Candidate object automatic screening method, electronic equipment and computer readable storage medium

Abstract

The invention provides a candidate object automatic screening method, electronic equipment and a computer readable storage medium, and belongs to the technical field of data processing. The method comprises the steps that candidate object data corresponding to candidate objects are obtained from a target data platform, and the candidate objects are obtained based on screening parameters input by a user or determined based on a reference object specified by the user; performing feature extraction on the candidate object data to obtain candidate object feature information; matching the candidate object feature information with the target information to obtain a matching degree score; determining whether the candidate object meets a target requirement based on the matching degree score; when the candidate objects meet the target requirements, the candidate objects are added into an effective object set; and when the number of the effective objects in the effective object set does not reach the preset number, continuing to obtain the candidate object data for screening. The candidate object screening efficiency can be improved, and the accuracy of candidate object determination under the multi-dimensional condition is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an automatic candidate selection method, electronic device, and computer-readable storage medium. Background Technology

[0002] With the development of mobile internet technology and the continuous maturation of the digital marketing industry, brand promotion and product marketing based on internet platforms have become common marketing methods for enterprises. In such marketing activities, enterprises or advertising agencies usually need to screen out suitable partners from a large number of online influencer accounts, and conduct a comprehensive evaluation based on information such as the influencer's content area, persona tags, fan base, audience profile, and historical content performance data to determine the influencer resources suitable for participating in marketing promotion activities.

[0003] In related technologies, the selection process for influencer accounts typically relies heavily on manual methods. Media personnel set target requirements through business platforms, review the detailed data of each searched influencer account, and make judgments and arrangements based on business needs. Because the number of candidate influencer accounts is usually large, and the selection process often requires comprehensive consideration of multiple data dimensions, such as follower size, content performance metrics, audience structure, and collaboration pricing, manual selection is not only inefficient but also prone to judgment bias or omissions when faced with complex combinations of conditions, making it difficult to meet the efficiency and accuracy requirements of actual marketing activities.

[0004] Therefore, in the process of selecting influencer accounts for internet marketing campaigns, the low efficiency of screening and the difficulty in identifying candidates under multi-dimensional conditions have become urgent problems that need to be solved. Summary of the Invention

[0005] This application provides an automatic candidate selection method, electronic device, and computer-readable storage medium to solve the problems of low selection efficiency and difficulty in determining candidates under multi-dimensional conditions in the existing talent account selection process.

[0006] Firstly, this application provides a method for automatically filtering candidate objects, the method comprising: Candidate object data corresponding to the candidate object is obtained from the target data platform; wherein, the candidate object is obtained from the target data platform based on the filtering parameters input by the user, or the candidate object is determined based on a benchmark object, wherein the benchmark object is the object specified by the user; Feature extraction is performed on the candidate object data to obtain the feature information of the candidate object; The feature information of the candidate object is matched with the target information to obtain a matching score; wherein, when the candidate object is determined based on the filtering parameters input by the user, the target information is the filtering parameters; when the candidate object is determined based on the benchmark object, the target information is the feature information of the benchmark object; Based on the matching score, it is determined whether the candidate object meets the target requirements; When the candidate object meets the target requirement, the candidate object is determined as a valid object and added to the valid object set; When the number of valid objects in the set of valid objects does not reach the preset number, the step of obtaining the candidate object data corresponding to the candidate object from the target data platform is repeated until the number of valid objects in the set of valid objects reaches the preset number.

[0007] In one possible design, matching the feature information of the candidate object with the target information to obtain a matching score includes: Construct a candidate object feature vector based on the feature information of the candidate object; Based on the target information, construct a target information vector; Based on the candidate object feature vector and the target information vector, a matching score is calculated; wherein, the matching score is calculated by a weighted function, and the weight parameters in the weighted function are used to characterize the importance of different matching dimensions.

[0008] In one possible design, determining whether the candidate object meets the target requirements based on the matching score includes: Determine whether the matching score is less than a preset scoring threshold; When the matching score is greater than or equal to the preset score threshold, it is determined that the candidate object meets the target requirement; When the matching score is less than the preset score threshold, it is determined that the candidate does not meet the target requirements.

[0009] In one possible design, the step of extracting features from the candidate object data to obtain the feature information of the candidate object includes: Extract the basic attribute data, audience structure data, and content performance data of the candidate objects from the candidate object data; Based on the candidate object's basic attribute data, audience structure data, and content performance data, the feature information of the candidate object is generated. The basic attribute data includes at least one of object identification information, object profile information, and object tag information; the audience structure data includes at least one of fan number data, gender distribution data, age distribution data, and geographic distribution data; and the content performance data includes at least one of readership data, interaction data, interaction rate data, viral article rate data, and completion rate data.

[0010] In one possible design, the audience structure data and content performance data are extracted from the candidate object data, including: The graphical interface elements in the target data platform page are parsed to obtain the audience structure data of the candidate object; wherein, the graphical interface elements include Canvas elements or SVG elements, and the audience structure data includes at least one of gender distribution data, age distribution data, and geographic distribution data; Obtain the historical content data of the candidate object; Based on the historical content data, the content performance data of the candidate objects are determined.

[0011] In one possible design, when the candidate objects are obtained from the target data platform based on user-input filtering parameters, the method further includes: Access the target data platform via a browser-automated execution program; The search criteria are automatically set based on the filtering parameters, and the candidate objects are retrieved based on the search criteria. The step of obtaining candidate object data corresponding to the candidate object from the target data platform includes: Data scraping is performed on the details page corresponding to the candidate object on the target data platform to obtain the candidate object data; wherein, the data scraping includes scraping at least one of text data, numerical data, graphical interface data, and historical content data.

[0012] In one possible design, when the candidate object is determined based on the baseline object, the method further includes: Obtain the baseline object data of the baseline object; Feature extraction is performed on the reference object data to obtain the feature information of the reference object; Based on the feature information of the benchmark object, objects similar to the benchmark object are identified in the target data platform as candidate objects.

[0013] In one possible design, the filtering parameters include at least one of the following: basic information parameters, object tag parameters, object data representation parameters, audience structure parameters, and price range parameters. The basic information parameters include at least one of the following: platform account, project identifier, task name, and number of tasks. The object tag parameters include at least one of object tags, content categories, and keywords; The object data performance parameters include at least one of the following: readership, interaction rate, viral article rate, and completion rate; The audience structure parameters include at least one of the following: fan number range, fan gender ratio, fan age distribution, and fan geographic distribution.

[0014] Secondly, this application provides an automatic candidate screening device, comprising: a module for performing the aforementioned method embodiment of the first aspect.

[0015] Thirdly, this application provides an electronic device, including: a memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method described in the first aspect or various possible designs of the first aspect.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the method described in the first aspect or various possible designs of the first aspect.

[0017] Fifthly, this application provides a computer program product, which includes computer program code that, when run on a computer, causes the computer to implement the method described in the first aspect or various possible designs of the first aspect.

