Commodity Information Push Method, Intermediate Service Platform, Device and Storage Medium

Abstract

The present invention relates to the field of artificial intelligence technology, and discloses a method for pushing product information, an intermediate service platform, a computer device, and a storage medium. The method receives a product recommendation request sent by a preset business platform; determines a target server from the main server and all the backup servers; the target server is equipped with a virtual IP; sends the product recommendation request to the target server through the virtual IP, and determines the target recommendation model from the preset model platform and obtains initial recommendation information through the target server according to the scene call information; filters the initial recommendation information to select the target recommended product from all the initial recommended products; generates a product recommendation list based on all the target recommended products, and sends the product recommendation list to the preset business platform. The present invention improves the efficiency and accuracy of product recommendations.

Description

Technical Field

[0001] The present invention relates to the technical field of prediction models, and in particular, to a method for pushing commodity information, an intermediate service platform, a computer device, and a storage medium. Background Art

[0002] With the development of science and technology, more and more businesses with offline sales methods have gradually expanded their online sales methods. For example, by combining technologies such as artificial intelligence, corresponding products or commodities are recommended to users online.

[0003] In the prior art, generally, an artificial intelligence algorithm model trained by set business recommendation rules is used to recommend products or commodities. However, the above method has the following deficiencies: The business recommendation rules will change with the rapid changes of products, and thus new models need to be continuously trained according to the new business rules, resulting in low efficiency of product or commodity recommendation and prone to mismatched recommendations, thereby leading to low accuracy of product or commodity recommendation. Summary of the Invention

[0004] Embodiments of the present invention provide a method for pushing commodity information, an intermediate service platform, a computer device, and a storage medium to solve the problem of low accuracy of product or commodity recommendation in the prior art.

[0005] A method for pushing commodity information, which is applied to an intermediate service platform. The intermediate service platform includes a main server and at least one backup server. The method includes:

[0006] Receiving a commodity recommendation request sent by a preset service platform; the commodity recommendation request includes scenario call information;

[0007] Determining a target server from the main server and all the backup servers; a virtual IP is carried on the target server;

[0008] Sending the commodity recommendation request to the target server through the virtual IP, and determining a target recommendation model from a preset model platform and obtaining initial recommendation information by the target server according to the scenario call information; the initial recommendation information is generated by the target recommendation model according to the scenario call information; the initial recommendation information includes at least one initial recommended commodity;

[0009] Filtering commodities from the initial recommendation information to screen out target recommended commodities from all the initial recommended commodities;

[0010] Generating a commodity recommendation list according to all the target recommended commodities and sending the commodity recommendation list to the preset service platform.

[0011] A commodity information push device, comprising:

[0012] A recommendation instruction receiving module, configured to receive a commodity recommendation request sent by a preset service platform; the commodity recommendation request includes scenario invocation information;

[0013] A server selection module, configured to determine a target server from the primary server and all the backup servers; a virtual IP is carried on the target server;

[0014] A recommendation information acquisition module, configured to send the commodity recommendation request to the target server through the virtual IP, and determine a target recommendation model and acquire initial recommendation information from a preset model platform by the target server according to the scenario invocation information; the initial recommendation information is generated by the target recommendation model according to the scenario invocation information; the initial recommendation information includes at least one initial recommended commodity;

[0015] A recommendation information filtering module, configured to filter commodities from the initial recommendation information to screen out target recommended commodities from all the initial recommended commodities;

[0016] A recommendation list generation module, configured to generate a commodity recommendation list according to all the target recommended commodities and send the commodity recommendation list to the preset service platform.

[0017] A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, where when the processor executes the computer program, the above-mentioned commodity information push method is implemented.

[0018] A computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned commodity information push method is implemented.

[0019] The above-mentioned method for pushing commodity information, intermediate service platform, computer device and storage medium. The method includes receiving a commodity recommendation request sent by a preset service platform; the commodity recommendation request includes scenario call information; determining a target server from the primary server and all the backup servers; a virtual IP is carried on the target server; sending the commodity recommendation request to the target server through the virtual IP, and the target server determines a target recommendation model from a preset model platform and obtains initial recommendation information according to the scenario call information; the initial recommendation information is generated by the target recommendation model according to the scenario call information; the initial recommendation information includes at least one initial recommended commodity; filtering the initial recommendation information to screen out target recommended commodities from all the initial recommended commodities; generating a commodity recommendation list according to all the target recommended commodities and sending the commodity recommendation list to the preset service platform.

