Resource recommendation method and device, electronic equipment and storage medium

Abstract

The invention provides a resource recommendation method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence, in particular to the technical fields of cloud computing, big data, intelligent search, deep learning and the like. According to the specific implementation scheme, the method comprises the steps of determining object features of a first object according to portrait data of the first object and target behavior data of the first object on an e-commerce platform; determining a target recommendation channel interested by the first object from a plurality of resource recommendation channels of the e-commerce platform according to the object features; determining a to-be-recommended target resource from a plurality of candidate resources according to the object features and article features and article labels of promoted articles in the plurality of candidate resources under the target recommendation channel; and recommending the target resource to the first object.

Description

Technical Field

[0001] The present disclosure relates to the field of AI (Artificial Intelligence), specifically to technical fields such as cloud computing, big data, intelligent search, and deep learning, and in particular to resource recommendation methods, devices, electronic devices, and storage media. Background Art

[0002] With the continuous advancement of Internet technology, especially the continuous development of big data and AI technology, personalized recommendation systems have become an important tool for various online platforms to improve user experience and service quality. These recommendation systems can significantly improve user engagement, satisfaction and retention by deeply understanding user interests and preferences and providing highly customized resource recommendations accordingly. Summary of the invention

[0003] The present disclosure provides a resource recommendation method, device, electronic device and storage medium.

[0004] According to one aspect of the present disclosure, a resource recommendation method is provided, comprising:

[0005] Determining object features of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform;

[0006] Determining, according to the object characteristics, a target recommendation channel of interest to the first object from a plurality of resource recommendation channels of the e-commerce platform;

[0007] Determining a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel;

[0008] The target resource is recommended to the first object.

[0009] According to another aspect of the present disclosure, there is provided a resource recommendation device, comprising:

[0010] A first determination module, configured to determine an object feature of the first object based on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform;

[0011] A second determination module is used to determine, according to the object characteristics, a target recommendation channel that the first object is interested in from a plurality of resource recommendation channels of the e-commerce platform;

[0012] A third determination module is used to determine a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel;

[0013] A recommendation module is used to recommend the target resource to the first object.

[0014] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0015] at least one processor; and

[0016] a memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the resource recommendation method proposed in the above aspect of the present disclosure.

[0018] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium of computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the resource recommendation method proposed in the above aspect of the present disclosure.

[0019] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the resource recommendation method proposed in the above aspect of the present disclosure is implemented.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0022] Figure 1 A schematic diagram of a process flow of a resource recommendation method provided in Embodiment 1 of the present disclosure;

[0023] Figure 2 A flowchart of a resource recommendation method provided in Embodiment 2 of the present disclosure;

[0024] Figure 3 A schematic diagram of a flow chart of a resource recommendation method provided in Embodiment 3 of the present disclosure;

[0025] Figure 4 A flowchart of a resource recommendation method provided in Embodiment 4 of the present disclosure;

[0026] Figure 5 A flowchart of a resource recommendation method provided in Embodiment 5 of the present disclosure;

[0027] Figure 6A schematic diagram of the architecture of the data infrastructure of the e-commerce business provided by the embodiment of the present disclosure;

[0028] Figure 7 A schematic diagram of the application and analysis of e-commerce business data provided by an embodiment of the present disclosure;

[0029] Figure 8 A schematic diagram of the implementation principle of the recommended e-commerce data project provided in the embodiment of the present disclosure;

[0030] Fig. 9 This is a schematic diagram of the structure of a resource recommendation device provided in Embodiment 6 of the present disclosure;

[0031] Fig.10 A schematic block diagram of an example electronic device that can be used to implement embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0032] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0033] The resource recommendation method, device, electronic device, and storage medium of the embodiments of the present disclosure are described below with reference to the accompanying drawings.

[0034] Figure 1 This is a flowchart of the resource recommendation method provided in the first embodiment of the present disclosure.

[0035] The embodiment of the present disclosure is illustrated by taking the resource recommendation method configured in a resource recommendation device as an example. The resource recommendation device can be applied to any electronic device so that the electronic device can perform a resource recommendation function.

[0036] Among them, the electronic device can be any device with computing capabilities, such as a server, a personal computer, a mobile terminal, etc. The mobile terminal can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.

[0037] like Figure 1 As shown, the resource recommendation method may include the following steps S101 to S104:

[0038] Step S101, determining the object characteristics of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform.

[0039] The first object may be any user object using the e-commerce platform. For example, the first object may be a user object currently logged into the e-commerce platform, and the e-commerce platform is in the foreground running state.

[0040] Among them, portrait data is a set of information used to describe the user characteristics and behavior patterns of the first object, including but not limited to: basic attributes (such as demographic characteristics, social attributes, etc.), statistical characteristics (such as the first object’s activity on the e-commerce platform, whether the first object is a new or old user of the e-commerce platform, etc.), group information to which the first object belongs (such as exquisite women, etc.), predictive indicators (for example, when the first object is a new user of the e-commerce platform, indicators predicted by deep learning technology), etc.

[0041] In the disclosed embodiment, feature extraction may be performed on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform to obtain the object features of the first object.

[0042] Step S102: determining a target recommendation channel that the first object is interested in from a variety of resource recommendation channels of the e-commerce platform according to the object characteristics.

[0043] Among them, resource recommendation channels include but are not limited to: short video recommendation channels (or short video sales recommendation channels) and live broadcast room recommendation channels (or live broadcast room sales recommendation channels), etc.

[0044] The number of target recommendation channels may be one or more, and the embodiment of the present disclosure does not limit this.

[0045] As an example, deep learning technology can be used to predict the target recommendation channel that the first object is interested in or prefers from multiple resource recommendation channels of the e-commerce platform based on the object characteristics of the first object.

[0046] For example, if the first subject only browses short videos on the e-commerce platform but does not watch live broadcasts (or crashes after entering the live broadcast room), the target promotion channel may be a short video recommendation channel; if the first subject only watches live broadcasts on the e-commerce platform but does not browse short videos, the target promotion channel may be a live broadcast room recommendation channel; if the first subject both browses short videos and watches live broadcasts on the e-commerce platform, the target promotion channels may include short video recommendation channels and live broadcast room recommendation channels.

[0047] Step S103, determining a target resource to be recommended from multiple candidate resources according to the object characteristics, and according to the item characteristics and item tags of the promoted items in multiple candidate resources under the target recommendation channel.

[0048] Among them, candidate resources include but are not limited to: short videos, live broadcast rooms, etc. Taking the target recommendation channel as the live broadcast room recommendation channel as an example, the candidate resources may include the live broadcast room in the live broadcast state; taking the target recommendation channel as the short video recommendation channel as an example, the candidate resources may include short videos or small videos.

[0049] Among them, promotional items include physical goods, virtual goods and services, etc. For example, promotional items can include online courses, subscription services, digital products, etc.

[0050] In an embodiment of the present disclosure, item features and item tags of promoted items in multiple candidate resources under a target recommendation channel can be obtained, and based on the item features and item tags of the promoted items in the multiple candidate resources and based on the object features of the first object, a target resource of interest to the first object can be determined from the multiple candidate resources.

[0051] As an example, deep learning technology can be used to predict the recommendation probability of multiple candidate resources based on the item features and item tags of the promoted items in multiple candidate resources, as well as the object features, wherein the recommendation probability is used to indicate the first subject's preference for the candidate resources and the promoted items in the candidate resources. Thus, in the present disclosure, the target resource of interest to the first subject can be determined from the multiple candidate resources based on the recommendation probabilities of the multiple candidate resources.

[0052] Exemplarily, a candidate resource whose recommendation probability is higher than a set probability threshold may be used as a target resource.

[0053] Exemplarily, multiple candidate resources can be sorted from large to small according to the recommendation probability to obtain a sorted sequence. Therefore, in the present disclosure, a set number of candidate resources ranked first in the sorted sequence can be used as target resources. Taking the set number as N as an example, the TopN candidate resources in the sorted sequence can be used as target resources.

[0054] Exemplarily, multiple candidate resources may be sorted from large to small according to the recommendation probability to obtain a sorting sequence, and thus in the present disclosure, a set proportion of candidate resources ranked first in the sorting sequence may be used as target resources.

[0055] Step S104: recommending a target resource to the first object.

[0056] Exemplarily, each target resource may be pushed to the account of the first object in the e-commerce platform.

[0057] The resource recommendation method of the disclosed embodiment determines the target recommendation channel that the first object is interested in according to the object characteristics, and recommends resources according to the object characteristics and the item characteristics and item tags of the promotional items in the resources under the target recommendation channel. This can achieve more effective allocation of marketing resources and ensure that resources are invested in the user groups that are most likely to generate conversions. It can not only optimize resource allocation, but also improve the conversion rate of promotional items, ensure that the recommended content is highly consistent with the actual preferences of the first object, reduce the interference of irrelevant information, and enhance the first object's trust and stickiness to the e-commerce platform.

[0058] It should be noted that in the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information are all carried out with the user's consent, and are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0059] In order to clearly explain how any embodiment of the present disclosure determines the object characteristics of the first object based on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform, the present disclosure also proposes a resource recommendation method.

[0060] Figure 2 This is a flowchart of the resource recommendation method provided in the second embodiment of the present disclosure.

[0061] like Figure 2 As shown, the resource recommendation method may include the following steps S201 to S206:

[0062] Step S201, obtain the portrait data of the first object, and obtain the life cycle of the first object on the e-commerce platform.

[0063] It should be noted that the explanation of the portrait data in the aforementioned embodiment is also applicable to this embodiment and will not be repeated here.

[0064] The life cycle is used to indicate the time range of the last continuous interaction between the first object and the e-commerce platform. That is, the life cycle is used to indicate the time range of the e-commerce platform being in the foreground operation state after the first object last logged into the e-commerce platform.

[0065] In any embodiment of the present disclosure, the portrait data of the first object can be obtained by following steps A to C:

[0066] Step A: Acquire an initial label system of the first object according to basic information of the first object.

[0067] Among them, basic information includes but is not limited to: demographic characteristics, social attributes, etc.