[0018] This application provides an automatic candidate selection method, electronic device, and computer-readable storage medium. In this automatic candidate selection method, firstly, candidate data corresponding to the candidate object is obtained from a target data platform, and feature extraction is performed on the candidate data to obtain the candidate object's feature information. Secondly, the feature information of the candidate object is matched with the target information to obtain a matching score. This transforms the multi-dimensional object evaluation process, which originally relied on manual comparison and subjective judgment, into an automated quantitative judgment process based on the matching score. Therefore, even when faced with multi-dimensional conditions such as fan base size, audience profile, and content performance, it can still perform a unified, objective, and continuous selection judgment on whether the candidate object meets the target requirements. Based on this, by identifying candidate objects that meet the target requirements as valid objects and adding them to the valid object set, and continuing to acquire candidate objects and perform subsequent filtering processes until the number of valid objects reaches the preset number, continuous automatic filtering guided by the target number can be achieved. Therefore, compared with the existing technology that mainly relies on manual browsing, recording and judging of influencer accounts one by one, this application can reduce the time consumption and subjective bias caused by manual participation and judgment, improve the efficiency of candidate object filtering, and improve the accuracy of candidate object determination under multi-dimensional conditions, thereby solving the problems of low filtering efficiency and difficulty in determining candidate objects under multi-dimensional conditions in related technologies. Attached Figure Description

[0019] Figure 1 A flowchart illustrating an automatic candidate selection method provided in an embodiment of this application; Figure 2 A flowchart illustrating another automatic candidate selection method provided in this application embodiment; Figure 3 A flowchart illustrating another automatic candidate selection method provided in this application embodiment; Figure 4 A flowchart illustrating yet another method for automatically selecting candidate objects provided in an embodiment of this application; Figure 5 A schematic diagram of the structure of an automatic candidate screening device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and drawings of this application are intended to cover non-exclusive inclusion.

[0022] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of the phrase "embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists, A and B can exist simultaneously, and B exists. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] Furthermore, the terms "first," "second," etc., in the specification and claims of this application or in the aforementioned drawings are used to distinguish different objects rather than to describe a specific order, and may explicitly or implicitly include one or more of the features.

[0025] In the description of this application, unless otherwise stated, "multiple" and "at least two" mean two or more (including two), and similarly, "multiple groups" and "at least two groups" mean two or more (including two groups).

[0026] In the description of this application, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, "connected" or "linked" can refer not only to a physical connection, but also to an electrical connection or a signal connection. For instance, it can be a direct connection, i.e., a physical connection, or an indirect connection through at least one intermediate component, as long as the circuit is connected. It can also refer to the internal connection between two components. A signal connection can refer not only to a signal connection through a circuit, but also to a signal connection through a medium, such as radio waves. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0027] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, different technical features in this application can be combined with each other.

[0028] In internet marketing scenarios, businesses or advertising agencies typically need to sift through a large number of influencer accounts on internet platforms to select partners that meet their marketing needs. For example, when executing brand promotion or content marketing projects, influencer accounts need to be screened based on project requirements to determine candidates for participating in marketing and promotional activities.

[0029] In relevant business processes, companies typically use internal business platforms to search and filter influencer accounts. Specifically, media personnel log into the platform's backend and input basic filtering criteria—such as follower count range, content area, or account tags—into the platform's filtering interface based on the marketing project's needs, to obtain an initial list of influencer accounts. Subsequently, media personnel need to access each influencer account's details page to view multi-dimensional data, including recent content views, engagement rates, follower demographics, audience geographic distribution, and collaboration pricing. Based on this data, they manually assess whether the influencer account meets the business requirements. For influencer accounts deemed suitable, media personnel typically need to record and organize relevant information to form the final candidate influencer list.

[0030] However, in practical applications, manual screening methods have significant limitations due to the large number of influencer accounts on internet platforms and the complexity of the data dimensions involved in each account. First, facing a large amount of influencer account data, media personnel need to browse and evaluate each account individually, resulting in a time-consuming and inefficient screening process. Second, influencer screening typically involves a combination of multiple data dimensions, such as fan base size, content performance metrics, audience structure, and collaboration pricing. Manual methods struggle to accurately compare multi-dimensional data in a short time, easily leading to omissions or biases, thus affecting the accuracy of the screening results. Furthermore, when companies are running multiple marketing projects simultaneously, manual methods cannot handle multiple screening tasks concurrently, resulting in weak task processing capabilities. Finally, the lack of a unified structured storage method for data obtained during manual screening makes it difficult to reuse historical screening results, hindering the accumulation and utilization of data assets.

[0031] Therefore, in internet marketing, how to efficiently screen candidates when faced with a large amount of influencer account data and multi-dimensional screening conditions, thereby improving screening efficiency and accuracy, has become a pressing technical problem that needs to be solved.

[0032] To address the aforementioned issues, this application provides an automatic candidate object screening method. This method involves acquiring candidate object data from a target data platform and extracting features from the candidate object data to obtain the candidate object's feature information. Further, the feature information of the candidate object is matched with target information, and a matching score is calculated. Based on the matching score, it is determined whether the candidate object meets the target requirements, thereby identifying valid objects and adding them to a valid object set. When the number of valid objects in the valid object set does not reach a preset number, new candidate object data is acquired, and the above process is repeated until the number of valid objects reaches the preset number. This achieves automated candidate object screening, improving screening efficiency and accuracy.

[0033] The automatic candidate selection method provided in this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Figure 1 This is a flowchart illustrating an automatic candidate selection method provided in an embodiment of this application. Figure 1 As shown, the automatic candidate selection method provided in this application embodiment specifically includes S101 to S107, and S101 to S107 will be described in detail below.

[0035] It should be noted that this automatic candidate selection method can be executed by an electronic device, which can be a server, cloud processing device, terminal device, or other device with data processing capabilities. The electronic device can complete the automatic candidate selection process by calling instructions stored in its memory to execute computer commands.

[0036] In some embodiments, the automatic candidate selection method provided in this application can run on a data processing system based on automated execution technology. This system can adopt a layered architecture, mainly including a user interaction layer, a business logic layer, an execution layer, and a data storage layer.

[0037] The user interaction layer provides users with a visual interface, which can include modules for task creation, task list display, similar object search, filter parameter display, and data export. Users can submit filtering tasks, view task status, and obtain filtering results through the user interaction layer.

[0038] The business logic layer receives user requests and executes business processes related to the filtering tasks. Specifically, it performs operations such as parameter validation, task creation, task status management, and task scheduling control. Through the business logic layer, user-input filtering parameters can be converted into filtering tasks to be executed, and the execution process of multiple filtering tasks can be managed uniformly.

[0039] The execution layer is used to perform automatic candidate selection tasks. The execution layer can be deployed on servers, virtual machines, or other execution nodes with computing power.

[0040] In some embodiments, the execution layer may include a browser simulation module, a page parsing module, a task execution module, and a matching processing module. The browser simulation module accesses the target data platform, the page parsing module extracts candidate object data from the page, the task execution module executes the candidate object acquisition and processing flow, and the matching processing module performs matching calculations between candidate objects and target information.

[0041] The data storage layer is used to store data related to the filtering task. Specifically, the data storage layer can store task configuration information, structured data of candidate objects, project information, and system logs. Task configuration information can be stored in a structured format, candidate object structured data can include candidate object profile data, content performance data, and cooperation quotation data, and system logs can be used to record task execution status and system running status.

[0042] In some embodiments, the system may further include a scheduling control center. The scheduling control center may include a task queue manager, a resource pool manager, and a dynamic scheduling engine. The task queue manager receives user-submitted filtering tasks, performs parameter validation on the filtering tasks, generates task identifiers, and adds the filtering tasks to the task queue. The resource pool manager maintains a dynamic resource pool of available execution resources and monitors the status of each execution instance. The dynamic scheduling engine performs resource allocation or resource reclamation control on the filtering tasks based on task priority, resource requirements, and system load.