[0020] The present invention sets an intermediate service platform, and the intermediate service platform provides a unified service interface externally (such as a preset service platform). Other platforms only need to input according to the parameter form set by the service interface, and then can call the corresponding algorithm model through the intermediate service platform for commodity recommendation. In this way, the efficiency of commodity recommendation is improved, and at the same time, the cost of repeated development is reduced. Further, by completing the operation of commodity filtering through the intermediate service platform, the corresponding filtering rules can be incorporated into commodity recommendation, thereby improving the accuracy of commodity recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 is a schematic diagram of an application environment of the method for pushing commodity information in an embodiment of the present invention;

[0023] Figure 2 is a flowchart of the method for pushing commodity information in an embodiment of the present invention;

[0024] Figure 3 is a schematic block diagram of an intermediate service platform in an embodiment of the present invention;

[0025] Figure 4 is a schematic diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] The commodity information push method provided by the embodiment of the present invention can be applied to the application environment as Figure 1 shown. Specifically, this commodity information push method is applied in a commodity information push system, and this commodity information push system includes a preset service platform, an intermediate service platform, and a preset model platform as Figure 1 shown. The preset service platform and the intermediate service platform, as well as between the intermediate service platform and the preset model platform, communicate through a network to solve the problem of low accuracy in product or commodity recommendation in the prior art. Among them, the preset service platform can be a user terminal, which refers to a program that provides local services for customers corresponding to the server. The client can be installed on, but not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The intermediate service platform can be implemented by an independent server or a server cluster composed of multiple servers. Among them, the server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0028] In one embodiment, as Figure 2 shown, a commodity information push method is provided. Taking the method applied to the Figure 1 intermediate service platform as an example, it includes the following steps:

[0029] S10: Receive a commodity recommendation request sent by the preset service platform; the commodity recommendation request includes scenario call information.

[0030] Understandably, the preset business platform represents the platform of the business channel party. The product recommendation request sent by the preset business platform can be automatically generated after the user or the salesperson inputs the corresponding scenario call information, or the user or the salesperson can transmit the product recommendation request containing the scenario call information to the preset business platform, which is not limited here. Among them, the scenario call information includes information such as scenario ID, scenario name (i.e., model name), recommended personalized information, whether to use ABTest (AB testing), version ID1 (one version of the model), version ID2 (another version of the model), etc.

[0031] Furthermore, a unified external interface is provided to other platforms (such as the preset business platform) in the intermediate service platform. The intermediate service platform provides several standard forms of output values, such as dictionaries, lists, etc. After determining the parameters and their representative meanings, the external interface of the intermediate service platform can be set. This external interface is applicable to the calls of all external interfaces of the same type of input to the model. It avoids the drawback that each model needs to provide a different external interface, realizes the unification of the external interface specifications of all algorithm models in the preset model platform, and supports the dynamic orchestration and combination of services.

[0032] S20: Determine a target server from the main server and all the backup servers; a virtual IP is carried on the target server.

[0033] Understandably, in the present invention, there are multiple servers in the intermediate service platform, that is, the intermediate service platform is equivalent to a server group, including a main server and multiple backup servers. Therefore, the target server is the main server or any one of the backup servers. The virtual IP is the public network virtual IP, which is used to forward the product recommendation request; furthermore, the virtual IP is pre-fixed on the main server. If the main server is in a normal service state, that is, the main server can normally execute steps such as the recommendation of the target recommendation model, then the main server can be used as the target server; if the server is not in a normal service state (such as a service outage state), the virtual IP can be carried to any one of the backup servers in a normal service state, so that the backup server carrying the virtual IP is used as the target server and completes steps such as the recommendation of the target recommendation model. In this way, when the server (main server or backup server) carrying the virtual IP fails, the virtual IP can be quickly transferred to other servers in a normal service state, reducing the impact of server failure on the response to service access requests, thereby improving the response rate of service access requests.

[0034] Further, the primary server can send multicast messages. If the backup servers do not receive the multicast messages sent by the primary server within a certain period of time, it is considered that the primary server is not in a normal service state (such as a service downtime state); if the backup servers can receive the multicast messages sent by the primary server within a certain period of time, it is considered that the primary server is in a normal service state. When the primary server is in a normal service state, the primary server is determined as the target server; when the primary server is not in a normal service state, the virtual IP can be transferred to other backup servers in a normal service state, so that the backup server carrying the virtual IP serves as the target server.