[0068] Step B: Based on the historical behavior data of the first object on the e-commerce platform, the initial label system of the first object is updated to obtain a portrait label system of the first object in multiple dimensions.

[0069] The historical behavior data includes, but is not limited to, the purchase record, browsing record, search record, etc. of the first subject on the e-commerce platform, which is used to indicate the first subject's short-term and long-term resource preferences, item preferences, etc.

[0070] Among them, multiple dimensions include but are not limited to the following dimensions: basic attribute dimension, statistical feature dimension, user grouping dimension, predictive indicator dimension, etc.

[0071] In the embodiments of the present disclosure, the historical behavior data of the first object on the e-commerce platform can be processed based on statistical analysis methods and deep learning methods to obtain portrait labels of the first object in multiple dimensions, and the portrait labels of the first object in multiple dimensions can be used to update the initial label system of the first object to obtain a portrait label system of the first object in multiple dimensions.

[0072] It should be noted that, in the related art, different service nodes in a distributed cluster are responsible for different task types. If there are a large number of tasks under the same task type, it may cause the task to be in a queue for a long time and not be executed in time. To address this problem, in any one of the embodiments of the present disclosure, the service nodes in the distributed cluster can be dynamically scheduled to collect the historical behavior data of the first object. Exemplarily, the method for obtaining the historical behavior data of the first object is, for example, based on the load information of each service node in the distributed cluster, determining the target node with the lowest load from each service node, and scheduling the target node to perform the data collection task, so as to collect the initial behavior data associated with the first object from multiple data sources, and converting and splicing the initial behavior data into a format to obtain the historical behavior data.

[0073] For example, in the process of collecting historical behavior data, the collected data may be format converted and spliced ​​while being collected, so as to improve the data collection efficiency and processing efficiency.

[0074] This can not only improve the resource utilization efficiency in the distributed cluster, but also improve the execution efficiency of data collection tasks.

[0075] It should be noted that, considering that different industries in the relevant technologies are often in a relatively isolated state when using the mobile Internet to collect data, that is, a single industry’s understanding of users is limited to its own business scope, and the data obtained only partially reflects the user’s behavioral characteristics within the industry. Moreover, the data between different industries are isolated from each other and cannot be interconnected and shared. In the field of e-commerce business, this data island phenomenon is particularly prominent. For example, the e-commerce industry is closely related to the logistics industry, the financial industry, etc., but because the data is not interoperable, e-commerce companies find it difficult to fully understand the complete behavior of users in the entire consumer chain and cannot accurately grasp user needs, thereby affecting the refined operation of the business and the improvement of service quality. In response to this problem, in any embodiment of the present disclosure, historical behavior data of the first object can be collected from multiple business directions and multiple industries, that is, the historical behavior data of the first object includes but is not limited to at least one of the following:

[0076] The first item is the historical behavior data of the first object in multiple business directions in the e-commerce platform.

[0077] Among them, business directions include but are not limited to: live e-commerce, video e-commerce, digital humans, search and smart shopping guides, etc.

[0078] The second item is the historical behavior data of the first object in multiple target industries; wherein the target industry is an industry associated with the consumption chain of the first object.

[0079] Among them, the target industries include but are not limited to: e-commerce industry, logistics industry, financial industry, etc.

[0080] In summary, collecting the historical behavior data of the first object in different business directions and the historical behavior data of the first object in different industries can improve the integrity and comprehensiveness of the historical behavior data collection, and thus improve the richness and comprehensiveness of the object feature construction.

[0081] Step C: Determine the portrait data of the first object according to the portrait label system.

[0082] As an example, portrait tags of multiple dimensions in the portrait tag system can be used as portrait data of the first object.

[0083] In summary, the initial label system is generated based on the basic information of the first object, which can provide a basic user profile. Through the historical behavior data of the first object on the e-commerce platform, such as purchase records, browsing habits, search keywords, etc., the initial label system can be further updated, which can refine the interest preferences and consumption patterns of the first object, thereby more accurately locating the needs of the first object and improving the first object's usage experience in the resource recommendation scenario.

[0084] Step S202, obtaining target behavior data of the first object on the e-commerce platform during its life cycle.

[0085] In the embodiment of the present disclosure, the target behavior data of the first object in the above life cycle may be collected, wherein the collection method of the target behavior data is similar to the collection method of the historical behavior data, which will not be elaborated here.

[0086] Step S203, determining the object features of the first object according to the portrait data and the target behavior data.

[0087] As an example, feature extraction may be performed on the portrait data and the target behavior data of the first object to obtain the object features of the first object.

[0088] In any embodiment of the present disclosure, the object feature of the first object can be obtained by following steps D to F:

[0089] Step D: Determine historical resources that the first object interacts with during its life cycle based on the target behavior data.

[0090] Among them, historical resources refer to resources on which the first object has performed interactive behaviors (browsing, viewing, sharing, liking, etc.).

[0091] Step E: Obtain the resource tag of the historical resource (referred to as the first resource tag in this disclosure).

[0092] Among them, the first resource label includes but is not limited to: group labels (or customer labels) facing historical resources, regional labels, emotional labels (such as positive energy, inspirational, touching, etc.), etc.

[0093] Step F: Determine the object characteristics of the first object based on the portrait data, the target behavior data and the first resource tag.

[0094] As an example, feature extraction may be performed on the portrait data, the target behavior data, and the first resource tag to obtain object features of the first object.

[0095] In summary, the object features of the first object are generated by integrating the resource tags, portrait data, and target behavior data of the historical resources that have performed interactive behaviors during the life cycle of the first object. This can further improve the richness and comprehensiveness of the object feature construction, thereby improving the accuracy of resource recommendations.

[0096] Step S204: determining a target recommendation channel that the first object is interested in from a variety of resource recommendation channels of the e-commerce platform according to the object characteristics.

[0097] Step S205 , determining a target resource to be recommended from multiple candidate resources according to the object characteristics, and according to the item characteristics and item tags of the promoted items in multiple candidate resources under the target recommendation channel.

[0098] Step S206: recommending a target resource to the first object.

[0099] The explanation of steps S204 to S206 can be found in the relevant description in any embodiment of the present disclosure, and will not be repeated here.

[0100] The resource recommendation method of the disclosed embodiment determines the object characteristics of the first object based on the target behavior data of the first object during its life cycle in the e-commerce platform. The object characteristics can more accurately capture the instant preferences of the first object. This immediacy helps the e-commerce platform to adjust the recommendation algorithm in a timely manner, thereby providing resources and promotional items that better meet the current needs of the first object, thereby improving the accuracy and efficiency of resource recommendations.

[0101] In order to clearly explain how the target resource to be recommended is determined from multiple candidate resources based on object features, and item features and item tags of promoted items in multiple candidate resources in the above embodiment, the present disclosure also proposes a resource recommendation method.

[0102] Figure 3 This is a flowchart of the resource recommendation method provided in Embodiment 3 of the present disclosure.

[0103] like Figure 3 As shown, the resource recommendation method may include the following steps S301 to S309:

[0104] Step S301, determining the object characteristics of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform.

[0105] Step S302: determining a target recommendation channel that the first object is interested in from a variety of resource recommendation channels of the e-commerce platform according to the object characteristics.

[0106] For explanations of steps S301 to S302, reference may be made to the relevant descriptions in any embodiment of the present disclosure and will not be repeated here.

[0107] Step S303: determining the LTV (Lifetime Value) of the first object based on the historical behavior data of the first object on the e-commerce platform.

[0108] It should be noted that the explanation of the historical behavior data in the aforementioned embodiment is also applicable to this embodiment and will not be elaborated here.

[0109] In the embodiment of the present disclosure, the LTV of the first object may be calculated based on the historical behavior data of the first object.

[0110] As an example, the core indicators such as the value of each order, annual purchase frequency, and repurchase rate of the first object on the e-commerce platform can be determined based on the historical behavior data of the first object, and the LTV of the first object can be calculated based on the above core indicators. Among them, LTV is positively correlated with the above core indicators.

[0111] Step S304: acquiring an item screening condition that matches the first object according to the LTV of the first object.

[0112] Among them, the item screening conditions include but are not limited to: price screening range, credit screening range, etc.

[0113] In an embodiment of the present disclosure, the association relationship between different item filtering conditions and LTV intervals can be pre-configured, so that in the present disclosure, the above-mentioned relationship association can be queried based on the LTV of the first object to determine the item filtering condition adapted to the first object, wherein the LTV of the first object is within the LTV interval associated with the item filtering condition.

[0114] Step S305 , based on the item screening condition, multiple resources under the target recommendation channel are screened to obtain multiple candidate resources; wherein the promoted items in the candidate resources meet the item screening condition.

[0115] In the disclosed embodiment, based on the above-mentioned item screening conditions, the massive resources under the target recommendation channel can be screened to obtain a plurality of candidate resources, wherein the promotional items in the candidate resources meet the item screening conditions.

[0116] Step S306: Acquire item features of promoted items from a plurality of candidate resources.

[0117] In any embodiment of the present disclosure, the item features of a promotion item in any candidate resource can be obtained by following steps G to I:

[0118] Step G: Query the knowledge graph associated with the promoted items in multiple candidate resources; wherein the nodes in the knowledge graph are used to indicate the promoted items, and the edges are used to indicate the association relationship between the nodes.

[0119] The knowledge graph is constructed based on the item information of all promoted items on the e-commerce platform. The nodes in the knowledge graph are used to indicate the promoted items, and the edges are used to indicate the association between different nodes. Taking the promoted items as an example, the knowledge graph can also be called the commodity knowledge graph.

[0120] Step H: For any candidate resource among the multiple candidate resources, extract a sub-graph associated with the promoted items in the candidate resource from the knowledge graph.

[0121] The sub-graph includes the promotional items in the candidate resource and other promotional items associated with the promotional items in the candidate resource.

[0122] Step I: Based on the feature extraction algorithm, feature extraction is performed on the above sub-graph to obtain the item features of the promoted items in the candidate resource.

