and each item feature, candidate items that are closest to that user feature are selected. For example, the server calls the user feature extraction network in the dual-tower model to extract features based on the first account information, obtaining the first account feature. It then calls the item feature extraction network to extract features based on multiple item information, obtaining multiple item features. Finally, the similarity between the first account feature and each item feature is determined, and the target number of items with high similarity are selected as candidate items.

[0163] It should be noted that the above-mentioned methods of selecting alternative items can be combined in any way, and this disclosure does not limit this.

[0164] Figure 7 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment. Figure 6 Based on the illustrated embodiment, the process of screening candidate items is described. For example... Figure 7 As shown, the method includes the following steps.

[0165] 701. In response to the item recommendation instruction for the first account, the server retrieves multiple alternative items.

[0166] The implementation method for this step is as described above. Figure 6The embodiments shown are not described in detail here.

[0167] In some embodiments, after the server obtains multiple candidate items, it needs to filter these candidate items. For example, the server filters out duplicate candidate items, candidate items that do not meet the item recommendation criteria in the item recommendation request, and candidate items that have been recommended to the first account within a certain period of time before the current point in time. This can improve the accuracy of item recommendations.

[0168] 702. For any candidate item, the server obtains the number of third accounts corresponding to the candidate item. The server detects the acquisition operation of the candidate item in the item supply platform for the third account within the second time period before the current time point.

[0169] For any candidate item, if the server detects an acquisition operation of that candidate item from any item supply platform within a second time period prior to the current time, then the account corresponding to that item supply platform becomes the third account corresponding to that candidate item. The second time period can be any duration, for example, 3 days.

[0170] 703. The server sorts the multiple candidate items in descending order of the number of third-party accounts corresponding to them.

[0171] 704. The server retrieves the second-highest number of candidate items from the sorted list.

[0172] The second quantity can be any quantity, for example, 20.

[0173] In this embodiment of the disclosure, multiple candidate items are sorted in descending order of the number of third accounts corresponding to the multiple candidate items, and the second number of candidate items ranked first are obtained. This ensures that the number of third accounts corresponding to the obtained candidate items is relatively large. That is, within the second time period before the current time point, the number of item supply platforms that successfully obtain the item after it is listed by other users is relatively large, indicating that the candidate item is more likely to be liked by users. Therefore, the accuracy of item recommendation is improved.

[0174] It should be noted that steps 702-704 are only one method for filtering candidate items. Other methods can also be used. For example, the server can sort the candidate items obtained in step 701 in descending order of their acquisition quantity, resulting in a sixth item sequence. Based on the order of each candidate item in this sixth item sequence, the fourth selection probability of each item is determined, and then the second number of candidate items with the highest fourth selection probability are selected. Alternatively, the server can sort the candidate items obtained in step 701 in descending order of their allocated resource quantity, resulting in a seventh item sequence. Based on the order of each candidate item in this seventh item sequence, the fifth selection probability of each item is determined, and then the second number of candidate items with the highest fifth selection probability are selected. Alternatively, the server can filter candidate items by combining the number of corresponding third accounts, their acquisition quantity, and the allocated resource quantity. For example, the server sorts the multiple candidate items obtained from 701 in descending order of the number of corresponding third accounts to obtain the eighth item sequence. Based on the order of each candidate item in the eighth item sequence, the sixth selection probability of each item is determined. The fifth, sixth, and seventh selection probabilities of each candidate item are weighted to obtain the final selection probability of each candidate item. Then, the second number of candidate items with the highest final selection probability are selected from the multiple candidate items.

[0175] 705. The server determines the availability and acquisition rate of multiple candidate items.

[0176] The listing rate indicates the probability that, when an alternative item is recommended to the first account, the alternative item will be listed on the first account's item supply platform. The acquisition rate indicates the probability that, when an alternative item is listed on the item supply platform, other than the first account will be detected acquiring the listed alternative item.

[0177] 706. The server determines the recommendation probability of multiple candidate items based on their listing rate and acquisition rate. The recommendation probability is positively correlated with the listing rate and acquisition rate.