[0043] In some embodiments, the execution layer may also employ a containerized execution architecture. Specifically, the execution layer can run in an isolated execution environment built on container technology to provide a lightweight virtualized runtime environment for filtering tasks. The containerized execution architecture may include a pre-built execution image and multiple task sandbox instances. The execution image may pre-install a browser runtime environment, an automated execution runtime environment, and a script execution environment; each running filtering task may correspond to an independent task sandbox instance. The task sandbox instance may include an independent browser user data directory, an independent network namespace, an independent file system space, and a limited CPU and / or memory resource quota. This approach improves the execution isolation between different filtering tasks and reduces resource interference during concurrent execution of multiple tasks.

[0044] Through the above-described overall system architecture, the entire process of automatic candidate selection can be realized, from user submission, task scheduling, automated execution to result storage, thereby improving the system's scalability, stability, and resource utilization efficiency in multi-task concurrent scenarios.

[0045] S101. Obtain candidate object data corresponding to the candidate object from the target data platform.

[0046] The target data platform can be a platform used to display object information, object profiles, object content performance data, or object business data. Taking the internet marketing scenario as an example, the target data platform can be a business platform, content platform, social platform, influencer management platform, or other platforms that can provide object retrieval and detail display functions. This embodiment does not impose specific limitations.

[0047] In one implementation, candidate objects can be obtained from a target data platform based on user-input filtering parameters. Specifically, users can input filtering parameters through the business platform interface, such as a range of follower counts, object tags, content categories, interaction rate thresholds, or audience structure parameters. The electronic device performs an object search in the target data platform based on the filtering parameters, thereby obtaining a list of candidate objects that meet the initial search criteria, and further retrieving the corresponding candidate object data.

[0048] In another implementation, candidate objects can be determined based on baseline objects.

[0049] The base object can be a user-specified object, such as the object corresponding to the user-input object homepage, the object corresponding to the object name, or the object corresponding to the object identifier.

[0050] Electronic devices can first identify a benchmark object and acquire relevant data about the benchmark object, and then identify objects in the target data platform that have a certain relationship with the benchmark object, such as objects similar to the benchmark object, thereby obtaining candidate objects.

[0051] In this embodiment, candidate object data may include various types of raw or semi-structured data related to the candidate object, such as text data, numerical data, graphical interface data, historical content data, etc.

[0052] Textual data may include object name, object description, object tags, object category, etc.; numerical data may include number of followers, number of views, number of interactions, interaction rate, price data, etc.; graphical interface data may include visual charts presented by graphical interface elements in the target data platform page; historical content data may include a preset number of historical posts published by the candidate object and the corresponding data.

[0053] S102. Extract features from the candidate object data to obtain the feature information of the candidate objects.

[0054] It should be noted that candidate object data usually comes from complex sources, has diverse data formats, and different data types are expressed in different ways. Therefore, it is necessary to extract features from the original candidate object data to convert it into candidate object feature information that can be used for subsequent matching processing.

[0055] In some embodiments, electronic devices can extract basic attribute data, audience structure data, and content performance data of candidate objects from candidate object data.

[0056] Basic attribute data may include at least one of object identification information, object description information, and object tag information; audience structure data may include at least one of fan number data, gender distribution data, age distribution data, and geographic distribution data; content performance data may include at least one of readership data, interaction data, interaction rate data, viral article rate data, and completion rate data.

[0057] In this embodiment, after extracting the basic attribute data, audience structure data, and content performance data of the candidate object, the electronic device can uniformly organize, encode, or concatenate these various types of data to generate the feature information of the candidate object. The feature information of the candidate object can be a structured feature set, a combination of attribute fields, vectorized input pre-feature information, or other data representations that can characterize the features of the candidate object.

[0058] S103. Match the feature information of the candidate object with the target information to obtain a matching score.

[0059] In this embodiment, when the candidate object is determined based on the filtering parameters input by the user, the target information is the filtering parameters; when the candidate object is determined based on the benchmark object, the target information is the feature information of the benchmark object.

[0060] It should be noted that, in order to determine the degree of matching between the candidate and the user's current filtering target, the electronic device can match the feature information of the candidate with the target information and calculate the corresponding matching score.

[0061] In one implementation, when candidate objects are obtained from a target data platform based on user-input filtering parameters, the target information can be the filtering parameters. In this case, the electronic device determines the degree of matching between the candidate objects and the user's current filtering target by matching the feature information of the candidate objects with the filtering parameters.

[0062] In another implementation, when the candidate object is determined based on the reference object, the target information can be the feature information of the reference object. In this case, the electronic device determines the degree of similarity or matching between the candidate object and the reference object by matching the feature information of the candidate object with the feature information of the reference object.

[0063] It should be noted that the matching score is used to quantify the degree of matching between the candidate and the target information. The higher the matching score, the higher the degree of matching between the candidate and the target information; the lower the matching score, the lower the degree of matching between the candidate and the target information.

[0064] S104. Based on the matching score, determine whether the candidate object meets the target requirements.

[0065] It should be noted that the target requirement is used to characterize the screening task for candidate objects under the current screening task. Specifically, it can be implicitly determined by the screening parameters input by the user, or it can be implicitly determined by the similarity matching requirements corresponding to the benchmark object.

[0066] When the candidate object meets the target requirements, the electronic device executes the method steps shown in S105; when the candidate object does not meet the target requirements, the electronic device repeats the method steps shown in S101.

[0067] S105. Determine the candidate objects as valid objects and add the valid objects to the set of valid objects.

[0068] In this embodiment, when the electronic device determines that a candidate object meets the target requirements, it marks the candidate object as a valid object. A valid object refers to a candidate object that meets the current filtering task after being judged through screening.

[0069] In some embodiments, the electronic device can create corresponding record entries for valid objects and add the valid objects to a valid object set. The valid object set is a collection of objects identified as valid objects in the current filtering task. Each valid object in the valid object set can store its object identifier, object characteristic information, matching score, and additional information related to the filtering task.

[0070] In some embodiments, the electronic device may also update the number of objects in the valid object set when a valid object is added to the valid object set, so as to determine whether the number of valid objects in the valid object set has reached a preset number.

[0071] S106. Determine whether the number of valid objects in the valid object set has reached the preset number.

[0072] The preset quantity can be the number of target objects that the user sets in advance when creating a filtering task.

[0073] For example, in a marketing campaign, users can specify the number of target influencers to be filtered out as 10, 20, or other numbers in the task configuration interface. The electronic device can continuously count the number of valid individuals in the valid target set during the filtering process and compare the number of valid individuals in the valid target set with the preset number.

[0074] In some embodiments, in addition to determining whether the number of valid objects in the valid object set has reached a preset number, the electronic device can also monitor the execution status of the current filtering task in real time to predict whether the current filtering task can reach the preset number.

[0075] Specifically, the electronic device can continuously record the number of processed candidate objects and the number of valid objects matched during the current screening task, and calculate the hit rate of the current screening task in real time based on the above data. At the same time, the electronic device can also record the matching score distribution information of the valid objects matched, such as the average score, the highest score, the lowest score, and the standard deviation of the score, to reflect the overall quality level of the currently matched objects.