[0035] S30: Send the commodity recommendation request to the target server through the virtual IP, and through the target server, determine a target recommendation model from a preset model platform according to the scenario call information and obtain initial recommendation information; the initial recommendation information is generated by the target recommendation model according to the scenario call information; the initial recommendation information includes at least one initial recommended commodity.

[0036] Understandably, the preset model platform stores algorithm models for iterative update of model interfaces. The intermediate service platform records relevant information of these algorithm models, as well as the corresponding relationship between the algorithm models and the model interfaces. The target recommendation model refers to one or two models in the preset model platform, and the number of models is determined according to whether the A / B test is adopted in the scenario call information. For example, when the scenario call information indicates that the A / B test needs to be adopted, it means that different versions of models in the same scenario need to be called. The initial recommendation information refers to after determining the target recommendation model, calling the target recommendation model through the interface corresponding to the target recommendation model, so as to generate the initial recommendation information according to the recommended personalized information in the scenario call information by the target recommendation model; further, the initial recommendation information includes at least one initial recommended commodity.

[0037] Specifically, after determining the target server from the primary server and all the backup servers, the target server can parse the corresponding model requirement information (such as the above-mentioned scenario ID, scenario name, etc.) according to the scenario call information in the commodity recommendation request, and then determine the corresponding target recommendation model from the preset model platform according to the model requirement information. Further, after determining the target recommendation model, the target recommendation model can be called according to the interface corresponding to the target recommendation model, and the scenario call information is input into the target recommendation model, so that the target recommendation model can output the corresponding initial recommended commodities according to the scenario call information, and all the initial recommended commodities are combined to form the initial recommendation information.

[0038] S30: Filter the commodities in the initial recommendation information to screen out the target recommended commodities from all the initial recommended commodities.

[0039] Understandably, after the target recommendation model outputs the corresponding initial recommendation information, the user may have their own purchase habits (for example, the user does not like products of certain brands), or in the historical recommendation process, the user feedbacks that some products are not suitable for recommendation in this scenario. Therefore, it is necessary to filter the initial recommended products in the initial recommendation information. In this embodiment, the initial recommendation information is filtered by two methods. One is: filtering the initial recommendation information according to the product filtering rule information in the product recommendation request sent by the preset business platform; wherein, the product filtering rule information is a filtering rule customized by the user or the salesperson. For example, the product filtering rule information indicates that products of a certain brand need to be filtered when recommending products; the other is filtering the initial recommendation information according to the scenario filtering rule information in the preset scenario-rule mapping table stored in the intermediate service platform that matches the scenario call information; wherein, the scenario filtering rule information records that products in certain scenarios are not allowed to be sold. Therefore, it is meaningless to recommend such products to users. Therefore, these products can be filtered according to the scenario filtering rule information. Further, the examples explained above are only one example of the rules, and other suitable rules are also applicable, and no more examples are given here.

[0040] S40: Generate a product recommendation list based on all the target recommended products, and send the product recommendation list to the preset business platform.

[0041] Specifically, after filtering the initial recommendation information to screen out the target recommended products from all the initial recommended products, generate a product recommendation list based on all the target recommended products, and send the product recommendation list to the preset business platform, so as to complete the complete product recommendation process.

[0042] In this embodiment, by setting an intermediate service platform, and this intermediate service platform provides a unified service interface to the outside (such as the preset business platform), other platforms only need to input according to the parameter form set by this service interface, and then they can call the corresponding algorithm model through this intermediate service platform for product recommendation. In this way, the efficiency of product recommendation is improved, and at the same time, the cost of repeated development is reduced. Further, by completing the operation of product filtering through the intermediate service platform, the product recommendation can incorporate the corresponding filtering rules, thereby improving the accuracy of product recommendation.

[0043] In one embodiment, in step S20, that is, determining the target recommendation model and obtaining the initial recommendation information from the preset model platform according to the scenario call information, includes:

[0044] Parse the scene call information to obtain the scene call table fields; the scene call table fields include basic model call information and model call version information.

[0045] Understandably, parsing the scene call information can be a process such as entity recognition of the scene call information, so as to determine the basic model call information and model call version information included in the scene call information. Further, as pointed out in the above description, the scene call information includes information such as scene ID, scene name (i.e., model name), recommended personalized information, whether to use A / B test (A / B testing), version ID1 (one of the model versions), version ID2 (another version of the model), etc. Therefore, the basic model call information referred to here is the scene ID, scene name, recommended personalized information, and whether to use A / B test; the model call version information is version ID1 and version ID2 (if A / B test is not used, the model call version information is one version ID).