[0178] 707. The server recommends multiple alternative items to the first account based on the recommendation probability of multiple alternative items.

[0179] For the implementation of steps 705-707, please refer to steps 302-305 above, which will not be repeated here.

[0180] It should be noted that the above embodiments can be combined in any way, and this disclosure does not impose any restrictions on this.

[0181] Figure 8This is a schematic diagram illustrating the item recommendation process provided in an embodiment of this disclosure. (See reference...) Figure 8 After receiving an item recommendation request, the server uses a recall module to recall items based on various recall methods; that is, the recall module identifies multiple candidate items. These candidate items are then filtered. After filtering, the items are processed by a coarse-sorting module according to... Figure 7 The method shown sorts multiple candidate items and retrieves the top-ranked items. Among them, Figure 8 The truncation process involves retaining the top-ranked candidate items in the sorted item sequence while discarding the rest. The server then uses a fine-ranking module to determine the listing rate and acquisition rate of candidate items using both models. These probabilities are combined to obtain a recommendation probability, which is then sorted in descending order. The resulting item sequence is truncated, retaining the top-ranked items and discarding the rest. Next, the server uses a re-ranking module to re-rank the retained candidates as described in step 607, effectively dispersing them based on the diversity of the three-level item categories. Finally, the re-ranked item sequence is returned to the terminal that sent the item recommendation request.

[0182] Figure 9 This is a block diagram illustrating an information recommendation device according to an exemplary embodiment. (Refer to...) Figure 9 The device includes: The item acquisition unit 901 is configured to execute an item recommendation instruction in response to the first account and acquire multiple candidate items; The first determining unit 902 is configured to determine the listing rate and acquisition rate of multiple candidate items. The listing rate indicates the probability that candidate items are listed on the item supply platform of the first account when candidate items are recommended to the first account. The acquisition rate indicates the probability that acquisition operations of other accounts besides the first account are detected when candidate items are listed on the item supply platform. The second determining unit 903 is configured to perform the determination of the recommendation probability of multiple candidate items based on the listing rate and acquisition rate of multiple candidate items, wherein the recommendation probability is positively correlated with the listing rate and acquisition rate. Item recommendation unit 904 is configured to perform recommendation probabilities based on multiple candidate items, recommending multiple candidate items to the first account.

[0183] In this embodiment of the disclosure, since the listing rate of alternative items indicates the probability that alternative items will be listed on the first account's item supply platform when alternative items are recommended to the first account, it can indicate the degree of liking of the alternative items by the user to which the first account belongs. And the acquisition rate of alternative items indicates the probability that an acquisition operation of the listed alternative items will be detected when alternative items are listed on the first account's item supply platform, it can indicate the degree of liking of the alternative items by other users. Therefore, the recommendation probability of alternative items is determined based on the listing rate and acquisition rate of alternative items, and alternative items are recommended to the first account based on the recommendation probability, so that the items recommended to the first account are not only liked by the user to which the first account belongs, but also liked by other users. Therefore, the accuracy of item recommendation can be improved.

[0184] In some embodiments, the first determining unit 902 is configured to perform the following: obtain first account information, second account information, and item information of multiple candidate items, wherein the second account is an account that follows the first account; for any candidate item, invoke the acquisition rate prediction model to determine the acquisition rate of the candidate item based on the item information of the candidate item, the first account information, and the second account information.

[0185] In some embodiments, the training process of the acquisition rate prediction model includes: Obtain first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, other accounts besides the first sample account are detected to obtain the listed sample item, or no acquisition operation of the sample item is detected. The acquisition rate prediction model is invoked to determine the first predicted label of the sample item based on the first sample account information, the second sample account information, and the sample item information; Based on the first sample label and the first predicted label, a acquisition rate prediction model is trained to reduce the difference between the first predicted label obtained by the trained acquisition rate prediction model and the first sample label.