[0076] In some embodiments, the electronic device can construct a hit rate prediction model based on processed candidate object data to predict the probability that the current screening task will ultimately achieve its target number of targets. For example, the electronic device can use a Bayesian inference method to estimate the current hit rate and continuously update the confidence interval of the hit rate as the number of processed candidate objects increases. An unbiased prior can be used in the initial stage, and the estimated interval of the hit rate gradually converges as the processed data gradually increases. Furthermore, the electronic device can predict the number of valid objects that the current screening task can subsequently obtain based on a conservative estimate in the hit rate confidence interval, such as a 95% confidence lower bound.

[0077] In some embodiments, the electronic device can predict the number of candidate objects that still need to be processed after reaching a preset number based on a conservatively estimated hit rate, and compare the predicted number of candidate objects required with the current number of remaining candidate objects. When the predicted number of candidate objects required is significantly greater than the number of remaining candidate objects, the electronic device can further calculate the probability that the current screening task cannot complete the preset number, thereby assessing the probability of completing the current screening task.

[0078] When the electronic device determines that the number of valid objects in the valid object set has reached the preset number, the electronic device can execute the method steps shown in S107; when the number of valid objects in the valid object set has not reached the preset number, the electronic device can return to execute S101, continue to obtain new candidate object data from the target data platform, and repeat the subsequent feature extraction, matching calculation and filtering judgment steps to further supplement the valid objects in the valid object set.

[0079] S107. Stop automatic filtering of candidate objects.

[0080] In this embodiment, when the electronic device determines that the number of valid objects in the valid object set has reached a preset number, the automatic selection process of candidate objects can be stopped.

[0081] Specifically, the electronic device can terminate the subsequent processes of acquiring candidate objects, extracting features, performing matching calculations, and making filtering judgments, thereby ending the current filtering task. After stopping automatic filtering, the electronic device can output the information of the valid objects in the valid object set for users to view, analyze, or export later.

[0082] In some embodiments, the electronic device can organize the valid object information from the valid object set and generate filtering result data. For example, it can generate a result list containing object identification information, object feature information, and matching score, or export the filtering results as a data table for users to make further business decisions.

[0083] In other embodiments, the electronic device may also store the screening results in a business database or task management system for subsequent task reuse or historical data analysis.

[0084] This embodiment sets up a valid object quantity judgment mechanism, which enables electronic devices to automatically terminate the screening process after obtaining a number of valid objects that meet the business requirements, thereby avoiding invalid data processing and improving the efficiency of candidate object screening.

[0085] In some embodiments, when the electronic device predicts that the current screening task has a high probability of failure based on the task execution status, it can also generate task execution suggestions and dynamically control the current screening task. Specifically, when the electronic device determines that the probability of the current screening task completing a preset number of tasks is lower than a preset threshold, it can determine that the current screening task is a high-failure-probability task and generate an early termination suggestion to stop the current screening task and release computing resources.

[0086] In some embodiments, the electronic device can also generate different types of decision suggestions based on the task execution status. For example, when the current screening task is progressing normally and it is predicted that a preset number can be completed, the electronic device can generate a suggestion to continue execution normally; when the current hit rate is lower than expected but there is still a possibility of completion, the electronic device can generate a strategy adjustment suggestion, such as lowering the preset scoring threshold, relaxing some screening conditions, or adjusting the matching parameters; when the current hit rate is significantly too low and it is predicted that the preset number cannot be completed, the electronic device can generate a suggestion to terminate early, thereby terminating the current screening task and releasing system resources.

[0087] In this embodiment, the execution status of the current screening task is monitored in real time during the automatic candidate screening process. The task hit rate is dynamically estimated based on the number of processed candidate objects, the number of valid objects hit, and the matching score distribution information. At the same time, a statistical prediction model is used to predict the probability of the current screening task finally achieving the target number. This allows the system to identify low-probability tasks in advance during the screening task execution process and generate decision suggestions for continuing execution, adjusting strategies, or terminating early based on the prediction results. This enables the system to reduce the ineffective execution of low-probability tasks while ensuring the quality of screening results, reducing system resource waste, improving the overall efficiency of screening task completion, and providing users with a data analysis-based basis for task execution decisions.

[0088] This application provides an automatic candidate selection method. First, candidate data corresponding to the candidate object is obtained from a target data platform, and feature extraction is performed on the candidate data to obtain the candidate object's feature information. Second, the feature information of the candidate object is matched with the target information to obtain a matching score. This transforms the original multi-dimensional object evaluation process, which relied on manual comparison and subjective judgment, into an automated quantitative judgment process based on the matching score. Therefore, even when faced with multi-dimensional conditions such as fan base size, audience profile, and content performance, it can still perform a unified, objective, and continuous selection judgment on whether candidate objects meet the target requirements. Based on this, by selecting candidates that meet the target requirements... The selected objects are identified as valid objects and added to the valid object set. If the number of valid objects in the valid object set does not reach the preset number, the process of acquiring candidate objects and subsequent filtering continues until the number of valid objects reaches the preset number. This enables continuous automatic filtering guided by the target number. Therefore, compared with the existing technology that mainly relies on manual browsing, recording and judging of influencer accounts one by one, this application can reduce the time consumption and subjective bias caused by manual participation and judgment, improve the efficiency of candidate object filtering, and improve the accuracy of candidate object determination under multi-dimensional conditions. This solves the problems of low filtering efficiency and difficulty in determining candidate objects under multi-dimensional conditions in related technologies.

[0089] In the above embodiments, the electronic device needs to match the feature information of the candidate object with the target information to obtain a matching score. Next, the specific process by which the electronic device matches the feature information of the candidate object with the target information to obtain a matching score will be described in detail.

[0090] Figure 2 This is a flowchart illustrating another automatic candidate selection method provided in an embodiment of this application. Figure 2 As shown, in one possible embodiment, the method steps shown in S103 can be implemented by S1031 to S1033, which will be described in detail below.

[0091] S1031. Construct a feature vector of the candidate object based on the feature information of the candidate object.

[0092] Among them, the candidate object feature vector is a data representation form used to characterize the multi-dimensional features of the candidate object. The candidate object feature vector can uniformly map the attribute information of the candidate object in different dimensions into a vector structure suitable for subsequent calculation and processing.

[0093] In this embodiment, the feature information of the candidate object may include multiple feature dimensions. For example, the feature information of the candidate object may include basic attribute-related features, audience structure-related features, and content performance-related features.

[0094] Among them, basic attribute-related features may include information such as object identifier, object description, object tag, and content classification; audience structure-related features may include information such as number of fans, fan gender distribution, fan age distribution, and fan geographic distribution; content performance-related features may include information such as number of reads, number of interactions, interaction rate, viral article rate, and completion rate; electronic devices can organize and encode the above-mentioned feature information of different dimensions to form a candidate object feature vector.

[0095] In one implementation, the electronic device can directly use the numerical features in the candidate object's feature information as vector dimension values. For example, numerical indicators such as the number of followers, interaction rate, viral article rate, and completion rate can be mapped to different dimensions of the candidate object's feature vector to form a numerical vector representation.

[0096] In another implementation, for non-numerical features, the electronic device can first perform feature encoding and then construct the vector. For example, for discrete features such as object tags, content categories, and geographic categories, they can be converted into corresponding numerical representations through preset mapping relationships, one-hot encoding methods, or other encoding methods, and then further written into the corresponding dimensions of the candidate object feature vector.

[0097] In some embodiments, the electronic device can also perform standardization or normalization processing on feature information from different sources to reduce the impact of features with different dimensions on subsequent matching calculation results. For example, absolute numerical indicators such as the number of followers, number of views, and number of interactions can be converted into standardized values ​​under a unified dimension, thereby improving the rationality and comparability of subsequent scoring calculations.