[0123] In summary, by extracting sub-graphs, we can capture various information related to the promotional items, including direct and indirect associations. This correlation analysis helps to reveal the intrinsic connections between promotional items, such as substitutes, complements or competitive relationships, thereby providing richer, more diverse and accurate product features. It should be understood that with the addition of new promotional items, the knowledge graph and sub-graph can be continuously updated and improved, which enables the feature extraction process to keep pace with the times and adapt to the ever-changing market environment.

[0124] Step S307, obtaining item tags of promoted items from multiple candidate resources.

[0125] In any embodiment of the present disclosure, the item label of each promotional item may be manually labeled, or each promotional item may be automatically labeled using an algorithm, and the present disclosure does not limit this.

[0126] Exemplarily, the item tag of each promotional item includes but is not limited to at least one of the following:

[0127] The first item is the industry label of the industry to which the promotional item belongs; the industry label is used to indicate whether promotional items within the industry are allowed to be recommended repeatedly.

[0128] The second item is the category label of the category to which the promoted item belongs; wherein the category label is used to indicate whether the promoted items under the category are allowed to be recommended repeatedly.

[0129] The third item is a weight label of the category to which the promotional item belongs; wherein the weight label is used to indicate the probability of repeated consumption of the promotional item in the category by various objects (such as user objects).

[0130] The fourth item is a time tag of the category to which the promotional item belongs; wherein the time tag is used to indicate the cycle (or referred to as repurchase cycle) during which the promotional item in the category is repeatedly consumed by each object.

[0131] For example, the repurchase period for dairy products may be 2 weeks, and the repurchase period for toothbrushes may be 3 months, etc.

[0132] The fifth item is the group label (or customer label) that the promotional item is targeted at.

[0133] It is understandable that if a user still receives multiple promotional messages about the same or similar items after completing a purchase, it may cause the user to feel troubled or disgusted, thereby reducing user satisfaction and loyalty. At the same time, this may also waste the company's marketing resources and reduce marketing efficiency. In response to this situation, in this disclosure, the item tags of the promoted items are introduced in the resource recommendation scenario, which can reduce the probability of repeated exposure after the user places an order and improve the user's experience in the resource recommendation scenario.

[0134] Step S308, determining a target resource to be recommended from the multiple candidate resources according to the object characteristics, and according to the item characteristics and item tags of the promoted items in the multiple candidate resources.

[0135] As an example, deep learning technology can be used to predict the recommendation probability of multiple candidate resources based on the item features and item tags of the promoted items in multiple candidate resources and the object features of the first object, wherein the recommendation probability is used to indicate the first object's preference for the candidate resources and the promoted items in the candidate resources. Therefore, in the present disclosure, a target resource can be determined from multiple candidate resources based on the recommendation probabilities of multiple candidate resources.

[0136] Step S309: recommending target resources to the first object.

[0137] For the explanation of step S309, reference may be made to the relevant description in any embodiment of the present disclosure, and no further details will be given here.

[0138] The resource recommendation method of the disclosed embodiment can recommend resources to each user in a targeted manner according to the life cycle value of each user, thereby improving the personalization and accuracy of resource recommendations.

[0139] In order to clearly explain how any embodiment of the present disclosure determines a target resource to be recommended from multiple candidate resources based on object features, and based on item features and item tags of promoted items in multiple candidate resources, the present disclosure also proposes a resource recommendation method.

[0140] Figure 4 This is a flowchart of the resource recommendation method provided in Embodiment 4 of the present disclosure.

[0141] like Figure 4 As shown, the resource recommendation method may include the following steps S401 to S407:

[0142] Step S401, determining the object characteristics of the first object based on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform.

[0143] Step S402: determining a target recommendation channel that the first object is interested in from a variety of resource recommendation channels of the e-commerce platform according to the object characteristics.

[0144] For explanations of steps S401 to S402, reference may be made to the relevant descriptions in any embodiment of the present disclosure, and will not be repeated here.

[0145] Step S403, determining a target recommendation strategy from the resource recommendation strategies according to the quality score and confidence of at least one resource recommendation strategy; wherein the confidence is used to indicate the credibility of the quality score of the resource recommendation strategy.

[0146] Among them, the quality score and confidence of each resource recommendation strategy are obtained by experimenting, testing or evaluating each resource recommendation model.

[0147] In the embodiment of the present disclosure, a target recommendation strategy can be determined from each resource recommendation strategy according to the quality score and confidence of each resource recommendation strategy. For example, a resource recommendation strategy with a relatively high quality score and a relatively high confidence can be used as the target recommendation strategy.

[0148] In any embodiment of the present disclosure, the quality score and confidence of each resource recommendation strategy can be determined by following steps J to L:

[0149] Step J: Based on the business needs associated with multiple business directions in the e-commerce platform, obtain the standard data caliber and standard indicator caliber of the first business indicator adapted to each business direction.

[0150] The business directions include but are not limited to: live e-commerce, video e-commerce, digital human, search and smart shopping guide, etc. The business needs associated with different business directions may be different, and the business indicators adapted to different business directions (referred to as the first business indicators in this disclosure) may also be different.

[0151] Among them, the first business indicator includes but is not limited to: CTCVR (Click-Through & Conversion Rate, which is the product of CTR (Click-Through Rate) and CVR (Conversion Rate), CTR, CVR, retention rate and other business indicators.

[0152] Among them, standard data caliber refers to the standard definition and specification of data items during the data collection stage, including but not limited to data source, data type, time range, geographical area, etc. It ensures that all relevant parties have a consistent understanding of the same data item.

[0153] Among them, the standard indicator caliber refers to the statistics or metrics defined for specific business needs, including the calculation formula, parameter settings, data processing details, units, upper and lower limits, etc. of the first business indicator.

[0154] Step K: In the process of recommending resources to objects in multiple groups based on any resource recommendation strategy, monitoring information of each group is obtained; wherein the monitoring information includes a first indicator value of a first business indicator corresponding to the group monitored based on a standard data caliber and a standard indicator caliber.

[0155] Among them, the number of objects in each group is consistent.

[0156] In the present disclosure, for any resource recommendation strategy, in the process of recommending resources to objects in multiple groups (such as test users or experimental users) using a resource recommendation model adapted to the resource recommendation strategy, monitoring information of each group can be obtained, wherein the monitoring information of each group includes the first indicator value of the first business indicator corresponding to each group obtained by monitoring based on the standard data caliber and standard indicator caliber of the first business indicator adapted to each business direction.

[0157] For example, the number of groups is M, and the first service indicator includes CTCVR. Within a specified period of time, resource recommendation strategy 1 is adopted to recommend M groups (A1, A2, ..., A M ) in the process of recommending resources to each object, the number of each group (A1, A2, ..., A M ) in the specified time period; and, in the specified time period, adopt resource recommendation strategy 2 to recommend M groups (respectively B1, B2, …, B M ) in the process of recommending resources to each object in the group (B1, B2, ..., B M )CTCVR value within a specified period of time.

[0158] Step L: Determine the quality score and confidence of any resource recommendation strategy based on the monitoring information of multiple groups.

[0159] In the embodiment of the present disclosure, the quality score and confidence of the resource recommendation strategy may be determined based on monitoring information of multiple groups.

[0160] Still taking the above example as an example, for resource recommendation strategy 1, each group (A1, A2, …, A M ) in a specified period of time, as the quality score of resource recommendation strategy 1. For resource recommendation strategy 2, each group (B1, B2, …, B M) The average, cumulative value or weighted sum of the CTCVR values ​​within the specified time period is used as the quality score of resource recommendation strategy 2.

[0161] And, A i (i=1,2,…,10) The CTCVR value and B in the specified period i The CTCVR values ​​within the specified period are compared. If A i The CTCVR value during the specified period is greater than B i The CTCVR value within the specified time period increases the confidence score of resource recommendation strategy 1. i The CTCVR value during the specified period is less than B i The CTCVR value within the specified time period increases the confidence score of resource recommendation strategy 2.

[0162] For example, if the CTCVR values ​​of 6 groups in resource recommendation strategy 1 are higher than those in resource recommendation strategy 2, and the CTCVR values ​​of 4 groups are lower than those in resource recommendation strategy 2, then the confidence of resource recommendation strategy 1 is higher than that of resource recommendation strategy 2.

[0163] It can be understood that even if the indicator value of the first business indicator in a certain group of a resource recommendation strategy is abnormally high (for example, a certain object places 200 orders within a specified period of time, while other objects only place 1 or 2 orders), and the indicator values ​​of the first business indicators in other groups are all low, according to the above calculation method, although the calculated quality score of the resource recommendation strategy is relatively high, the calculated confidence level of the resource recommendation strategy is relatively low, and thus the resource recommendation strategy will not be used as the target recommendation strategy for resource recommendation in the end, which can reduce the impact of outliers, optimize resource allocation, and improve the conversion rate of promoted items.

[0164] In any embodiment of the present disclosure, after obtaining the monitoring information of multiple groups corresponding to each resource recommendation strategy (including the first indicator value of the first business indicator adapted for each business direction), a business report for each business direction can be generated according to the first indicator value of the first business indicator adapted for each business direction, and the business report for each business direction can be stored; wherein, the business report is used for data access to update the resource recommendation strategy.

[0165] As a result, business personnel or technical personnel in each business direction can access business reports in a timely manner, and update or optimize resource recommendation strategies based on the indicator values ​​of various business indicators under the corresponding business direction in the business report. This can not only meet the data needs of different business directions, but also improve the accuracy of resource recommendations.

[0166] Step S404: query a resource recommendation model that is compatible with the target recommendation strategy.

[0167] Each resource recommendation strategy has an associated (or adapted) resource recommendation model, and each resource recommendation model is trained using the associated resource recommendation strategy.

[0168] Step S405 , using a resource recommendation model to predict the recommendation probability of multiple candidate resources according to the object characteristics, the first resource characteristics of multiple candidate resources under the target recommendation channel, and the item characteristics and item tags of the promoted items in the multiple candidate resources.

[0169] The recommendation probability is used to indicate the first subject's preference for the candidate resources and the promotional items in the candidate resources.

[0170] The first resource feature is a resource feature obtained by extracting features from the candidate resource.