[0186] In some embodiments, obtaining the first sample label includes: If an operation to acquire a sample item is detected within the first time period after the sample item is listed, the first sample tag corresponding to the sample item will be determined as the first value. The listing time is the time when the sample item is listed on the item supply platform of the first sample account. If no acquisition operation of the sample item is detected within the first time period after the listing time, the first sample tag corresponding to the sample item will be determined as the second value.

[0187] In some embodiments, the second determining unit 903 is configured to perform the following operation: for any candidate item, multiply the listing rate of the candidate item by the acquisition rate to obtain the recommendation probability of the candidate item.

[0188] In some embodiments, the second determining unit 903 is configured to perform the following operations: determining a first item sequence and a second item sequence based on the listing rate and acquisition rate of multiple candidate items; the first item sequence is an item sequence obtained by sorting multiple candidate items in descending order of listing rate, and the second item sequence is an item sequence obtained by sorting multiple candidate items in descending order of acquisition rate; for any candidate item, determining a first recommendation probability of the candidate item based on its order in the first item sequence, and determining a second recommendation probability of the candidate item based on its order in the second item sequence; and weighting the first recommendation probability and the second recommendation probability to obtain the recommendation probability of the candidate item.

[0189] In some embodiments, the first determining unit 902 is configured to perform the following actions: obtaining first account information and item information of multiple candidate items; for any candidate item, invoking an availability prediction model to determine the availability rate of the candidate item based on the item information of the candidate item and the first account information.

[0190] In some embodiments, the training process of the shelf availability prediction model includes: Obtain second sample information, which includes third sample account information, sample item information, and second sample tag. The second sample tag indicates whether the sample item is listed on the third sample account's item supply platform after recommending the sample item to the third sample account, or whether the sample item is not listed on the item supply platform. The product listing rate prediction model is invoked, and the second predicted label for the sample product is determined based on the third sample account information and sample item information; Based on the second sample label and the second predicted label, a shelf-life prediction model is trained to reduce the difference between the second predicted label and the second sample label obtained by the trained shelf-life prediction model.

[0191] In some embodiments, obtaining the second sample label includes: After recommending sample items to the third sample account, and then listing the sample items on the third sample account's item supply platform, the second sample tag corresponding to the sample items is determined as the third value; If a sample item is recommended to a third sample account but is not listed on the third sample account's item supply platform, the second sample tag corresponding to the sample item will be designated as the fourth value.

[0192] In some embodiments, the item acquisition unit 901 is configured to execute an action in response to an item recommendation instruction, determine a second account, wherein the second account is an account that follows a first account; for any item category among multiple item categories, determine the number of second accounts that have acquired items within the item category from the determined second accounts; determine a first number of target item categories from the multiple item categories, wherein the number of second accounts corresponding to the target item categories is the largest; and select multiple candidate items from the target item categories.

[0193] In some embodiments, the item acquisition unit 901 is further configured to perform the following: for any candidate item, acquire the number of third accounts corresponding to the candidate item, wherein the server detects the acquisition operation of the candidate item in the item supply platform for the third account within a second time period before the current time point; sort the multiple candidate items in descending order of the number of third accounts corresponding to the multiple candidate items; and acquire the second number of candidate items that are ranked first.

[0194] Figure 10 This is a block diagram illustrating a model training apparatus according to an exemplary embodiment. (Refer to...) Figure 10 The device includes: The information acquisition unit 1001 is configured to acquire first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item by an account other than the first sample account is detected, or no acquisition operation of the sample item is detected. Model calling unit 1002 is configured to execute a call acquisition rate prediction model to determine the first predicted label of a sample item based on the first sample account information, the second sample account information, and the sample item information. The model training unit 1003 is configured to train the acquisition rate prediction model based on the first sample label and the first predicted label, so as to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

[0195] In this embodiment, considering that the first sample account information, the second sample account information, and the sample item information are all important factors that can influence the successful acquisition of sample items by other users in the item supply platform of the first sample account, an acquisition rate prediction model is trained based on these three types of information. This allows the model to learn the relationship between these three types of information and the first sample tag, i.e., the relationship between these three types of information and whether the sample items in the item supply platform of the first sample account are successfully acquired by other users. Therefore, based on the information of any account, the information of other accounts following that account, and the information of any item, it can accurately determine the probability that an item in the item supply platform of an account is acquired by other users. Subsequently, based on this probability, the recommendation probability of the item is determined, and items are recommended based on this recommendation probability. This increases the probability that the recommended items will be liked by users, thus improving the accuracy of item recommendations.