[0098] S1032. Based on the target information, construct the target information vector.

[0099] Among them, the target information vector is a data representation used to characterize the features corresponding to the target requirements in the current screening task, so as to perform matching calculations with the feature vectors of candidate objects in the subsequent process.

[0100] In this embodiment, the specific content of the target information can vary depending on the source path of the candidate object. When the candidate object is determined based on the filtering parameters input by the user, the target information can be the filtering parameters; when the candidate object is determined based on the benchmark object, the target information can be the feature information of the benchmark object.

[0101] In one implementation, when the target information is a filtering parameter, the electronic device can construct a target information vector based on the filtering parameter. Specifically, the filtering parameters input by the user can include multiple dimensions of filtering conditions, such as a range of fan counts, object tags, content categories, interaction rate thresholds, audience structure conditions, and price ranges. The electronic device can vectorize the filtering parameters according to preset dimensional mapping rules to form a target information vector. For example, the upper and lower limits of the fan count range can be mapped to corresponding numerical dimensions, object tags and content categories can be mapped to corresponding category dimensions, and interaction rate thresholds, audience proportion thresholds, etc., can be mapped to corresponding target constraint dimensions.

[0102] In another implementation, when the target information is the feature information of a reference object, the electronic device can construct a target information vector based on the feature information of the reference object. Since the feature information of the reference object itself may have the same or similar feature dimensions as the feature information of the candidate object, the electronic device can organize and encode the feature information of the reference object in the same or compatible way as constructing the feature vector of the candidate object, thereby forming the target information vector.

[0103] In some embodiments, to ensure the comparability of subsequent matching calculations, electronic devices can maintain consistency or compatibility between the target information vector and the candidate object feature vector in terms of dimensional definition. For example, when the candidate object feature vector includes tag dimension, audience structure dimension, and content performance dimension, the target information vector can also be defined in the corresponding dimensions so that subsequent matching score calculations can be performed on the same dimension basis.

[0104] S1033. Calculate the matching score based on the candidate object feature vector and the target information vector.

[0105] The matching score is calculated using a weighted function, where the weight parameters represent the importance of different matching dimensions.

[0106] It should be noted that the matching score is used to characterize the degree of matching between the candidate object and the target information.

[0107] In one implementation, the electronic device can calculate the local matching results between the candidate object feature vector and the target information vector according to each matching dimension, and summarize the local matching results through a weighting function to obtain the overall matching score.

[0108] Local matching results can reflect the degree of matching between the candidate and the target information in a single dimension, such as the degree of matching in the dimension of the number of fans, the degree of matching in the dimension of object tags, the degree of matching in the dimension of audience structure, or the degree of matching in the dimension of content presentation.

[0109] In some embodiments, the weight parameters in the weighting function are used to characterize the importance of different matching dimensions. That is, different feature dimensions may play different roles in the candidate selection process, so different weight parameters can be set for different dimensions.

[0110] For example, in some marketing tasks, audience structure may be more important than content presentation; in this case, electronic devices can assign a higher weight to the audience structure dimension. In other tasks, object tags and content categories may be more important; in this case, electronic devices can correspondingly increase the weight of the tag dimension. By introducing weight parameters, the matching score can be made more in line with actual business needs.

[0111] In one specific implementation, the electronic device can obtain the matching results for each dimension based on the difference, similarity, or satisfaction degree between the candidate object feature vector and the target information vector in each dimension, and then perform weighted summation according to preset weights to obtain the matching score.

[0112] In another specific implementation, the matching score can also be determined based on the weighted average of the matching results of each dimension, the normalized weighted result, or other comprehensive calculation results.

[0113] In some embodiments, since the influence of each feature dimension on the candidate selection results may vary across different platforms, industries, or business types, the weight parameters in the matching score can be adaptively determined according to different business scenarios, rather than using fixed weights.

[0114] For example, in the beauty industry, the percentage of women in the target audience can be considered a significant feature; however, in the maternal and infant industry, the age distribution of the target audience may be more important. Therefore, consistently using fixed weight parameters may result in matching scores that fail to accurately reflect the actual screening needs in different business scenarios.

[0115] In one implementation, the electronic device can establish a historical success case database, which records sample data corresponding to multiple historical screening tasks. The sample data includes at least the characteristic information of the objects that ultimately achieved cooperation and the characteristic information of the objects that did not achieve cooperation. When a new screening task is initiated, the electronic device can extract historical sample data of the same or similar type as the current screening task from the historical success case database, and analyze the importance of each characteristic dimension in distinguishing successful and unsuccessful samples based on the historical sample data, thereby determining the weight parameters corresponding to each matching dimension.

[0116] In another implementation, the electronic device can also utilize statistical learning methods to analyze historical sample data to determine the contribution of different feature dimensions. For example, a logistic regression algorithm can be used to model the influence of different feature dimensions in distinguishing successful and unsuccessful samples, thereby generating a set of initial weight parameters that match the current business scenario. In this way, the electronic device can automatically learn a matching model suitable for the current screening task without manual adjustment of the weight parameters, improving the adaptability of the matching score to different business scenarios.

[0117] In this embodiment, the matching score can be a single numerical value used to quantify the overall degree of matching between the candidate and the target information. Generally, a higher matching score indicates that the candidate is more consistent with the current filtering target; a lower matching score indicates that the candidate is less consistent with the current filtering target.

[0118] It should be noted that electronic devices can use different decision functions for different types of feature dimensions when calculating matching scores. If a simple 0 / 1 decision method is always used, it is easy to produce abrupt decision results under boundary conditions, causing some candidate objects near the boundary to be directly excluded.

[0119] For example, when the target requirement is at least 100,000 followers, a candidate with 99,999 followers might be directly judged as not meeting the requirement, while a candidate with 100,001 followers would be judged as meeting the requirement, even though the difference in actual business value might be small. To address this, electronic devices can employ soft thresholding functions for at least some feature dimensions to convert discrete judgment results into continuous values ​​between 0 and 1, thereby smoothly reflecting the degree of matching between the candidate and the target requirement in the corresponding feature dimensions.

[0120] In some embodiments, the soft threshold function can be implemented using an S-curve, the steepness of which can be adjusted according to business requirements.

[0121] In some embodiments, electronic devices can select different judgment functions based on different field types. For numerical range features (such as the number of followers, number of reads, number of interactions, interaction rate, viral article rate, completion rate, or price range), a hard threshold mode or a soft threshold mode can be used to determine the degree of matching by judging whether the corresponding indicator of the candidate object is within or close to the preset range; for categorical attribute features (such as tag attributes such as object tags, content categories, or keywords), the matching result can be determined by judging whether the corresponding value of the candidate object is within the set of allowed values; for audience structure-related features (such as the gender distribution of followers, the age distribution of followers, or the geographical distribution of followers), the matching score can be calculated based on the proportion or distribution of the candidate object on the corresponding indicator; for textual features (such as text information in the object introduction or content title), the degree of matching between the candidate object and the target keyword can be calculated based on the semantic similarity of the text.

[0122] In some embodiments, the electronic device may also support composite decision rules. These composite decision rules express the logical combination relationship between multiple business conditions. The electronic device can obtain the corresponding composite decision result by logically combining the results of multiple decision functions. This approach avoids the limitations of simply expressing complex business logic through linear weighting, thus allowing the matching score process to better adapt to diverse business screening needs.