[0046] Obtain the model maintenance list; the model maintenance list includes at least one model maintenance triple; one model maintenance triple includes basic model maintenance information, model maintenance version information, and model interface information.

[0047] Understandably, there are multiple algorithm models stored in the preset model platform, and each algorithm model may have different versions. Therefore, the model maintenance list is stored in the database of the intermediate service platform. The model maintenance list includes at least one model maintenance triple, and the model maintenance triple is used to represent different algorithm models or different model versions; among them, one model maintenance triple includes basic model maintenance information, model maintenance version information, and model interface information. Further, the basic model maintenance information may include, for example, scene ID, scene name (i.e., model name); the model maintenance version information refers to the version information corresponding to the algorithm model; the model interface information refers to the interface information corresponding to each different version of the algorithm model or different algorithm models, that is, one algorithm model corresponds to one call interface.

[0048] Determine the target triple from all the model maintenance triples according to the basic model call information, model call version information, basic model maintenance information, and model maintenance version information.

[0049] Specifically, after obtaining the model maintenance list, the basic model call information can be matched with the basic model maintenance information, and the model call version information can be matched with the model maintenance version information, so as to determine a model maintenance triple with the same basic model maintenance information as the basic model call information and the same model maintenance version information as the model call version information, and record this model maintenance triple as the target triple.

[0050] Determine the target recommended model according to the basic model maintenance information and the model maintenance version information in the target triple, and call the target recommended model according to the model interface information in the target triple, so that the target recommended model outputs the initial recommendation information according to the scenario call information.

[0051] Specifically, after determining the target triple from all the model maintenance triples according to the basic model call information, the model call version information, the basic model maintenance information, and the model maintenance version information, determine the target recommended model according to the basic model maintenance information and the model maintenance version information in the target triple, and call the target recommended model according to the model interface information uniquely corresponding to the target recommended model indicated in the target triple, so that the target recommended model performs product recommendation according to the recommendation personalization information in the scenario call information, thereby obtaining the initial recommendation information.

[0052] In one embodiment, the step of calling the target recommended model according to the model interface information in the target triple, so that the target recommended model outputs the initial recommendation information according to the scenario call information, includes:

[0053] Determine the interface call method corresponding to the scenario call information according to the basic model call information and the model call version information.

[0054] It can be understood that in this embodiment, there are two interface call methods: one is the ordinary call method, that is, only one algorithm model needs to be called; the other is the shunt call method, that is, it is necessary to use the ABTest method for calling. At this time, the names of two models may appear in the basic model call information, or two different version IDs may appear in the model call version information. In this way, the interface call method corresponding to the scenario call information can be determined through the basic model call information and the model call version information.

[0055] When the interface call method is the shunt call method, the target server classifies the recommended population according to the scenario call information to obtain population classification labels.

[0056] Understandably, it is pointed out in the above description that the scenario call information includes recommended personalized information, and the recommended personalized information may include the characteristic information of multiple people. For example, when salespersons or testers are conducting tests, they can transmit the activity degree of user access page information and relevant characteristics of accessed sections, such as: interaction duration, click-through rate, accessed topics, etc. Furthermore, the recommended population can be classified according to the recommended personalized information in the scenario call information, so as to obtain population classification labels.

[0057] Apply the AB test distribution rule strategy, and the target server generates an AB recommendation plan according to the scenario call information, the scenario call information, and the population classification label.

[0058] Understandably, the AB test distribution rule strategy is a product recommendation strategy that can split multiple different versions to different users in real time. The number and proportion of population groups will be continuously adjusted according to the results of the AB test. Finally, the recommendation plan to be released to all users is determined. The process of function assembly is to locate the insertion position of the cloud function code that matches the population classification label in the main function code corresponding to the page information (which can be the page information accessed by the user to be tested), insert each cloud function code at the corresponding insertion position, package the inserted main function, and generate an AB test plan, and associate the AB test plan with the page information and the population group label.

[0059] Among them, the main function code is the code of the main line created for the upgrade or modification of the corresponding page information. The cloud function code is an independent function code that developers only need to provide a simple function method name and code snippet to go online and use. By adding a trigger to trigger its operation to the cloud function code, it can be triggered to respond and be used in the main function code.