[0196] It should be noted that the information recommendation device and model training device provided in the above embodiments are only illustrative examples of the division of the above functional modules when recommending information or training models. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the electronic device can be divided into different functional modules to complete all or part of the functions described above. In addition, the information recommendation device and information recommendation method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments. Similarly, the model training device and model training method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0197] In an exemplary embodiment, an electronic device is also provided, comprising one or more processors and volatile or non-volatile memory for storing one or more processor-executable instructions, wherein the one or more processors are configured to execute the instructions to implement the information recommendation method or model training method described above.

[0198] Optionally, the electronic device is provided as a terminal. Figure 11 A structural block diagram of a terminal 1100 provided in an exemplary embodiment of this disclosure is shown. The terminal 1100 may be a smartphone, tablet computer, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer, or desktop computer. The terminal 1100 may also be referred to as a user device, portable terminal, laptop terminal, desktop terminal, or other names.

[0199] Terminal 1100 includes a processor 1101 and a memory 1102.

[0200] Processor 1101 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1101 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1101 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1101 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 1101 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0201] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1102 are used to store at least one program code, which is executed by the processor 1101 to implement the information recommendation method or model training method provided in the method embodiments of this disclosure.

[0202] In some embodiments, the terminal 1100 may also optionally include a peripheral device interface 1103 and at least one peripheral device. The processor 1101, memory 1102, and peripheral device interface 1103 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 1103 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of the following: a radio frequency circuit 1104, a display screen 1105, a camera assembly 1106, an audio circuit 1107, a positioning assembly 1108, and a power supply 1109.

[0203] Peripheral device interface 1103 can be used to connect at least one I / O (Input / Output) related peripheral device to processor 1101 and memory 1102. In some embodiments, processor 1101, memory 1102 and peripheral device interface 1103 are integrated on the same chip or circuit board; in some other embodiments, any one or two of processor 1101, memory 1102 and peripheral device interface 1103 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0204] The radio frequency (RF) circuit 1104 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 1104 communicates with communication networks and other communication devices via electromagnetic signals. The RF circuit 1104 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals back into electrical signals. Optionally, the RF circuit 1104 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The RF circuit 1104 can communicate with other terminals via at least one wireless communication protocol. This wireless communication protocol includes, but is not limited to: metropolitan area networks (MANs), various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks (WLANs), and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 1104 may also include circuitry related to NFC (Near Field Communication), which is not limited in this disclosure.

[0205] Display screen 1105 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 1105 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 1101 for processing. In this case, display screen 1105 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 1105, which serves as the front panel of terminal 1100; in other embodiments, there may be at least two display screens, respectively disposed on different surfaces of terminal 1100 or in a folded design; in still other embodiments, display screen 1105 may be a flexible display screen, disposed on a curved or folded surface of terminal 1100. Furthermore, display screen 1105 may also be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. The display screen 1105 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0206] The camera assembly 1106 is used to acquire images or videos. Optionally, the camera assembly 1106 includes a front-facing camera and a rear-facing camera. The front-facing camera is disposed on the front panel of the terminal, and the rear-facing camera is disposed on the back of the terminal. In some embodiments, there are at least two rear-facing cameras, which are any one of a main camera, a depth-sensing camera, a wide-angle camera, and a telephoto camera, to achieve background blurring by fusion of the main camera and the depth-sensing camera, panoramic shooting by fusion of the main camera and the wide-angle camera, VR (Virtual Reality) shooting, or other fusion shooting functions. In some embodiments, the camera assembly 1106 may also include a flash. The flash may be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cool light flash, which can be used for light compensation at different color temperatures.