[0123] In this way, electronic devices can extend the matching results of candidate objects in each feature dimension from simple discrete judgments to continuous matching values. By combining multi-type judgment functions and composite judgment rules, they can more accurately reflect the degree of matching between candidate objects and target information, thereby reducing the omission of boundary objects, improving the precision of candidate object sorting and comparison, and enhancing the system's adaptability to complex screening requirements.

[0124] In this embodiment, a candidate object feature vector is constructed based on the candidate object's feature information, and a target information vector is constructed based on the target information, enabling the candidate object's feature information and the target information to be expressed in a unified feature space. Furthermore, a matching score is calculated based on the candidate object's feature vector and the target information vector, and a weighting function is used to assign corresponding weights to different matching dimensions. This allows for a quantitative evaluation of the matching degree between the candidate object and the target information, taking into account the importance of multiple feature dimensions. Therefore, this application can transform the process of manually comparing and comprehensively judging multi-dimensional data item by item into an automated scoring process based on vector matching and weighted calculation, thereby improving the efficiency and objectivity of candidate object matching evaluation and enhancing the accuracy of candidate object selection under multi-dimensional conditions.

[0125] In the above embodiments, when calculating the matching score between candidate objects and target information, the electronic device can also model the candidate object selection process as a multi-objective optimization problem.

[0126] Specifically, electronic devices can use evaluation metrics for candidate objects across different feature dimensions as independent objective functions. These could include objective functions characterizing the content quality of candidate objects, cooperation costs, and the degree of match between the candidate object's audience profile and target needs. Based on the values ​​of candidate objects on each objective function, the electronic device can determine the set of non-dominated objects in the feature space and consider this set as the Pareto front candidate set. For candidate objects located on the Pareto front, there is no situation where they can be simultaneously and better replaced by other candidate objects in at least one objective dimension, thus enabling a balanced selection among multiple business objectives.

[0127] By using the above methods, multiple business objectives such as candidate quality, cooperation costs, and audience matching can be considered simultaneously during the automatic candidate screening process, thereby improving the adaptability of the screening results to complex business needs.

[0128] In the above embodiments, the electronic device needs to determine whether the candidate object meets the target requirements based on the matching score. The specific process by which the electronic device determines whether the candidate object meets the target requirements based on the matching score will be described in detail below.

[0129] Figure 3 This is a flowchart illustrating another automatic candidate selection method provided in an embodiment of this application. Figure 3 As shown, in one possible embodiment, the method steps shown in S104 can be implemented by Sa1 to Sa3, which will be described in detail below.

[0130] Sa1. Determine whether the matching score is less than the preset score threshold.

[0131] The preset scoring threshold is a criterion used to determine whether a candidate meets the target requirements.

[0132] In one implementation, the preset scoring threshold used to determine whether a candidate meets the target requirements can be pre-set by the system. For example, a scoring threshold can be pre-configured based on historical screening task data, business experience rules, or platform default strategies, thus serving as a unified judgment standard for candidate screening.

[0133] In another implementation, the preset scoring threshold can also be dynamically adjusted based on the real-time status during the current screening task execution.

[0134] Specifically, the electronic device can monitor the execution progress of the current screening task in real time, including the number of processed candidates, the number of matched candidates, and the remaining time of the task, and adaptively adjust the preset scoring threshold based on the above information.

[0135] For example, if the current screening task hits slower than expected, the electronic device can appropriately lower the preset scoring threshold to increase the probability of candidates entering the effective object set, thereby ensuring that the target number can be reached within the preset time. If the current screening task hits too fast, it may indicate that the current preset scoring threshold is set too low, resulting in a large number of candidates with low matching degree being selected. In this case, the electronic device can appropriately raise the preset scoring threshold to ensure the quality of the screening results.

[0136] In another implementation, the adjustment range of the preset scoring threshold can also be dynamically determined based on the matching score distribution of the matched candidates.

[0137] For example, electronic devices can statistically analyze the average score, score fluctuation range, or score distribution interval of the matched candidates, and limit the adjustment range of the preset score threshold based on the statistical results, so as to avoid the preset score threshold being adjusted too much or too quickly, which would lead to unstable screening results.

[0138] In some embodiments, an electronic device can determine whether the degree of matching between a candidate object and target information meets the screening requirements by judging whether the matching score is less than a preset scoring threshold.

[0139] When the electronic device determines that the matching score is less than the preset scoring threshold, the electronic device executes the method steps shown in Sa3; when the electronic device determines that the matching score is greater than or equal to the preset scoring threshold, the electronic device executes the method steps shown in Sa2.

[0140] Sa2. Determine if the candidate object meets the target requirements.

[0141] When the matching score of a candidate object is greater than or equal to the preset scoring threshold, the degree of matching between the candidate object and the target information meets the preset requirements, and therefore the candidate object is determined to meet the target requirements.

[0142] When a candidate meets the target requirements, the electronic device records the matching score corresponding to the candidate and associates and stores the matching score with candidate identification information, candidate feature information and other data for subsequent sorting, display or analysis.

[0143] Sa3. Determine that the candidate does not meet the target requirements.

[0144] When the matching score of a candidate is less than the preset scoring threshold, the degree of matching between the candidate and the target information does not meet the preset requirements, and thus the candidate is deemed not to meet the target requirements.

[0145] When a candidate does not meet the target requirements, the electronic device can mark it as a failed candidate and ignore it in subsequent screening processes, preventing it from being added to the set of valid candidates. The electronic device can also record the matching score and related feature information of the failed candidates for use in subsequent screening strategy optimization or data analysis. For example, it can analyze the score distribution of different candidates to adjust the scoring threshold and improve the rationality of the screening strategy.

[0146] In the above embodiments, the electronic device needs to extract features from the candidate object data to obtain the feature information of the candidate objects. Next, the specific process of the electronic device extracting features from the candidate object data to obtain the feature information of the candidate objects will be described in detail.

[0147] Figure 4 This is a flowchart illustrating another automatic candidate selection method provided in an embodiment of this application. Figure 4 As shown, in one possible embodiment, the method steps shown in S102 can be implemented by S1021 and S1022, which are described in detail below.

[0148] S1021. Extract the basic attribute data, audience structure data, and content performance data of the candidate objects from the candidate object data.

[0149] Among them, basic attribute data is used to reflect the candidate's own identity attributes or content positioning attributes, audience structure data is used to reflect the composition of the candidate's corresponding audience group, and content performance data is used to reflect the candidate's dissemination effect or interaction effect during the content release process.

[0150] In one implementation, the electronic device can first perform field identification and classification processing on the candidate object data. For example, information such as the object name, object description, object tags, and object category explicitly displayed on the page can be identified and classified as basic attribute data; information such as the number of followers, gender distribution, age distribution, and geographic distribution displayed on the page can be identified and classified as audience structure data; and information such as the number of views, interactions, interaction rate, viral article rate, and completion rate displayed on the page can be identified and classified as content performance data. In this way, the electronic device can map candidate object data from different sources and in different forms to a unified data category system.

[0151] In another implementation, the electronic device can also extract corresponding data based on preset data extraction rules or field templates. For example, field rules, tag rules, or location rules corresponding to different types of data can be pre-configured. After obtaining candidate object data, the electronic device automatically extracts basic attribute data, audience structure data, and content performance data according to the pre-configured rules.