[0060] Call the target recommendation model according to the model interface information in the target triple, so that the target recommendation model outputs the initial recommendation information according to the AB recommendation plan.

[0061] Specifically, since there are two different models or two different versions of models in the split call method, there are two different model interface information in the split call method. Furthermore, after calling two target recommendation models according to the two different model interface information, the target recommendation model can recommend products according to the AB recommendation plan, and obtain the initial recommendation information output by the two target recommendation models. Furthermore, the algorithm model can be continuously improved according to the user feedback results corresponding to the initial recommendation information output by the two target recommendation models, so as to continuously improve the recommendation accuracy of the algorithm model.

[0062] In one embodiment, determining a target triple from all the model maintenance triples according to the basic information for model invocation, model invocation version information, basic model maintenance information, and model maintenance version information includes:

[0063] Matching the basic information for model invocation with the basic model maintenance information, and matching the model invocation version information with the model maintenance version information.

[0064] Determining the model maintenance triple with the basic model maintenance information that matches the basic information for model invocation and the model maintenance version information that matches the model invocation version information as the target triple.

[0065] Understandably, after obtaining the model maintenance list, the basic information for model invocation can be directly matched with the basic model maintenance information in the model maintenance list, and the model invocation version information can be matched with the model maintenance version information in the model maintenance list, so as to determine the model maintenance triple with the basic model maintenance information that matches the basic information for model invocation and the model maintenance version information that matches the model invocation version information as the target triple.

[0066] In one embodiment, the commodity recommendation request further includes commodity filtering rule information; filtering the initial recommendation information for commodities to screen out target recommended commodities from all the initial recommended commodities includes:

[0067] Obtaining the scenario filtering rule information corresponding to the scenario invocation information from a preset scenario-rule mapping table.

[0068] Understandably, the preset scenario-rule mapping table can be set in advance according to historical recommendation records (such as the feedback results of users after commodity recommendations), for example, some commodities are not allowed to be sold in corresponding scenarios in historical recommendation records, or there is no user consumption record after some commodities are recommended to users, etc., so as to construct the preset scenario-rule mapping table. Further, there is at least one set of scenario-rule mapping relationships in the preset scenario-rule mapping table, and a set of scenario-rule mapping relationships represents the corresponding commodity filtering rules in a certain scenario.

[0069] Further, after obtaining the initial recommendation information, obtain the preset scenario-rule mapping table, and query the scenario-rule mapping relationship corresponding to the basic information for model invocation included in the scenario invocation information from the preset scenario-rule mapping table, that is, query the scenario-rule mapping relationship including the basic information for model invocation from the preset scenario-rule mapping table, and then extract the filtering rules in the queried scenario-rule mapping relationship including the basic information for model invocation to obtain the scenario filtering rule information.

[0070] Perform a primary product filtering on the initial recommendation information according to the scenario filtering rule information, so as to screen out scenario recommendation products from all the initial recommended products.

[0071] Specifically, after obtaining the scenario filtering rule information corresponding to the scenario call information, the initial recommendation information can be subjected to a primary product filtering according to the scenario filtering rule information. For example, the corresponding product name or the link corresponding to the product is recorded in the scenario filtering rule information. Therefore, it is possible to directly query whether there is the corresponding product name in the initial recommendation information, or query through the link corresponding to the product, so as to realize the primary product filtering of the initial recommendation information, and further screen out scenario recommendation products from all the initial recommended products.

[0072] Perform a secondary product filtering on the scenario recommendation products according to the product filtering rule information, so as to screen out the target recommendation products from all the scenario recommendation products.

[0073] Specifically, after performing a primary product filtering on the initial recommendation information according to the scenario filtering rule information to screen out scenario recommendation products from all the initial recommended products, the product filtering rule information included in the product recommendation request is parsed to determine whether the product filtering rule information is empty. For example, a new user may need more product recommendations for selection, and this new user may not set product filtering rules, so the product filtering rule information is empty. If it is empty, no filtering is required; if it is not empty, a secondary product filtering is performed on the scenario recommendation products according to the product filtering rule information, so as to screen out the target recommendation products from all the scenario recommendation products. In this way, through the two-layer product filtering rules, the filtering of the recommended products can be completed, thereby improving the accuracy of product recommendations.