[0171] In the disclosed embodiment, object features, first resource features of multiple candidate resources under a target recommendation channel, item features and item tags of promoted items in the multiple candidate resources may be input into a resource recommendation model, and the resource recommendation model may output recommendation probabilities of the multiple candidate resources.

[0172] That is, for any candidate resource, the resource recommendation model can predict the recommendation probability of the candidate resource based on the object features, the first resource features of the candidate resource, the item features and item tags of the promoted items in the candidate resource.

[0173] Step S406: determining a target resource from the multiple candidate resources according to the recommendation probabilities of the multiple candidate resources.

[0174] Exemplarily, a candidate resource whose recommendation probability is higher than a set probability threshold may be used as a target resource.

[0175] Exemplarily, multiple candidate resources can be sorted from large to small according to the recommendation probability to obtain a sorted sequence. Therefore, in the present disclosure, a set number of candidate resources ranked first in the sorted sequence can be used as target resources. Taking the set number as N as an example, the TopN candidate resources in the sorted sequence can be used as target resources.

[0176] Exemplarily, multiple candidate resources may be sorted from large to small according to the recommendation probability to obtain a sorting sequence, and thus in the present disclosure, a set proportion of candidate resources ranked first in the sorting sequence may be used as target resources.

[0177] Step S407: recommending the target resource to the first object.

[0178] For explanation of step S407, reference may be made to the relevant description in any embodiment of the present disclosure, and will not be repeated here.

[0179] In any embodiment of the present disclosure, in the process of recommending resources based on a resource recommendation model adapted to a target recommendation strategy, the indicator value of the first business indicator (referred to as the second indicator value in the present disclosure) can be monitored in real time based on the standard data caliber and standard indicator caliber of the first business indicator adapted to multiple business directions in the e-commerce platform, and based on the second indicator value of each first business indicator, it can be determined whether there is a target business indicator in each first business indicator, wherein, when the target business indicator is a positive indicator, the indicator value of the target business indicator is lower than the indicator threshold associated with the target business indicator, and when the target business indicator is a negative indicator, the indicator value of the target business indicator is higher than the indicator threshold associated with the target business indicator.

[0180] If the target business indicator exists, an alarm message is generated according to the target business indicator and the second indicator value of the target business indicator, and at least one of the following is performed:

[0181] The first item is to display the alarm information.

[0182] The second item is to broadcast alarm information.

[0183] The third item is sending alarm information. For example, the alarm information can be sent to relevant personnel in the business direction adapted to the target business indicator, so that the relevant personnel can quickly locate the cause of the abnormality based on the alarm information and take targeted measures to make improvements.

[0184] In summary, in the case of abnormal business indicators, timely alarms can not only enable relevant personnel to quickly locate the cause of the abnormality based on the alarm information, so as to take targeted measures to improve and promote continuous improvement and optimization of the business, but also, abnormal business indicators may mean that there are problems with the recommended resource content or the content of the promotional items, the resource recommendation channels or the target audience positioning. After timely alarms, relevant personnel can review and optimize marketing strategies and resource recommendation strategies, etc., to improve the attractiveness and conversion rate of recommended promotional items.

[0185] The resource recommendation method of the embodiment of the present disclosure selects the optimal target recommendation strategy from various resource recommendation strategies based on the quality score and confidence of at least one resource recommendation strategy, and adopts a resource recommendation model adapted to the target recommendation strategy to perform resource recommendation, which can further improve the accuracy of resource recommendation.

[0186] In order to clearly illustrate any embodiment of the present disclosure, the present disclosure also proposes a resource recommendation method.

[0187] Figure 5This is a flowchart of the resource recommendation method provided in Embodiment 5 of the present disclosure.

[0188] like Figure 5 As shown, the resource recommendation method may include the following steps S501 to S506:

[0189] Step S501, determining the object characteristics of the first object based on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform.

[0190] Step S502: determining a target recommendation channel that the first object is interested in from a variety of resource recommendation channels of the e-commerce platform according to the object characteristics.

[0191] Step S503, determining a target resource to be recommended from multiple candidate resources according to the object characteristics, and according to the item characteristics and item tags of the promoted items in multiple candidate resources under the target recommendation channel.

[0192] Step S504: recommending a target resource to the first object.

[0193] For explanations of steps S501 to S504, reference may be made to the relevant descriptions in any embodiment of the present disclosure, and will not be repeated here.

[0194] In any embodiment of the present disclosure, when the target resource includes a live broadcast room, the indicator value of the second business indicator associated with the live broadcast room can also be determined based on the real-time live broadcast data in the live broadcast room, which is recorded as the third indicator value in the present disclosure, and the third indicator value of the second business indicator is sent to the anchor end of the live broadcast room, so that the anchor end can visualize the third indicator value of the second business indicator.

[0195] Among them, the second business indicators include but are not limited to atomic indicators and composite indicators. By way of example, the second business indicators include positive business indicators and negative business indicators. Among them, the positive business indicators include: real-time viewing number, cumulative viewing number, cumulative number of completed orders, cumulative number of transactions, number of new fans, number of likes in the past 5 minutes, number of comments in the past 5 minutes, etc.; negative business indicators include exit rate, unfollow rate, proportion of invalid comments, payment failure rate, loading failure rate, etc.

[0196] As a result, the anchor and the operation team can instantly understand the activity of the live broadcast room and the audience participation. When abnormal data or a decrease in audience participation is found, the operation team can quickly adjust the live broadcast strategy, such as changing the live broadcast content, adding interactive links, etc., to improve audience satisfaction and stickiness, improve the overall atmosphere and activity of the live broadcast room, and improve the live broadcast effect.

[0197] In any embodiment of the present disclosure, when the target resource includes a live broadcast room, and the number of live broadcast rooms is multiple, the following steps O to R may also be performed:

[0198] Step O: Determine a popular live broadcast room from the multiple live broadcast rooms based on the third indicator value of the second business indicator associated with the multiple live broadcast rooms.

[0199] As an example, the popularity value of each live broadcast room can be calculated based on the third indicator value of the second business indicator associated with each live broadcast room, and the popular live broadcast room can be determined from the multiple live broadcast rooms based on the popularity values ​​of the multiple live broadcast rooms.

[0200] Among them, the heat value is positively correlated with the third indicator value of the positive second business indicator, and the heat value is negatively correlated with the third indicator value of the negative second business indicator.

[0201] Exemplarily, the live broadcast room with the highest popularity value can be regarded as a popular live broadcast room.

[0202] Exemplarily, a live broadcast room whose popularity value is higher than a set popularity threshold can be regarded as a popular live broadcast room.

[0203] Step P: Obtain resource characteristics (referred to as second resource characteristics in this disclosure) and resource tags (referred to as second resource tags in this disclosure) of the popular live broadcast room, and obtain item characteristics and item tags of promoted items in the popular live broadcast room.

[0204] Step Q: Determine a third object from multiple candidate objects based on the second resource features and the second resource tags of the popular live broadcast room, and the item features and item tags of the promoted items in the popular live broadcast room.

[0205] As an example, a third object that is interested in the popular live broadcast room can be predicted from among multiple candidate objects based on the object features of multiple candidate objects, the second resource features and second resource tags of the popular live broadcast room, and the item features and item tags of the promoted items in the popular live broadcast room. The implementation principle is similar to step S103 and will not be elaborated here.

[0206] Step R: Recommend popular live broadcast rooms to the third party.

[0207] Exemplarily, popular live broadcast rooms may be pushed to the third party's account on the e-commerce platform.

[0208] In summary, by analyzing the resource features and resource tags of popular live broadcast rooms, as well as the item features and item tags of promoted items, we can more accurately match the interests and needs of third parties, thereby achieving personalized resource recommendations. In addition, accurate resource recommendations can reduce the time users spend on screening and searching, and improve user satisfaction and experience.

[0209] Step S505, obtaining a second object that belongs to the same group as the first object; wherein the object features of the objects in the same group are similar.

[0210] As an example, object features of different objects can be clustered based on a clustering algorithm to obtain multiple groups, where the object features of each object in the same group are similar. Therefore, in this disclosure, each object belonging to the same group as the first object can be queried, which is recorded as the second object in this disclosure.

[0211] Exemplarily, the second object may be a user object that is currently logged into the e-commerce platform and the e-commerce platform is in the foreground running state.

[0212] Step S506: recommend the target resource to the second object.

[0213] Exemplarily, each target resource may be pushed to the account of the second object in the e-commerce platform.

[0214] The resource recommendation method of the embodiment of the present disclosure takes into account that the overhead of computing resources is very huge if personalized resource recommendations are implemented individually for a large number of users. Therefore, in the present disclosure, the same resources can be recommended to homogeneous groups of people, which not only reduces the overhead and complexity of computing resources, but also improves the efficiency of resource recommendation and user experience.

[0215] It should be noted that at a time when e-commerce business is booming, traditional data infrastructure methods have exposed many thorny problems, such as the lack of attention to data infrastructure, the independent operations of various links in the data system, and the existence of data silos between industries, which have brought many obstacles to the efficient development and continuous expansion of e-commerce business.

[0216] In the related technologies, the solutions for data circulation system mainly include the following two types:

[0217] The first is the industry data construction system: in product research and development, the front-end and back-end R&D personnel each perform log collection according to their own product needs. The data R&D personnel collect and clean the logs and maintain the database, and build a data warehouse to manage the data. Finally, the data analysis, strategy and other R&D personnel deliver data visualization, data analysis reports, strategy models and other outputs based on the data warehouse.

[0218] The second type is a general data construction system: by interacting with upstream data from multiple industries, common attributes between industries are extracted to extract, store, manage, and apply data, and various general indicators are provided to the downstream.

[0219] The first solution mentioned above is a deep mining of industry data, which can bring more refined data indicators and analysis reports to the business. Its disadvantages include: the coordination of various links in the data circulation system is not high, resulting in unnecessary losses; R&D personnel are highly dependent on their respective industries, which is not suitable for horizontal support of multi-industry businesses and has low personnel efficiency; there are data islands between different industry data, which is not suitable for data fusion and cross-analysis.