[0207] The audio circuit 1107 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, converting the sound waves into electrical signals that are input to the processor 1101 for processing, or input to the radio frequency circuit 1104 for voice communication. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each positioned at a different location on the terminal 1100. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert electrical signals from the processor 1101 or the radio frequency circuit 1104 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 1107 may also include a headphone jack.

[0208] The positioning component 1108 is used to locate the current geographical location of the terminal 1100 in order to enable navigation or LBS (Location Based Service). The positioning component 1108 can be a positioning component based on the US GPS (Global Positioning System), China's BeiDou system, Russia's Granas system, or the EU's Galileo system.

[0209] Power supply 1109 is used to power the various components in terminal 1100. Power supply 1109 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 1109 includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0210] In some embodiments, the terminal 1100 further includes one or more sensors 1110. The one or more sensors 1110 include, but are not limited to: an accelerometer 1111, a gyroscope 1112, a pressure sensor 1113, a fingerprint sensor 1114, an optical sensor 1115, and a proximity sensor 1116.

[0211] Accelerometer 1111 can detect the magnitude of acceleration along the three axes of a coordinate system established with terminal 1100. For example, accelerometer 1111 can be used to detect the components of gravitational acceleration along the three axes. Processor 1101 can control display screen 1105 to display the user interface in either a landscape or portrait view based on the gravitational acceleration signal acquired by accelerometer 1111. Accelerometer 1111 can also be used for collecting game or user motion data.

[0212] The gyroscope sensor 1112 can detect the orientation and rotation angle of the terminal 1100. The gyroscope sensor 1112 can work in conjunction with the accelerometer sensor 1111 to collect the user's 3D movements on the terminal 1100. Based on the data collected by the gyroscope sensor 1112, the processor 1101 can perform the following functions: motion sensing (e.g., changing the UI based on the user's tilt), image stabilization during shooting, game control, and inertial navigation.

[0213] The pressure sensor 1113 can be disposed on the side bezel of the terminal 1100 and / or on the lower layer of the display screen 1105. When the pressure sensor 1113 is disposed on the side bezel of the terminal 1100, it can detect the user's grip signal on the terminal 1100, and the processor 1101 can perform left / right hand recognition or quick operation based on the grip signal collected by the pressure sensor 1113. When the pressure sensor 1113 is disposed on the lower layer of the display screen 1105, the processor 1101 can control the operable controls on the UI interface based on the user's pressure operation on the display screen 1105. The operable controls include at least one of button controls, scroll bar controls, icon controls, and menu controls.

[0214] The fingerprint sensor 1114 is used to collect the user's fingerprint. The processor 1101 identifies the user's identity based on the fingerprint collected by the fingerprint sensor 1114, or the fingerprint sensor 1114 identifies the user's identity based on the collected fingerprint. When the user's identity is identified as trusted, the processor 1101 authorizes the user to perform relevant sensitive operations, including unlocking the screen, viewing encrypted information, downloading software, making payments, and changing settings. The fingerprint sensor 1114 can be located on the front, back, or side of the terminal 1100. When the terminal 1100 has physical buttons or a manufacturer's logo, the fingerprint sensor 1114 can be integrated with the physical buttons or manufacturer's logo.

[0215] An optical sensor 1115 is used to collect ambient light intensity. In one embodiment, the processor 1101 can control the display brightness of the display screen 1105 based on the ambient light intensity collected by the optical sensor 1115. Specifically, when the ambient light intensity is high, the display brightness of the display screen 1105 is increased; when the ambient light intensity is low, the display brightness of the display screen 1105 is decreased. In another embodiment, the processor 1101 can also dynamically adjust the shooting parameters of the camera assembly 1106 based on the ambient light intensity collected by the optical sensor 1115.