[0074] In one embodiment, the performing a secondary product filtering on the scenario recommendation products according to the product filtering rule information to screen out the target recommendation products from all the scenario recommendation products includes:

[0075] Perform a null value detection on the product filtering rule information to determine whether the product filtering rule information is empty.

[0076] Specifically, after performing a primary product filtering on the initial recommendation information according to the scenario filtering rule information to screen out scenario recommendation products from all the initial recommended products, a null value detection is performed on the product filtering rule information, that is, it is detected whether the corresponding rule information is included in the product filtering rule information.

[0077] When the product filtering rule information is not empty, re-filter the scenario-recommended products according to the product filtering rule information to obtain the target recommended products.

[0078] When the product filtering rule information is empty, record the scenario-recommended products as the target recommended products.

[0079] Specifically, after detecting whether the product filtering rule information is null to determine whether the product filtering rule information is empty, if the product filtering rule information is not empty, parse the product filtering rule information to determine information such as the product name or product link that needs to be filtered, so as to re-filter the scenario-recommended products according to the product name or product link. If the scenario-recommended product is a product that needs to be filtered mentioned in the product filtering rule information, then exclude the scenario-recommended product, so as to determine the scenario-recommended products obtained after filtering as the target recommended products.

[0080] Furthermore, if the product filtering rule information is empty, it means that the user has not set the corresponding product filtering rule yet, and then the scenario-recommended products obtained above can be directly determined as the target recommended products.

[0081] In one embodiment, in step S20, that is, determining the target server from the main server and all the backup servers includes:

[0082] Detect whether the main server is in a normal service state; the virtual IP is carried on the main server.

[0083] Specifically, the main server can send multicast information. If the backup servers do not receive the multicast information sent by the main server within a certain period of time, it is considered that the main server is not in a normal service state (such as a service downtime state); if the backup servers can receive the multicast information sent by the main server within a certain period of time, it is considered that the main server is in a normal service state. When the main server is in a normal service state, determine the main server as the target server; when the main server is not in a normal service state, the virtual IP can be transferred to other backup servers in a normal service state, so that the backup server carrying the virtual IP is used as the target server.

[0084] When the main server is in a normal service state, record the main server as the target server.

[0085] Specifically, after detecting whether the main server is in a normal service state, if the main server is in a normal service state, the main server is used as the target server to perform operations such as target recommendation model selection.

[0086] When the main server is not in a normal service state, transfer the virtual IP to any backup server in a normal service state, and record the backup server carrying the virtual IP as the target server.

[0087] Specifically, after detecting whether the main server is in a normal service state, if the main server is not in a normal service state, that is, when the main server is in a service outage state, detect whether each backup server is in a normal service state. Then, the virtual IP can be transferred to any backup server, and the backup server carrying the virtual IP is used as the target server to perform operations such as target recommendation model selection.

[0088] It should be understood that the sequence numbers of the steps in the above embodiments do not indicate the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0089] In one embodiment, an intermediate service platform is provided, which corresponds to the commodity information push method in the above embodiment one by one. As Figure 3 shown, the intermediate service platform includes a recommendation instruction receiving module 10, a server selection module 20, a recommendation information obtaining module 30, a recommendation information filtering module 40, and a recommendation list generating module 50. The detailed description of each functional module is as follows:

[0090] The recommendation instruction receiving module 10 is used to receive a commodity recommendation request sent by a preset service platform; the commodity recommendation request includes scenario call information;

[0091] The server selection module 20 is used to determine a target server from the main server and all backup servers; the target server carries a virtual IP;

[0092] The recommendation information obtaining module 30 is used to send the commodity recommendation request to the target server through the virtual IP, and the target server determines a target recommendation model and obtains initial recommendation information from a preset model platform according to the scenario call information; the initial recommendation information is generated by the target recommendation model according to the scenario call information; the initial recommendation information includes at least one initial recommended commodity;

[0093] The recommendation information filtering module 40 is used to filter commodities from the initial recommendation information to screen out target recommended commodities from all the initial recommended commodities;

[0094] The recommendation list generating module 50 is used to generate a commodity recommendation list according to all the target recommended commodities and send the commodity recommendation list to the preset service platform.