[0220] The second solution mentioned above is a horizontal data integration of multiple industries, but due to the different focuses of different industries, for example, the service industry pays more attention to the user's conversion effect, and the community industry pays more attention to user retention, duration and other indicators. Excessive pursuit of common ground without considering differences can bring little help to various industries.

[0221] In response to the technical problems existing in the above-mentioned related technologies, the present disclosure provides a data infrastructure solution applicable to e-commerce business, covering the formatting standards of multi-dimensional heterogeneous data, metadata management system, data warehouse construction, and general data BI (Business Intelligence) model architecture, etc., bringing new vitality and opportunities to the data processing and operation management of e-commerce business. This data infrastructure solution can be applied to the data infrastructure construction of multiple industries, and can help various industries to obtain their current development status in a timely and stable manner, and accurately guide the direction of business development through rigorous data analysis.

[0222] For example, the data infrastructure architecture of the e-commerce business can be as follows: Figure 6 shown.

[0223] The data infrastructure solution for e-commerce business provided by the present disclosure has built a set of data middle-end systems commonly used in the e-commerce field, which can quickly and accurately realize the infrastructure construction of the e-commerce business data circulation system, and can be reused in the e-commerce field at a low cost, thereby increasing the data dimension, improving the efficiency of R&D personnel, and providing more data and human support for mining the deep value of e-commerce data. For example, the application and analysis of e-commerce business data can be as follows: Figure 7 shown.

[0224] Exemplarily, the data center system mainly includes the following parts:

[0225] Part 1: Recommended e-commerce data projects.

[0226] The current resource recommendation strategy faces multiple challenges, including insufficient accuracy of training samples, disconnection between data labels and actual conditions, insufficient timeliness of training samples, lack of feature dimensions, slow model updates, and imperfect experimental feedback mechanisms. To solve these problems, big data-related technologies can be introduced to comprehensively improve data governance and algorithm levels. For example, the implementation principle of recommending e-commerce data projects is as follows: Figure 8 As shown, it mainly includes the following sub-parts:

[0227] 1. Optimization of real-time full-link data splicing.

[0228] Build from a new starting point: Use log data and original labels as the cornerstone to reshape the processing flow of new training samples.

[0229] Improved efficiency: Optimized processing logic (such as collecting and stitching data at the same time), adjusted task priorities, and significantly reduced the generation cycle of training samples.

[0230] Quality control: Establish a multi-dimensional monitoring system for training samples to ensure the accuracy of training samples under high-speed processing.

[0231] 2. Resource utilization and cluster stability management.

[0232] Fine-grained resource allocation: Perform multiple version iterations on resources, finely control resources, increase quantity without adding resources, and save a lot of resource costs.

[0233] Enhanced cluster stability: Intelligent task scheduling ensures stable operation of the cluster during peak hours.

[0234] 3. Immersive full link.

[0235] Migration pre-assessment: Comprehensively evaluate the live link to confirm the feasibility of low-cost and efficient migration to an immersive development environment.

[0236] Efficient reuse: Iterate cognition, reuse real-life scenario strategies, and efficiently deliver new training sample outputs.

[0237] 4. Build data timeliness.

[0238] Improved real-time data statistics capabilities: Complete multi-dimensional real-time live data statistics of the live broadcast room, including the number of followers, user flow rate, average viewing time, etc., and achieve rapid refresh of live broadcast data at multiple time granularities such as 1 minute, 5 minutes, and 1 hour, providing real-time and accurate data support for live broadcast operations.

[0239] 5. The timeliness of features is enhanced.

[0240] Real-time capture of user behavior: Focus on the instant collection and accumulation of live e-commerce data to accurately grasp the dynamic changes in user behavior.

[0241] Multi-scenario data fusion: Integrate short videos (or small videos) and live broadcast data to enrich the user behavior feature library and improve the timeliness and breadth of user behavior features.

[0242] Real-time data optimization on the business side: Update CTCVR data in real time to significantly improve the predictive performance of the resource recommendation model.

[0243] 6. Optimize model timeliness.

[0244] Innovation in training mode: The training method of the resource recommendation model is changed from sequence+merge (combining sequence learning and merge operations) to pure sequence mode, thereby reducing the model update cycle from 8 hours to 6 hours, greatly improving the response speed of the model.

[0245] Sample update: Updating the attribution samples not only enhances the interpretability of the model, but also improves its ability to adapt to the rapid changes in the live broadcast environment.

[0246] Among them, attribution samples refer to data samples associated with specific results or behaviors used to train models. In the live broadcast environment, these samples can include user behavior data, interaction data, purchase data, etc.

[0247] The second part is to search for e-commerce data.

[0248] At present, the search e-commerce business has encountered a series of serious problems in data construction. On the one hand, the data points are unclear, resulting in uneven quality of data sources and difficulty in accurately obtaining key information. On the other hand, the extremely fast business iteration speed has caused confusion in indicator management and it is impossible to form a unified and effective data evaluation standard. At the same time, business needs show a high degree of customization and personalization. The existing data construction model is difficult to meet the diverse data application scenarios. These problems have seriously hindered the development path of the search e-commerce business based on data-driven, affecting the timeliness and accuracy of business decisions.

[0249] In order to break this dilemma, it is urgent to carry out comprehensive search e-commerce data construction, introduce advanced data management technologies and concepts, reshape the data system, and provide solid data support for the development of e-commerce business. It mainly includes the following sub-parts:

[0250] 1. Data infrastructure.

[0251] Adapt and sort out upstream data points and data caliber changes. Unify indicator calibers, complete data accuracy verification, and align with standard data calibers.

[0252] Complete the construction of each theme wide table, such as: card theme, traffic theme, middle page and other wide tables. For example, taking the promotion items as an example, the construction of search wide tables includes: Aladdin card optimization with product iteration, construction of product wide table, Aladdin card wide table, and search traffic wide table. Support the data construction of commodity knowledge graph, complete the access of commodity library data, and develop the base table data such as new middle page, decision page, new industry data, and recommendation officer project.

[0253] 2. Data tool construction.

[0254] Continue to iterate the construction of BI reports to meet the daily general data viewing needs of product / operation personnel. Optimize report display and interaction logic, collect user experience pain points (such as business data with negative sentiment categories), and optimize user experience. Ensure the completion of the construction of Aladdin's general indicator experiment dashboard. Persist in building data tools other than BI to improve tool coverage.

[0255] 3. Indicator collation and monitoring platform.

[0256] Sort out the commonly used data indicators (or business indicators) and their calibers for search e-commerce to ensure data consistency. Implement output delay alarms and abnormal data fluctuation alarms for configured indicators. Consider the reusability in other directions, the boundaries of user use, the timeliness of alarms, and the comprehensiveness of data monitoring (indicators and dimensions of observation).

[0257] 4. Exploration of data value.

[0258] Explore collaborative projects in areas such as training samples, feature mining, knowledge graphs, and resource recommendation strategies.

[0259] By upgrading the version control solution, the impact of differences in the effectiveness of upstream data can be resolved. The output time of wide tables is improved from T+1 19 o'clock to T+1 10 o'clock, meeting the timeliness requirements of downstream data users. Users can view the scope, definition, and SQL (Structured Query Language)-level data retrieval logic of search common indicators, including common indicators under the four modules of [Traffic], [Card], [Product], and [Order].

[0260] The third part is about e-commerce user portraits. By improving the portrait label system of e-commerce users, the coverage and accuracy of portrait labels can be improved to achieve a comprehensive portrayal and accurate insight into e-commerce users. It mainly includes the following sub-parts:

[0261] 1. Improve the portrait label system of e-commerce users, perform primary stratification according to multiple dimensions, and accurately describe user characteristics through the arrangement and combination of labels in each dimension, thereby improving the coverage of core labels in the benchmark industry.

[0262] 2. Collaborate with the Eris portrait platform to enrich portrait label dimensions and application scenarios, and improve the conversion efficiency of e-commerce live broadcasts, shopping malls and other fields.

[0263] 3. Provide comprehensive and accurate portrait labels for merchant support, model feature mining, etc., to help achieve business goals.

[0264] Solution:

[0265] 1) Promote the construction of global_session: Guided by strategic needs, enrich the basic factual data of e-commerce business, and based on this, continuously improve the construction of the portrait label system under the e-commerce business.

[0266] 2) Continue to build behavioral feature labels: including user attribute labels, user value (preference) labels, etc. The constructed labels will be imported into the Eris platform and UMS (Unified Messaging Service) as needed for downstream use.

[0267] 3) Assisting model label construction: Provide feature samples for the construction of some model labels, assist model training, and improve model accuracy and efficiency.

[0268] 4) Perform feature mining and correlation analysis: Based on the existing portrait labels, mine the correlation between features, and assist in model iteration in a goal-oriented manner.

[0269] The fourth part is the host’s real-time large screen.

[0270] With the rapid development of the live broadcast e-commerce industry, the importance of data timeliness and accuracy in business decision-making has become increasingly prominent. In order to solve the timeliness problems encountered by the model in practical applications and further improve business benefits, the real-time large screen project for e-commerce anchors can be implemented, and the project data results can be applied to the timeliness project. It aims to provide merchants and advertisers with more accurate and timely data support by reconstructing the real-time data capabilities of the live broadcast room to optimize operational strategies. It mainly includes the following sub-sections:

[0271] 1. Reconstruction of underlying data architecture: In view of the complexity of upstream data flow of live e-commerce business and the limitations of historical data architecture, comprehensive business indicator verification and underlying data architecture reconstruction can be carried out. By adopting full-link data precision water level monitoring and stateless computing solutions, efficient data flow and accurate presentation in different business scenarios can be ensured.

[0272] 2. Ensure data accuracy and stability: The data source connection ensures 100% accuracy of the log data, reducing the frequency of impact on data timeliness.

[0273] 3. Expand business application scenarios: Based on the upgrade of real-time data capabilities, expand the application functions of scenarios such as user large screens, commercial large screens and central consoles. By optimizing real-time data docking and visualization capabilities, it provides merchants, anchors and advertisers with richer and more intuitive data display and interactive experience.