[0152] For different target data platforms, a data mapping relationship corresponding to the platform's page structure can also be established to adapt to the data display methods of different platforms.

[0153] In some embodiments, the electronic device may also perform preprocessing operations on the extracted data, such as removing outliers, standardizing field names, standardizing units of measurement, processing missing data, or performing format conversion, to improve the standardization and consistency of subsequent data use, thereby making the subsequently generated candidate object feature information more stable and comparable.

[0154] S1022. Based on the candidate object's basic attribute data, audience structure data, and content performance data, generate the candidate object's feature information.

[0155] In one implementation, an electronic device can combine basic attribute data, audience structure data, and content performance data in a preset order to form unified candidate object feature information. For example, basic attribute data can be used first as a description of object identity and content positioning, then audience structure data can be used as a description of the object's audience profile, and finally content performance data can be used as a description of the object's dissemination effect, thus obtaining complete candidate object feature information. This method preserves the individual meanings of different data categories while achieving a unified expression of multiple data types.

[0156] In another implementation, the electronic device can fuse basic attribute data, audience structure data, and content performance data based on the importance of different categories of data to generate candidate object feature information. For example, it can perform weighted integration of key fields, feature splicing of multidimensional data, or filter and retain different categories of data according to a preset strategy, thereby generating feature information representations more suitable for subsequent matching processes.

[0157] In some embodiments, the electronic device can also adjust the way candidate object feature information is generated according to the needs of different screening tasks. For example, in some tasks, the proportion of audience structure data in the feature information can be increased, while in other tasks, content performance data or basic attribute data can be highlighted, so that the generated candidate object feature information better meets the current target requirements.

[0158] In this embodiment, by extracting basic attribute data, audience structure data, and content performance data of candidate objects from the candidate object data, and generating feature information of candidate objects based on the basic attribute data, audience structure data, and content performance data, the multi-source and multi-type data originally scattered in the candidate object details page can be uniformly organized and formed into a structured feature expression. Since the basic attribute data of candidate objects can reflect the object's own positioning information, the audience structure data can reflect the characteristics of the corresponding audience group of the object, and the content performance data can reflect the dissemination effect of the object's content, by extracting and comprehensively generating the feature information of candidate objects from the above multi-dimensional data, the overall characteristics of candidate objects can be described more completely and accurately. This provides a more sufficient data foundation for the subsequent matching calculation between candidate objects and target information, improves the completeness and effectiveness of the candidate object feature expression, and thus helps to improve the accuracy of candidate object screening and judgment.

[0159] In the above embodiments, the electronic device needs to extract audience structure data and content performance data from the candidate object data. Since some audience structure data and content performance data are not displayed as direct fields, but rather through page graphics or historical content statistics, the electronic device can obtain these data by parsing the target data platform's page data and performing statistical analysis on historical content data. The specific process of the electronic device extracting audience structure data and content performance data from the candidate object data will be described in detail below.

[0160] In one possible embodiment, the method further includes Sc1 to Sc3, which are described in detail below.

[0161] Sc1: Parse the graphical interface elements in the target data platform page to obtain the audience structure data of the candidate objects.

[0162] The graphical interface elements include Canvas elements or SVG elements, and the audience structure data includes at least one of the following: gender distribution data, age distribution data, and geographic distribution data.

[0163] It should be noted that some target data platforms typically display candidate audience profile information in the form of visual charts, such as pie charts, bar charts, or distribution maps. These charts are often rendered using graphical interface elements on the webpage. For example, the corresponding graphical interface content can be generated using Canvas or SVG elements. Therefore, electronic devices can obtain the audience structure data corresponding to the candidate by parsing the graphical interface elements on the target data platform's webpage.

[0164] In one implementation, after loading the candidate details page, the electronic device can obtain the graphical interface elements corresponding to the audience profile display area on the page and identify the Canvas or SVG elements within them. Subsequently, the electronic device can extract data related to the audience structure by parsing the Canvas drawing data or SVG structural node information. For example, by parsing the proportion information, label information, or numerical information in the graphical elements, the device can obtain the candidate's gender distribution data, age distribution data, and geographical distribution data.

[0165] Sc2: Obtain historical content data of candidate objects.

[0166] In this embodiment of the application, in order to further analyze the content dissemination effect and user interaction of the candidate objects, the electronic device can acquire the historical content data of the candidate objects. The historical content data may include the historical content published by the candidate objects on the target data platform and the data indicators corresponding to the historical content.

[0167] In one implementation, the electronic device can retrieve several historical posts published by a candidate from the candidate's content list page. For example, it can retrieve content posted within a recently preset time range, or retrieve a recently preset number of historical posts. For each historical post, the electronic device can further retrieve its corresponding data metrics, such as views, likes, comments, favorites, or shares.

[0168] Sc3. Based on historical content data, determine the content performance data of candidate objects.

[0169] Among them, content performance data is used to characterize the dissemination effect or interaction effect of candidate objects when publishing content.

[0170] In one implementation, the electronic device can statistically obtain various content performance indicators based on the historical content data of the candidate content. For example, it can calculate the average or maximum number of reads based on the reading volume data corresponding to the historical content, the average number of interactions or the interaction rate based on the interaction volume data, and the viral article rate or completion rate based on the dissemination effect of the historical content. By statistically analyzing the historical content data, content performance data that reflects the content performance level of the candidate content can be obtained.

[0171] In another implementation, the electronic device can also generate multiple content performance metrics based on historical content data, and construct a content performance data set for candidate objects based on these metrics. For example, it can simultaneously generate metrics related to readership, interaction rate, and content dissemination stability, thereby reflecting the content performance capabilities of candidate objects from multiple dimensions.

[0172] In some embodiments, in order to improve the stability of content performance data, electronic devices may also preprocess historical content data, such as removing abnormal content data, correcting extreme values, or weighting data from different time periods, so that the final content performance data can more accurately reflect the actual content performance level of the candidate.

[0173] In this embodiment, audience structure data of candidate objects is obtained by parsing the graphical interface elements in the target data platform page, and historical content data of candidate objects is further obtained. Based on the historical content data, content performance data of candidate objects is determined. This allows audience information and content performance information, which were originally displayed graphically or scattered in historical content, to be automatically extracted and transformed into structured data. Since audience structure data can reflect the composition characteristics of the audience group corresponding to the candidate object, and content performance data determined based on historical content data can reflect the dissemination effect and interaction of the candidate object's content, the key data of candidate objects can be obtained more comprehensively and accurately through the parsing of graphical interface elements and statistical analysis of historical content data. This improves the completeness and accuracy of candidate object feature information extraction and provides a more reliable data foundation for subsequent candidate object matching calculation and screening judgment.

[0174] In the above embodiments, after completing the feature extraction of candidate objects, the electronic device can also cache the feature information of the candidate objects to reduce repeated extraction and calculation of the same candidate object in different screening tasks. Specifically, the electronic device can establish a multi-level feature caching mechanism. After the feature information of a candidate object is extracted, the feature information of the candidate object along with the extraction timestamp is written into the cache, and an expiration period is set for the feature information. Subsequently, when the same candidate object is involved again, the electronic device can preferentially read the feature information of the candidate object from the cache without having to re-execute the complete feature extraction process.

[0175] In some embodiments, the multi-level feature caching mechanism can manage cached content based on a least recently used strategy. When the cache capacity reaches a preset limit, the electronic device can prioritize evicting candidate object feature information that has not been accessed for the longest time. Simultaneously, the electronic device can also record the access frequency of each candidate object feature information and appropriately extend the cache validity period for frequently accessed candidate object feature information to improve the cache hit rate.