[0095] Preferably, the recommended information acquisition module 30 includes:

[0096] An information parsing unit, configured to parse the scenario call information to obtain scenario call table fields; the scenario call table fields include model call basic information and model call version information;

[0097] A list acquisition unit, configured to acquire a model maintenance list; the model maintenance list includes at least one model maintenance triple; one model maintenance triple includes model maintenance basic information, model maintenance version information, and model interface information;

[0098] An information matching unit, configured to determine a target triple from all the model maintenance triples according to the model call basic information, the model call version information, the model maintenance basic information, and the model maintenance version information;

[0099] A model call unit, configured to determine the target recommended model according to the model maintenance basic information and the model maintenance version information in the target triple, and call the target recommended model according to the model interface information in the target triple, so that the target recommended model outputs the initial recommended information according to the scenario call information.

[0100] Preferably, the model call unit includes:

[0101] An interface call method determination subunit, configured to determine an interface call method corresponding to the scenario call information according to the model call basic information and the model call version information;

[0102] A population classification subunit, configured to perform recommended population classification according to the scenario call information through the target server to obtain population classification labels when the interface call method is a shunt call method;

[0103] A recommended solution determination subunit, configured to generate an AB recommendation solution according to the scenario call information and the population classification labels through the target server by using an AB test distribution rule strategy;

[0104] A model call subunit, configured to call the target recommended model according to the model interface information in the target triple, so that the target recommended model outputs the initial recommended information according to the AB recommendation solution.

[0105] Preferably, the information matching unit includes:

[0106] An information matching subunit, configured to match the model call basic information with the model maintenance basic information, and match the model call version information with the model maintenance version information;

[0107] A target triple determination subunit, configured to determine, as the target triple, a model maintenance triple that matches the model call basic information and corresponds to the model maintenance version information that matches the model call version information.

[0108] Preferably, the recommendation information filtering module 40 includes:

[0109] An information acquisition unit, configured to acquire, from a preset scenario - rule mapping table, scenario filtering rule information corresponding to the scenario call information;

[0110] A first filtering unit, configured to perform an initial product filtering on the initial recommendation information according to the scenario filtering rule information, so as to screen out scenario - recommended products from all the initial recommended products;

[0111] A second filtering unit, configured to perform a secondary product filtering on the scenario - recommended products according to the product filtering rule information, so as to screen out the target recommended products from all the scenario - recommended products.

[0112] Preferably, the second filtering unit includes:

[0113] A null value detection unit, configured to perform a null value detection on the product filtering rule information to determine whether the product filtering rule information is null;

[0114] A first target recommended product determination unit, configured to, when the product filtering rule information is not null, perform a secondary product filtering on the scenario - recommended products according to the product filtering rule information to obtain the target recommended products;

[0115] A second target recommended product determination unit, configured to, when the product filtering rule information is null, record the scenario - recommended products as the target recommended products.

[0116] Preferably, the server selection module 20 includes:

[0117] A server detection unit, configured to detect whether the primary server is in a normal service state; the virtual IP is carried on the primary server.

[0118] A first server selection unit, configured to, when the primary server is in a normal service state, record the primary server as the target server;

[0119] A second server selection unit, configured to, when the primary server is not in a normal service state, transfer the virtual IP to any backup server in a normal service state, and record the backup server carrying the virtual IP as the target server.

[0120] For the specific limitations of the commodity information push device, reference can be made to the limitations of the commodity information push method in the foregoing text, which will not be elaborated herein. Each module in the above commodity information push device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0121] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 4 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the data used in the commodity information push method in the above embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a commodity information push method.

[0122] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the commodity information push method in the above embodiment.

[0123] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the commodity information push method in the above embodiment.

[0124] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0125] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or equivalently replace some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention and should all be included in the protection scope of the present invention.