[0274] Part 5: E-commerce business data insight analysis.

[0275] Through in-depth data mining and analysis, we provide strong support for business decisions and promote the sustainable growth of e-commerce business. It mainly includes the following sub-sections:

[0276] 1. Homogeneous population analysis: Use data mining and machine learning algorithms to accurately identify similar groups of live broadcast users, and provide users with personalized recommendations based on these analysis results, thereby improving user stickiness and conversion rate.

[0277] 2. Short video reversal experiment analysis: Through experimental analysis of the impact of short videos on live streaming e-commerce, it can be found that short video e-commerce has a positive effect on live streaming business.

[0278] 3. Analysis of differences in live broadcasts driven by short videos: In-depth data analysis of users who use short videos to drive live broadcasts shows that these users have higher click-through rates and conversion rates than other users. Through detailed analysis of user behavior data, data support can be provided for optimizing resource recommendation strategies.

[0279] 4. Analysis of experimental private domain conversion effects: An experiment was conducted on private domain conversion. The experimental group was significantly better than the control group in terms of order volume and payment orders from added-fan users, proving the effectiveness of the private domain operation strategy.

[0280] 5. Anchor transaction and exposure analysis: Through the analysis of anchor transaction data, it can be found that there is a problem of repeated contact after the user places an order. This paper proposes targeted optimization suggestions.

[0281] 6. Other analyses: In addition to the above-mentioned key analyses, multi-dimensional data analyses can be conducted on e-commerce expansion, non-live streaming users, expert anchors, short video or small video fusion fast scrolling rate, and user repurchase, etc. to fully support business decision-making.

[0282] Through in-depth analysis of data such as anchor transactions and user repurchases, we can find problems in e-commerce business: for example, the proportion of low-quality videos in some industries is high, and the anchor participation in some categories such as jewelry and collectibles is low. Targeted solutions can be proposed for these problems, laying the foundation for continuous improvement of the business.

[0283] For example, a video content review mechanism can be introduced to improve the quality of recommended videos. For example, AI technology can be introduced to automatically detect video quality, such as image clarity, content relevance, etc., and only recommend high-quality videos to users.

[0284] For example, for categories with low participation, special support plans can be provided, providing more recommendation resources and preferential policies to attract more anchors to participate.

[0285] Part 6, data application support.

[0286] Due to the rapid development of e-commerce business and its multi-faceted development, the overall data construction faces complex cross-team collaboration challenges. It not only requires close cooperation with multiple field teams within the e-commerce BU (Business Unit), such as production and operation, strategy, R&D, and strategy, but also requires efficient collaboration with the data middle platform, live broadcast middle platform, feed content distribution platform and other teams within the BG (Business Group). Such a working environment requires the construction of a unified, efficient and widely adaptable data set-report development-maintenance system to support the data needs of each team and promote the sustainable development of the business. It mainly includes the following sub-parts:

[0287] 1. Focus on extensive support for data application construction. This includes unified caliber, unified construction, and uniform cognition; it also requires the construction of business reports covering various business directions to ensure that data can fully cover all BU businesses. Whether it is live e-commerce, video e-commerce, or emerging fields such as digital people, search, and smart shopping guides, a detailed data application system is required to meet the current data needs of various departments.

[0288] 2. Targeted customization of business data. In the process of report construction, we carried out targeted customization by deeply understanding the specific data needs of each department. For example, we created experimental reports for the experimental links of the recommendation e-commerce team to meet their experimental analysis needs in the direction of live e-commerce and video e-commerce. At the same time, we provide reports and data sets for the business characteristics of digital people, search, and smart shopping guides to ensure that they can make timely and reasonable decisions based on their own business conditions.

[0289] 3. Rationally apply various data construction tools such as TDA (Topological Data Analysis) data sets, develop wide tables and SQL templates, and continuously improve the flexibility and scalability of delivered data.

[0290] In summary, the data infrastructure solution for e-commerce business provided by the present invention is conveniently designed and can reuse different data infrastructure solutions to improve labor efficiency and tap into the deep value of data; it sets the direction for products: it improves the efficiency and accuracy of data output, assists in multi-dimensional data analysis, and provides guidance for product positioning and development; it improves the efficiency of R&D: it improves abnormal data monitoring, quickly locates problems, improves R&D efficiency, and ensures product quality; it improves the efficiency of strategies: the construction of the experiment management platform improves the efficiency of model iteration by about 30%, and improves the efficiency of model iteration; unified standard construction reduces the cost of human communication and business understanding.

[0291] In addition, we will build an efficient and easy-to-use data warehouse to standardize the management of full-process data such as user behavior data, business data, third-party data, etc. in various fields of e-commerce business; build a unified and standardized data access and data processing process to improve data output efficiency, realize the universality of the process, and ensure that new vertical data can be quickly accessed in the future, with reasonable resource allocation and high computing efficiency; build a complete and comprehensive big data operation and maintenance platform, including metadata management tools, indicator monitoring and alarm tools, task scheduling tools, etc., to ensure daily operation and maintenance efficiency; create a powerful data analysis BI system, deeply explore the business value of data, export data analysis and data mining capabilities to the outside world, and use data to support business development and iteration; based on a complete and accurate data infrastructure, through data mining, data analysis and other technical methods, explore the implementation of projects in the direction of strategies, models, user growth, business insights, etc., to drive sustainable business development.

[0292] With the above Figures 1 to 5 Corresponding to the resource recommendation method provided in the embodiment, the present disclosure also provides a resource recommendation device. Figures 1 to 5 The resource recommendation method provided in the embodiment corresponds to the resource recommendation method, so the implementation of the resource recommendation method is also applicable to the resource recommendation device provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.

[0293] Fig. 9 This is a structural diagram of a resource recommendation device provided in Embodiment 6 of the present disclosure.

[0294] like Fig. 9 As shown, the resource recommendation device 900 may include: a first determination module 910 , a second determination module 920 , a third determination module 930 and a recommendation module 940 .

[0295] The first determination module 910 is used to determine the object characteristics of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform;

[0296] A second determination module 920 is used to determine a target recommendation channel that the first subject is interested in from a plurality of resource recommendation channels of the e-commerce platform according to the subject characteristics;

[0297] The third determination module 930 is used to determine the target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel;

[0298] The recommendation module 940 is configured to recommend a target resource to the first object.

[0299] In a possible implementation of the embodiments of the present disclosure, the first determination module 910 is used to: obtain portrait data of the first object, and obtain the life cycle of the first object on the e-commerce platform; wherein the life cycle is used to indicate the time range of the first object's most recent continuous interaction with the e-commerce platform; obtain target behavior data of the first object on the e-commerce platform within the life cycle; and determine the object characteristics of the first object based on the portrait data and the target behavior data.

[0300] In a possible implementation of the embodiment of the present disclosure, the first determination module 910 is used to: determine the historical resources that the first object interacts with during its life cycle based on the target behavior data; obtain the first resource tag of the historical resource; and determine the object characteristics of the first object based on the portrait data, the target behavior data and the first resource tag.

[0301] In a possible implementation of the embodiment of the present disclosure, the first determination module 910 is used to: obtain an initial label system of the first object based on the basic information of the first object; update the initial label system based on the historical behavior data of the first object on the e-commerce platform to obtain a portrait label system of the first object in multiple dimensions; and determine the portrait data of the first object based on the portrait label system.

[0302] In a possible implementation of the embodiment of the present disclosure, the first determination module 910 is also used to: determine the target node from each service node based on the load information of each service node in the distributed cluster; schedule the target node to perform a data collection task to collect initial behavior data associated with the first object from multiple data sources, and convert and splice the initial behavior data to obtain historical behavior data.

[0303] In a possible implementation of the embodiment of the present disclosure, the historical behavior data includes at least one of the following: historical behavior data of the first object in multiple business directions in the e-commerce platform; historical behavior data of the first object in multiple target industries; wherein the target industry is an industry associated with the consumption chain of the first object.

[0304] In a possible implementation of the embodiments of the present disclosure, the item tags of the promotional items in the candidate resources include at least one of the following: an industry tag of the industry to which the promotional item belongs; wherein the industry tag is used to indicate whether repeated recommendations are allowed for the promotional items in the industry; a category tag of the category to which the promotional item belongs; wherein the category tag is used to indicate whether repeated recommendations are allowed for the promotional items under the category; a weight tag under the category; wherein the weight tag is used to indicate the probability that the promotional items under the category are repeatedly consumed by each object; a time tag under the category; wherein the time tag is used to indicate the period for which the promotional items under the category are repeatedly consumed by each object; and a group tag for the promotional item.

[0305] In a possible implementation of the embodiments of the present disclosure, the third determination module 930 is used to: query the knowledge graph associated with the promotional items in multiple candidate resources; wherein the nodes in the knowledge graph are used to indicate the promotional items, and the edges are used to indicate the association relationship between the nodes; extract a sub-graph associated with the promotional items in any candidate resource from the knowledge graph; and perform feature extraction on the sub-graph to obtain item features of the promotional items in any candidate resource.

[0306] In a possible implementation of the embodiment of the present disclosure, the third determination module 930 is used to: determine the life cycle value LTV of the first object based on the historical behavior data of the first object on the e-commerce platform; obtain item screening conditions adapted to the first object based on the LTV of the first object; based on the item screening conditions, screen multiple resources under the target recommendation channel to obtain multiple candidate resources; wherein the promotion items in the candidate resources meet the item screening conditions; determine the target resource to be recommended from the multiple candidate resources based on the object characteristics, and based on the item characteristics and item tags of the promotion items in the multiple candidate resources.

[0307] In a possible implementation of the embodiments of the present disclosure, the third determination module 930 is used to: determine a target recommendation strategy from various resource recommendation strategies based on a quality score and a confidence level of at least one resource recommendation strategy; wherein the confidence level is used to indicate the degree of credibility of the quality score of the resource recommendation strategy; query a resource recommendation model adapted to the target recommendation strategy; use the resource recommendation model to predict the recommendation probability of multiple candidate resources based on object features, first resource features of multiple candidate resources, and item features and item tags of promoted items in the multiple candidate resources; and determine a target resource from multiple candidate resources based on the recommendation probabilities of the multiple candidate resources.