[0216] The proximity sensor 1116, also known as a distance sensor, is installed on the front panel of the terminal 1100. The proximity sensor 1116 is used to detect the distance between the user and the front of the terminal 1100. In one embodiment, when the proximity sensor 1116 detects that the distance between the user and the front of the terminal 1100 is gradually decreasing, the processor 1101 controls the display screen 1105 to switch from a screen-on state to a screen-off state; when the proximity sensor 1116 detects that the distance between the user and the front of the terminal 1100 is gradually increasing, the processor 1101 controls the display screen 1105 to switch from a screen-off state to a screen-on state.

[0217] Those skilled in the art will understand that Figure 11 The structure shown does not constitute a limitation on terminal 1100 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0218] Alternatively, the electronic device may be provided as a server. Figure 12 This is a schematic diagram of a server structure provided in an embodiment of this disclosure. The server 1200 can vary significantly due to differences in configuration or performance. It may include one or more central processing units (CPUs) 1201 and one or more memories 1202. The memories 1202 store at least one line of program code, which is loaded and executed by the processors 1201 to implement the information recommendation method or model training method provided in the various method embodiments described above. Of course, the server may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The server may also include other components for implementing device functions, which will not be elaborated upon here.

[0219] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in an electronic device to perform the information recommendation method or model training method described above. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.

[0220] In an exemplary embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the information recommendation method or model training method described above.

[0221] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0222] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. An information recommendation method, characterized in that, include: In response to the item recommendation command for the primary account, retrieve multiple alternative items; The listing rate and acquisition rate of the plurality of candidate items are determined. The listing rate indicates the probability that the candidate item is listed on the item supply platform of the first account when the candidate item is recommended to the first account. The acquisition rate indicates the probability that, when the candidate item is listed on the item supply platform, other accounts besides the first account are detected to acquire the listed candidate item. Based on the listing rate and acquisition rate of the multiple candidate items, the recommendation probability of the multiple candidate items is determined, and the recommendation probability is positively correlated with the listing rate and the acquisition rate. Based on the recommendation probability of the multiple candidate items, the multiple candidate items are recommended to the first account; The determination of the recommendation probability of the multiple candidate items based on their listing rate and acquisition rate includes: For any candidate item, the recommendation probability of the candidate item is obtained by multiplying its listing rate by its acquisition rate; or, Based on the listing rate and acquisition rate of the plurality of candidate items, a first item sequence and a second item sequence are determined. The first item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of listing rate, and the second item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of acquisition rate. For any candidate item, a first recommendation probability of the candidate item is determined based on its order in the first item sequence, and a second recommendation probability of the candidate item is determined based on its order in the second item sequence. The first recommendation probability and the second recommendation probability are weighted to obtain the recommendation probability of the candidate item.

2. The information recommendation method according to claim 1, characterized in that, Determining the acquisition rate of the plurality of candidate items includes: Obtain information about a first account, information about a second account, and information about the items of the multiple candidate items, wherein the second account is an account that follows the first account; For any candidate item, the acquisition rate prediction model is invoked to determine the acquisition rate of the candidate item based on the item information of the candidate item, the first account information, and the second account information.

3. The information recommendation method according to claim 2, characterized in that, The training process of the acquisition rate prediction model includes: Obtain first sample information, which includes first sample account information, second sample account information, sample item information, and first sample tag. The second sample account is an account that follows the first sample account. The first sample tag indicates that after the sample item is listed on the item supply platform of the first sample account, an acquisition operation of the listed sample item by an account other than the first sample account is detected, or no acquisition operation of the sample item is detected. The acquisition rate prediction model is invoked to determine the first predicted label of the sample item based on the first sample account information, the second sample account information, and the sample item information; Based on the first sample label and the first predicted label, the acquisition rate prediction model is trained to reduce the difference between the first predicted label and the first sample label obtained by the trained acquisition rate prediction model.

4. The information recommendation method according to claim 3, characterized in that, Obtaining the first sample label includes: If an operation to acquire the sample item is detected within a first time period after the listing time of the sample item, the first sample tag corresponding to the sample item is determined as the first value, and the listing time is the time when the sample item is listed on the item supply platform of the first sample account. If no acquisition operation of the sample item is detected within the first time period after the listing time, the first sample label corresponding to the sample item is determined as the second value.