Claims

1. A method for pushing commodity information, characterized in that, the method for pushing commodity information is applied to an intermediate service platform, and the intermediate service platform includes a main server and at least one backup server; the method includes: receiving a commodity recommendation request sent by a preset service platform; the commodity recommendation request includes scenario call information; the preset service platform refers to the user side of a program that provides local services corresponding to the intermediate service platform; determining a target server from the main server and all the backup servers; a virtual IP is carried on the target server; sending the commodity recommendation request to the target server through the virtual IP, and determining a target recommendation model from a preset model platform and obtaining initial recommendation information through the target server according to the scenario call information; the initial recommendation information is generated by the target recommendation model according to the scenario call information; the initial recommendation information includes at least one initial recommended commodity; filtering the initial recommendation information for commodities to screen out target recommended commodities from all the initial recommended commodities; generating a commodity recommendation list according to all the target recommended commodities, and sending the commodity recommendation list to the preset service platform; the determining a target recommendation model from a preset model platform and obtaining initial recommendation information according to the scenario call information includes: analyzing the scenario call information to obtain scenario call table fields; the scenario call table fields include model call basic information and model call version information; obtaining a model maintenance list; the model maintenance list includes at least one model maintenance triple; one model maintenance triple includes model maintenance basic information, model maintenance version information and model interface information; determining a target triple from all the model maintenance triples according to the model call basic information, model call version information, model maintenance basic information and model maintenance version information; determining the target recommendation model according to the model maintenance basic information and model maintenance version information in the target triple, and calling the target recommendation model according to the model interface information in the target triple, so that the target recommendation model outputs the initial recommendation information according to the scenario call information; the determining a target triple from all the model maintenance triples according to the model call basic information, model call version information, model maintenance basic information and model maintenance version information includes: matching the model call basic information with the model maintenance basic information, and matching the model call version information with the model maintenance version information; determining the model maintenance triple with model maintenance basic information matching the model call basic information and model maintenance version information matching the model call version information as the target triple.

2. The method for pushing commodity information according to claim 1, characterized in that, Invoking the target recommendation model according to the model interface information in the target triple, so that the target recommendation model outputs the initial recommendation information according to the scenario invocation information, includes: Determining an interface invocation method corresponding to the scenario invocation information according to the basic model invocation information and the model invocation version information; When the interface invocation method is a shunt invocation method, classifying the recommended population according to the scenario invocation information through the target server to obtain population classification labels; Applying the AB test distribution rule strategy to generate an AB recommendation plan through the target server according to the scenario invocation information and the population classification labels; Invoking the target recommendation model according to the model interface information in the target triple, so that the target recommendation model outputs the initial recommendation information according to the AB recommendation plan.

3. The commodity information push method according to claim 1, characterized in that, The commodity recommendation request further includes commodity filtering rule information; filtering the initial recommendation information for commodities to screen out target recommended commodities from all the initial recommended commodities, includes: Obtaining scenario filtering rule information corresponding to the scenario invocation information from a preset scenario-rule mapping table; Performing primary commodity filtering on the initial recommendation information according to the scenario filtering rule information to screen out scenario recommended commodities from all the initial recommended commodities; Performing secondary commodity filtering on the scenario recommended commodities according to the commodity filtering rule information to screen out the target recommended commodities from all the scenario recommended commodities.

4. The commodity information push method according to claim 3, characterized in that, Performing secondary commodity filtering on the scenario recommended commodities according to the commodity filtering rule information to screen out the target recommended commodities from all the scenario recommended commodities, includes: Performing a null value detection on the commodity filtering rule information to determine whether the commodity filtering rule information is null; When the commodity filtering rule information is not null, performing secondary commodity filtering on the scenario recommended commodities according to the commodity filtering rule information to obtain the target recommended commodities; When the commodity filtering rule information is null, recording the scenario recommended commodities as the target recommended commodities.

5. The commodity information push method according to claim 1, characterized in that, Determining a target server from the primary server and all the backup servers, includes: Detecting whether the primary server is in a normal service state; the virtual IP is carried on the primary server; When the primary server is in a normal service state, recording the primary server as the target server; When the primary server is not in a normal service state, transferring the virtual IP to any backup server in a normal service state, and recording the backup server carrying the virtual IP as the target server.

6. An intermediate service platform applying the commodity information push method according to claim 1, characterized in that, The intermediate service platform includes a main server and at least one backup server, and includes: A recommendation instruction receiving module, configured to receive a product recommendation request sent by a preset service platform; the product recommendation request includes scenario invocation information; A server selection module, configured to determine a target server from the main server and all the backup servers; a virtual IP is carried on the target server; A recommendation information acquisition module, configured to send the product recommendation request to the target server through the virtual IP, and determine a target recommendation model and acquire initial recommendation information from a preset model platform by the target server according to the scenario invocation information; the initial recommendation information is generated by the target recommendation model according to the scenario invocation information; the initial recommendation information includes at least one initial recommended product; A recommendation information filtering module, configured to filter products from the initial recommendation information to screen out target recommended products from all the initial recommended products; A recommendation list generation module, configured to generate a product recommendation list according to all the target recommended products, and send the product recommendation list to the preset service platform.

7. A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein, when the processor executes the computer program, the product information push method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium storing a computer program, wherein, when the computer program is executed by a processor, the product information push method according to any one of claims 1 to 5 is implemented.

Images