[0308] In a possible implementation of the embodiment of the present disclosure, the third determination module 930 is also used to: based on business needs associated with multiple business directions in the e-commerce platform, obtain a standard data caliber and a standard indicator caliber of a first business indicator adapted to each business direction; in the process of recommending resources to objects in multiple groups based on any resource recommendation strategy, obtain monitoring information of each group; wherein the monitoring information includes a first indicator value of the first business indicator corresponding to the group monitored based on the standard data caliber and the standard indicator caliber; and determine the quality score and confidence of any resource recommendation strategy based on the monitoring information of multiple groups.

[0309] In a possible implementation of the embodiment of the present disclosure, the resource recommendation device 900 further includes:

[0310] A first generating module, used to generate a business report for each business direction according to a first indicator value of a first business indicator adapted for each business direction;

[0311] The storage module is used to store business reports for each business direction; wherein the business reports are used for data access to update resource recommendation strategies.

[0312] In a possible implementation of the embodiment of the present disclosure, the resource recommendation device 900 further includes:

[0313] A monitoring module, used for monitoring a second indicator value of the first business indicator in real time based on a standard data caliber and a standard indicator caliber of the first business indicator adapted to multiple business directions in the e-commerce platform during a process of recommending resources based on a target recommendation strategy;

[0314] A second generating module is used to generate alarm information according to the target business indicator and the corresponding second indicator value when there is a target business indicator in each first business indicator; wherein, in response to the target business indicator being a positive indicator, the indicator value of the target business indicator is lower than the indicator threshold associated with the target business indicator, and in response to the target business indicator being a negative indicator, the indicator value of the target business indicator is higher than the indicator threshold associated with the target business indicator;

[0315] The execution module is used to execute at least one of the following: display alarm information; broadcast alarm information; send alarm information.

[0316] In a possible implementation of the embodiment of the present disclosure, the resource recommendation device 900 further includes:

[0317] An acquisition module, used for acquiring a second object belonging to the same group as the first object; wherein the objects in the same group have similar object features;

[0318] The recommendation module 940 is further configured to recommend a target resource to the second object.

[0319] In a possible implementation of the embodiment of the present disclosure, the resource recommendation device 900 further includes:

[0320] a fourth determination module, configured to determine, in response to the target resource including the live broadcast room, a third indicator value of the second business indicator associated with the live broadcast room based on the real-time live broadcast data in the live broadcast room;

[0321] The sending module is used to send the third indicator value of the second business indicator to the anchor end of the live broadcast room, so that the anchor end displays the third indicator value of the second business indicator.

[0322] In a possible implementation of the embodiment of the present disclosure, there are multiple live broadcast rooms, and the resource recommendation device 900 further includes:

[0323] A fifth determination module, configured to determine a popular live broadcast room from the plurality of live broadcast rooms according to a third indicator value of the second business indicator associated with the plurality of live broadcast rooms;

[0324] a sixth determination module, configured to determine a third object from a plurality of candidate objects based on a second resource feature and a second resource tag of the popular live broadcast room, and an item feature and an item tag of a promoted item in the popular live broadcast room;

[0325] The recommendation module 940 is also used to recommend popular live broadcast rooms to the third party.

[0326] The resource recommendation device of the disclosed embodiment determines the target recommendation channel that the first object is interested in according to the object characteristics, and recommends resources according to the object characteristics and the item characteristics and item tags of the promotional items in the resources under the target recommendation channel. This can achieve more effective allocation of marketing resources and ensure that resources are invested in the user groups that are most likely to generate conversions. It can not only optimize resource allocation, but also improve the conversion rate of promotional items, ensure that the recommended content is highly consistent with the actual preferences of the first object, reduce the interference of irrelevant information, and enhance the first object's trust and stickiness to the e-commerce platform.

[0327] In order to implement the above embodiments, the present disclosure also provides an electronic device, which may include at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the resource recommendation method proposed in any of the above embodiments of the present disclosure.

[0328] In order to implement the above embodiments, the present disclosure further provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the resource recommendation method proposed in any of the above embodiments of the present disclosure.

[0329] In order to implement the above embodiments, the present disclosure further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the resource recommendation method proposed in any of the above embodiments of the present disclosure.

[0330] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0331] Fig.10 A schematic block diagram of an example electronic device that can be used to implement embodiments of the present disclosure is shown.

[0332] Among them, the electronic device may include the server and the client in the above-mentioned embodiment. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0333] like Fig.10 As shown, the electronic device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 1002 or a computer program loaded from a storage unit 1007 to a RAM (Random Access Memory) 1003. In the RAM 1003, various programs and data required for the operation of the device 1000 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An I / O (Input / Output) interface 1005 is also connected to the bus 1004.

[0334] A number of components in the device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0335] The computing unit 1001 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a CPU (Central Processing Unit), a GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, a DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 1001 performs the various methods and processes described above, such as the resource recommendation method described above. For example, in some embodiments, the resource recommendation method described above may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the computing unit 1001, one or more steps of the resource recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 1001 may be configured to execute the above resource recommendation method in any other appropriate manner (for example, by means of firmware).

[0336] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general-purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0337] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0338] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0339] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0340] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0341] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services (Virtual Private Server). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0342] It should be noted that artificial intelligence is a discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0343] According to the technical solution of the embodiment of the present disclosure, the target recommendation channel that the first object is interested in is determined according to the object characteristics, and resource recommendations are performed according to the object characteristics and the item characteristics and item tags of the promotional items in the resources under the target recommendation channel. This can achieve more effective allocation of marketing resources and ensure that resources are invested in the user group that is most likely to generate conversions. This can not only optimize resource allocation, but also improve the conversion rate of promotional items, ensure that the recommended content is highly consistent with the actual preferences of the first object, reduce the interference of irrelevant information, and enhance the first object's trust and stickiness to the e-commerce platform.

[0344] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0345] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A resource recommendation method, comprising: Determining object features of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform; Determining, according to the object characteristics, a target recommendation channel of interest to the first object from a plurality of resource recommendation channels of the e-commerce platform; Determining a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel; The target resource is recommended to the first object.

2. The method according to claim 1, wherein: The determining the object feature of the first object according to the portrait data of the first object and the target behavior data of the first object on the e-commerce platform includes: Obtaining the portrait data of the first object, and obtaining the life cycle of the first object on the e-commerce platform; wherein the life cycle is used to indicate the time range of the most recent continuous interaction between the first object and the e-commerce platform; Acquire target behavior data of the first object on the e-commerce platform during the life cycle; The object characteristics of the first object are determined based on the portrait data and the target behavior data.

3. The method according to claim 2, wherein: The determining the object feature of the first object according to the portrait data and the target behavior data includes: Determining historical resources that the first object interacted with during the life cycle according to the target behavior data; Obtaining a first resource tag of the historical resource; Determine object characteristics of the first object based on the portrait data, the target behavior data and the first resource tag.

4. The method according to claim 2, wherein: The acquiring the portrait data of the first object includes: Acquire an initial label system of the first object according to the basic information of the first object; According to the historical behavior data of the first object on the e-commerce platform, the initial label system is updated to obtain a portrait label system of the first object in multiple dimensions; According to the portrait label system, portrait data of the first object is determined.

5. The method according to claim 4, wherein: The historical behavior data is obtained by following the steps below: Determine a target node from each of the service nodes based on load information of each service node in the distributed cluster; The target node is scheduled to execute a data collection task to collect initial behavior data associated with the first object from multiple data sources, and format conversion and splicing are performed on the initial behavior data to obtain the historical behavior data.

6. The method according to claim 4, wherein: The historical behavior data includes at least one of the following: Historical behavior data of the first object in multiple business directions on the e-commerce platform; The historical behavior data of the first object in multiple target industries; wherein the target industry is an industry associated with the consumption chain of the first object.

7. The method according to claim 1, wherein: The item tags of the promoted items in the candidate resources include at least one of the following: The industry label of the industry to which the promotional item belongs; wherein the industry label is used to indicate whether the promotional items in the industry are allowed to be recommended repeatedly; The category label of the category to which the promotional item belongs; wherein the category label is used to indicate whether the promotional item under the category is allowed to be recommended repeatedly; A weight label under the category; wherein the weight label is used to indicate the probability of the promoted item under the category being repeatedly consumed by each object; The time tag under the category; wherein the time tag is used to indicate the period during which the promotional items under the category are repeatedly consumed by each object; The label of the group that the promotional item is targeted at.

8. The method according to claim 1, wherein: The item features of the promoted items in the candidate resources are obtained by using the following steps: Querying a knowledge graph associated with the promotional items in the plurality of candidate resources; wherein the nodes in the knowledge graph are used to indicate the promotional items, and the edges are used to indicate the association relationship between the nodes; Extracting a subgraph associated with a promoted item in any candidate resource from the knowledge graph; Feature extraction is performed on the sub-graph to obtain item features of the promoted items in any of the candidate resources.

9. The method according to claim 1, wherein: The determining a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel includes: Determine the life cycle value LTV of the first object according to the historical behavior data of the first object on the e-commerce platform; According to the LTV of the first object, obtaining an item screening condition adapted to the first object; Based on the item screening condition, multiple resources under the target recommendation channel are screened to obtain the multiple candidate resources; wherein the promoted items in the candidate resources meet the item screening condition; According to the object features, and according to the item features and item tags of the promoted items in the multiple candidate resources, a target resource to be recommended is determined from the multiple candidate resources.

10. The method according to any one of claims 1 to 9, wherein: The determining a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel includes: Determine a target recommendation strategy from each of the resource recommendation strategies according to the quality score and confidence of at least one resource recommendation strategy; wherein the confidence is used to indicate the credibility of the quality score of the resource recommendation strategy; Querying a resource recommendation model that is compatible with the target recommendation strategy; The resource recommendation model is used to predict the recommendation probability of the multiple candidate resources according to the object feature, the first resource feature of the multiple candidate resources, and the item features and item tags of the promoted items in the multiple candidate resources; The target resource is determined from the multiple candidate resources according to the recommendation probabilities of the multiple candidate resources.