5. The information recommendation method according to claim 1, characterized in that, Determining the availability rate of the plurality of candidate items includes: Obtain the first account information and the item information of the multiple candidate items; For any candidate item, the listing rate prediction model is invoked, and the listing rate of the candidate item is determined based on the item information of the candidate item and the first account information.

6. The information recommendation method according to claim 5, characterized in that, The training process of the shelf availability prediction model includes: Obtain second sample information, which includes third sample account information, sample item information, and second sample tag. The second sample tag indicates whether the sample item is listed on the item supply platform of the third sample account after the sample item is recommended to the third sample account, or whether the sample item is not listed on the item supply platform. The product listing rate prediction model is invoked, and a second predicted tag for the sample item is determined based on the third sample account information and the sample item information; Based on the second sample label and the second predicted label, the shelf availability prediction model is trained to reduce the difference between the second predicted label obtained by the trained shelf availability prediction model and the second sample label.

7. The information recommendation method according to claim 6, characterized in that, Obtaining the second sample label includes: After recommending the sample item to the third sample account, and then listing the sample item on the item supply platform of the third sample account, the second sample tag corresponding to the sample item is determined as the third value. If the sample item is not listed on the item supply platform of the third sample account after recommending the sample item to the third sample account, the second sample tag corresponding to the sample item is determined as the fourth value.

8. The information recommendation method according to any one of claims 1-7, characterized in that, In response to the item recommendation instruction for the first account, multiple candidate items are obtained, including: In response to the item recommendation instruction, a second account is identified, which is an account that follows the first account; For any item category among multiple item categories, determine the number of second accounts that have acquired items within that item category from the identified second accounts; From the plurality of item categories, a first number of target item categories are determined, wherein the number of second accounts corresponding to the target item categories is the largest. Select the multiple candidate items from the target item category.

9. The information recommendation method according to any one of claims 1-7, characterized in that, Before determining the listing rate and acquisition rate of the plurality of candidate items, the information recommendation method further includes: For any candidate item, obtain the number of third accounts corresponding to the candidate item, wherein the server detects the acquisition operation of the candidate item in the item supply platform of the third account within a second time period before the current time point; The multiple candidate items are sorted in descending order of the number of third accounts corresponding to the multiple candidate items; Get the second number of the candidate items that are ranked first.

10. An information recommendation device, characterized in that, include: The item acquisition unit is configured to execute an item recommendation command in response to the first account and acquire multiple candidate items; The first determining unit is configured to determine the listing rate and acquisition rate of the plurality of candidate items, wherein the listing rate indicates the probability that the candidate items are listed on the first account's item supply platform when the candidate items are recommended to the first account, and the acquisition rate indicates the probability that acquisition operations of other accounts besides the first account are detected when the candidate items are listed on the item supply platform. The second determining unit is configured to perform a determination of the recommendation probability of the plurality of candidate items based on the listing rate and the acquisition rate of the plurality of candidate items, wherein the recommendation probability is positively correlated with the listing rate and the acquisition rate; The item recommendation unit is configured to recommend the multiple candidate items to the first account based on the recommendation probability of the multiple candidate items; The second determining unit is configured to perform the following operations for any candidate item: multiply the listing rate of the candidate item by the acquisition rate to obtain the recommendation probability of the candidate item; or, based on the listing rate and acquisition rate of the plurality of candidate items, determine a first item sequence and a second item sequence, wherein the first item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of listing rate, and the second item sequence is an item sequence obtained by sorting the plurality of candidate items in descending order of acquisition rate; for any candidate item, determine a first recommendation probability of the candidate item based on the order of the candidate item in the first item sequence, and determine a second recommendation probability of the candidate item based on the order of the candidate item in the second item sequence; The first recommendation probability and the second recommendation probability are weighted to obtain the recommendation probability of the candidate item.

11. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the information recommendation method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the information recommendation method as described in any one of claims 1-9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the information recommendation method according to any one of claims 1-9.

Images