11. The method according to claim 10, wherein: The quality score and confidence level of each resource recommendation strategy are determined using the following steps: Based on business needs associated with multiple business directions in the e-commerce platform, obtain a standard data caliber and a standard indicator caliber of a first business indicator adapted to each of the business directions; In the process of recommending resources to objects in multiple groups based on any resource recommendation strategy, monitoring information of each group is obtained; wherein the monitoring information includes a first indicator value of the first business indicator corresponding to the group obtained by monitoring based on the standard data caliber and the standard indicator caliber; The quality score and confidence level of any resource recommendation strategy are determined based on the monitoring information of the multiple groups.

12. The method according to claim 11, wherein: The method further comprises: Generate a business report for each of the business directions according to the first indicator value of the first business indicator adapted for each of the business directions; Business reports for each of the business directions are stored; wherein the business reports are used for data access to update resource recommendation strategies.

13. The method according to claim 10, wherein: The method further comprises: In the process of recommending resources based on the target recommendation strategy, based on the standard data caliber and standard indicator caliber of the first business indicator adapted to multiple business directions in the e-commerce platform, the second indicator value of the first business indicator is monitored in real time; In the case where there is a target business indicator in each of the first business indicators, an alarm message is generated according to the target business indicator and the corresponding second indicator value; wherein, in response to the target business indicator being a positive indicator, the indicator value of the target business indicator is lower than the indicator threshold associated with the target business indicator, and in response to the target business indicator being a negative indicator, the indicator value of the target business indicator is higher than the indicator threshold associated with the target business indicator; Perform at least one of the following: display the alarm information; broadcast the alarm information; send the alarm information.

14. The method according to any one of claims 1 to 9, wherein: The method further comprises: Acquire a second object that belongs to the same group as the first object; wherein the objects in the same group have similar object features; The target resource is recommended to the second object.

15. The method according to any one of claims 1 to 9, wherein: The method further comprises: In response to the target resource including a live broadcast room, determining a third indicator value of a second business indicator associated with the live broadcast room based on real-time live broadcast data in the live broadcast room; Send the third indicator value of the second business indicator to the anchor end of the live broadcast room, so that the anchor end displays the third indicator value of the second business indicator.

16. The method according to claim 15, wherein: The number of the live broadcast rooms is multiple, and the method further includes: Determining a popular live broadcast room from the multiple live broadcast rooms according to a third indicator value of the second business indicator associated with the multiple live broadcast rooms; Determine a third object from multiple candidate objects based on the second resource feature and the second resource tag of the popular live broadcast room, and the item feature and the item tag of the promoted item in the popular live broadcast room; Recommend the popular live broadcast room to the third object.

17. A resource recommendation device, comprising: A first determination module, configured to determine an object feature of the first object based on the portrait data of the first object and the target behavior data of the first object on the e-commerce platform; A second determination module is used to determine, according to the object characteristics, a target recommendation channel that the first object is interested in from a plurality of resource recommendation channels of the e-commerce platform; A third determination module is used to determine a target resource to be recommended from the multiple candidate resources according to the object characteristics and the item characteristics and item tags of the promoted items in the multiple candidate resources under the target recommendation channel; A recommendation module is used to recommend the target resource to the first object.

18. The device according to claim 17, wherein: The first determining module is used to: Obtaining the portrait data of the first object, and obtaining the life cycle of the first object on the e-commerce platform; wherein the life cycle is used to indicate the time range of the most recent continuous interaction between the first object and the e-commerce platform; Acquire target behavior data of the first object on the e-commerce platform during the life cycle; The object characteristics of the first object are determined based on the portrait data and the target behavior data.

19. The device according to claim 18, wherein: The first determining module is used to: Determining historical resources that the first object interacted with during the life cycle according to the target behavior data; Obtaining a first resource tag of the historical resource; Determine object characteristics of the first object based on the portrait data, the target behavior data and the first resource tag.

20. The device according to claim 18, wherein The first determining module is used to: Acquire an initial label system of the first object according to the basic information of the first object; According to the historical behavior data of the first object on the e-commerce platform, the initial label system is updated to obtain a portrait label system of the first object in multiple dimensions; According to the portrait label system, portrait data of the first object is determined.

21. The device according to claim 20, wherein: The first determining module is further used for: Determine a target node from each of the service nodes based on load information of each service node in the distributed cluster; The target node is scheduled to execute a data collection task to collect initial behavior data associated with the first object from multiple data sources, and format conversion and splicing are performed on the initial behavior data to obtain the historical behavior data.

22. The device according to claim 20, wherein: The historical behavior data includes at least one of the following: Historical behavior data of the first object in multiple business directions on the e-commerce platform; The historical behavior data of the first object in multiple target industries; wherein the target industry is an industry associated with the consumption chain of the first object.

23. The device according to claim 17, wherein: The item tags of the promoted items in the candidate resources include at least one of the following: The industry label of the industry to which the promotional item belongs; wherein the industry label is used to indicate whether the promotional items in the industry are allowed to be recommended repeatedly; The category label of the category to which the promotional item belongs; wherein the category label is used to indicate whether the promotional item under the category is allowed to be recommended repeatedly; A weight label under the category; wherein the weight label is used to indicate the probability of the promoted item under the category being repeatedly consumed by each object; The time tag under the category; wherein the time tag is used to indicate the period during which the promotional items under the category are repeatedly consumed by each object; The label of the group that the promotional item is targeted at.

24. The device according to claim 17, wherein: The third determining module is used to: Querying a knowledge graph associated with the promotional items in the plurality of candidate resources; wherein the nodes in the knowledge graph are used to indicate the promotional items, and the edges are used to indicate the association relationship between the nodes; Extracting a subgraph associated with a promoted item in any candidate resource from the knowledge graph; Feature extraction is performed on the sub-graph to obtain item features of the promoted items in any of the candidate resources.

25. The apparatus of claim 17, wherein: The third determining module is used to: Determine the life cycle value LTV of the first object according to the historical behavior data of the first object on the e-commerce platform; According to the LTV of the first object, obtaining an item screening condition adapted to the first object; Based on the item screening condition, multiple resources under the target recommendation channel are screened to obtain the multiple candidate resources; wherein the promoted items in the candidate resources meet the item screening condition; According to the object features, and according to the item features and item tags of the promoted items in the multiple candidate resources, a target resource to be recommended is determined from the multiple candidate resources.

26. The device according to any one of claims 17 to 25, wherein: The third determining module is used to: Determine a target recommendation strategy from each of the resource recommendation strategies according to the quality score and confidence of at least one resource recommendation strategy; wherein the confidence is used to indicate the credibility of the quality score of the resource recommendation strategy; Querying a resource recommendation model that is compatible with the target recommendation strategy; The resource recommendation model is used to predict the recommendation probability of the multiple candidate resources according to the object feature, the first resource feature of the multiple candidate resources, and the item features and item tags of the promoted items in the multiple candidate resources; The target resource is determined from the multiple candidate resources according to the recommendation probabilities of the multiple candidate resources.

27. The device according to claim 26, wherein: The third determining module is further used for: Based on business needs associated with multiple business directions in the e-commerce platform, obtain a standard data caliber and a standard indicator caliber of a first business indicator adapted to each of the business directions; In the process of recommending resources to objects in multiple groups based on any resource recommendation strategy, monitoring information of each group is obtained; wherein the monitoring information includes a first indicator value of the first business indicator corresponding to the group obtained by monitoring based on the standard data caliber and the standard indicator caliber; The quality score and confidence level of any resource recommendation strategy are determined based on the monitoring information of the multiple groups.

28. The device according to claim 27, wherein The device also includes: A first generating module, configured to generate a business report for each of the business directions according to a first indicator value of a first business indicator adapted for each of the business directions; A storage module is used to store business reports for each of the business directions; wherein the business reports are used for data access to update resource recommendation strategies.

29. The device according to claim 26, wherein: The device also includes: A monitoring module, configured to monitor in real time a second indicator value of the first business indicator based on a standard data caliber and a standard indicator caliber of the first business indicator adapted to multiple business directions in the e-commerce platform during a process of recommending resources based on the target recommendation strategy; A second generating module, configured to generate alarm information according to the target business indicator and the corresponding second indicator value when there is a target business indicator in each of the first business indicators; wherein, in response to the target business indicator being a positive indicator, the indicator value of the target business indicator is lower than the indicator threshold associated with the target business indicator, and in response to the target business indicator being a negative indicator, the indicator value of the target business indicator is higher than the indicator threshold associated with the target business indicator; The execution module is used to execute at least one of the following: display the alarm information; broadcast the alarm information; send the alarm information.

30. The device according to any one of claims 17 to 25, wherein: The device also includes: An acquisition module, configured to acquire a second object belonging to the same group as the first object; wherein the objects in the same group have similar object features; The recommendation module is further used to recommend the target resource to the second object.

31. The device according to any one of claims 17 to 25, wherein: The device also includes: a fourth determination module, configured to determine, in response to the target resource including a live broadcast room, a third indicator value of a second business indicator associated with the live broadcast room based on real-time live broadcast data in the live broadcast room; A sending module is used to send the third indicator value of the second business indicator to the anchor end of the live broadcast room, so that the anchor end displays the third indicator value of the second business indicator.

32. The device according to claim 31, wherein The number of the live broadcast rooms is multiple, and the device further includes: A fifth determination module, configured to determine a popular live broadcast room from the multiple live broadcast rooms according to a third indicator value of the second business indicator associated with the multiple live broadcast rooms; A sixth determination module, configured to determine a third object from a plurality of candidate objects based on the second resource feature and the second resource tag of the popular live broadcast room, and the item feature and the item tag of the promoted item in the popular live broadcast room; The recommendation module is also used to recommend the popular live broadcast room to the third object.

33. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the resource recommendation method according to any one of claims 1 to 16.

34. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the resource recommendation method according to any one of claims 1-16.

35. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the steps of the resource recommendation method according to any one of claims 1 to 16.

Images