[0176] In some embodiments, the electronic device can also perform feature pre-extraction based on a list of candidate objects to be processed in a task queue. Specifically, for candidate objects at the front of the task queue that are expected to enter the formal processing flow soon, the electronic device can start a lightweight feature extraction task in advance to extract the feature information of the corresponding candidate object and write it to the cache before the formal matching calculation. When the candidate object is subsequently subjected to formal screening processing, the electronic device can directly read the feature information from the cache, thereby reducing waiting time.

[0177] In some embodiments, the electronic device can also set different cache validity periods according to the timeliness requirements of different types of feature data. For example, for time-sensitive data such as quotation information and order availability status, a shorter cache validity period can be set; for relatively stable data such as object tags and object descriptions, a longer cache validity period can be set. When the user explicitly requests real-time data, the electronic device can also bypass the caching mechanism and re-extract the feature information of the corresponding candidate object from the target data platform to ensure the real-time nature and accuracy of the data.

[0178] In the embodiments of this application, the electronic device can reduce the repeated extraction and calculation of candidate object feature information, improve the feature extraction efficiency in the automatic screening process of candidate objects, and reduce the system resource consumption in multi-task parallel scenarios.

[0179] In the above embodiments, the electronic device needs to obtain candidate objects from the target data platform based on the filtering parameters input by the user. The specific process by which the electronic device obtains candidate objects from the target data platform based on the filtering parameters input by the user will be described in detail below.

[0180] In one possible embodiment, the electronic device first accesses the target data platform through a browser-automated execution program, then automatically sets search conditions based on filtering parameters, and finally retrieves candidate objects based on the search conditions.

[0181] In some implementations, electronic devices can invoke browser automation programs to simulate user access to a web platform, thereby automatically accessing the target data platform. For example, an electronic device can use automated scripts to control the browser to open the login or search page of the target data platform and complete the necessary authentication or login operations to access the target data platform's business interface. In this way, access to the target data platform can be completed without human intervention.

[0182] After accessing the target data platform, the electronic device can automatically set search criteria based on the user-input filtering parameters. These parameters can include multiple dimensions, such as a range of follower counts, object tags, content categories, interaction rate thresholds, audience structure criteria, or price ranges. The electronic device can map these filtering parameters to corresponding search fields in the target data platform's search interface and automatically fill in the relevant search criteria through a browser automation program. For example, users can enter a range of follower counts in the follower count filter, object tags in the tag filter, and the corresponding content category in the category filter, thus creating search criteria tailored to the user's filtering needs.

[0183] After the search criteria are set, the electronic device can trigger the search operation of the target data platform to obtain a list of candidate objects based on the search criteria.

[0184] In one implementation, the electronic device can trigger a search button or retrieval interface on a webpage via an automated program, thereby loading a list of candidate objects corresponding to the search criteria. Subsequently, the electronic device can extract information such as the identifiers, links, or IDs of the candidate objects from the search results page to determine the set of candidate objects.

[0185] In another implementation, the electronic device can also access the details pages corresponding to the candidate objects sequentially within the search results page, thereby further obtaining detailed data about the candidate objects. This method automates the process of discovering and acquiring candidate objects without requiring manual browsing of each page.

[0186] In this embodiment, the electronic device can automatically access the target data platform and perform object retrieval based on the filtering parameters input by the user, thereby obtaining a set of candidate objects, improving the automation level of the candidate object acquisition process, and reducing the time consumption caused by manual operation.

[0187] In the above embodiments, when candidate objects are not directly obtained based on user-input filtering parameters, but are determined by extending the selection based on a user-specified benchmark object, the electronic device can analyze the feature information of the benchmark object and search for objects with similar features to the benchmark object in the target data platform, thereby determining candidate objects. The specific process by which the electronic device determines candidate objects based on the benchmark object will be further explained below.

[0188] In one possible embodiment, the electronic device first acquires the reference object data of the reference object; second, it extracts features from the reference object data to obtain the feature information of the reference object; finally, based on the feature information of the reference object, it identifies objects similar to the reference object as candidate objects in the target data platform.

[0189] In some implementations, electronic devices can obtain benchmark object data based on object identification information input by the user. For example, the user can input information such as the account name, account homepage link, or object unique identifier of the benchmark object. The electronic device can then access the corresponding object page in the target data platform and obtain the data content corresponding to the benchmark object. The benchmark object data may include object profile information, object tag information, number of followers, audience structure information, and content performance information, thereby forming a data set that can characterize the overall features of the benchmark object.

[0190] After acquiring the reference object data, the electronic device can extract features from the reference object data to obtain the feature information of the reference object.

[0191] In one implementation, the electronic device can perform structured processing on the benchmark object data in a manner similar to or the same as that used for extracting features from candidate objects. For example, it can extract tag features, audience structure features, and content presentation features of the benchmark object, and then organize these features in a unified way to generate feature information for the benchmark object. In this way, the feature information of the benchmark object and the feature information of the candidate object can be kept consistent in terms of data structure or feature dimensions, thereby facilitating subsequent similarity judgment.

[0192] After obtaining the feature information of the benchmark object, the electronic device can identify objects with similar features to the benchmark object as candidate objects in the target data platform.

[0193] In one implementation, the electronic device can retrieve objects with the same or similar tags in the target data platform based on the tag information or content classification information of the reference object, thereby obtaining a set of objects that are similar to the content domain of the reference object.

[0194] In another implementation, the electronic device can also filter objects with similar audience characteristics or content performance levels in the target data platform based on the audience structure characteristics or content performance characteristics of the benchmark object, thereby obtaining objects with similar attribute characteristics to the benchmark object.

[0195] In some embodiments, the electronic device can also comprehensively consider multiple feature dimensions, such as object tag features, audience structure features, and content performance features, to filter or sort objects in the target data platform, thereby determining a set of objects that are most similar to the overall features of the benchmark object, and using these objects as a candidate object set. This approach enables candidate object expansion based on the reference object, helping users discover objects with similar attributes or potential collaborative value to the benchmark object.

[0196] In this embodiment, the electronic device can automatically expand to obtain candidate objects with similar characteristics based on the user-specified baseline object, thereby providing a new source of objects for subsequent candidate object screening, improving the efficiency of candidate object discovery, and expanding the scope of candidate object screening.

[0197] In influencer marketing scenarios, the data related to the benchmark object and platform object are usually distributed across multiple modalities, and the semantic representation of the data in each modality differs significantly, making direct and unified comparison difficult. Therefore, when electronic devices determine candidate objects based on the benchmark object, they can adopt a similar object determination method oriented towards multimodal heterogeneous feature fusion.

[0198] Specifically, multimodal data can include at least one of text modal data, visual modal data, behavioral modal data, and time-series modal data.

[0199] The text modal data can include unstructured text information such as object nicknames, object descriptions, object tags, and titles of historical content; the visual modal data can include image information such as content cover images and video cover images; the behavioral modal data can include structured statistical information such as fan age distribution, fan gender distribution, and fan geographic distribution; and the time-series modal data can include trend information on interaction rate, viral article rate, or other performance indicators within a preset time range.

[0200] Electronic devices can extract features from both the baseline object and the platform object in each modality, and then fuse these features to form a unified feature representation for similarity calculation. This approach reduces the semantic differences caused by the difficulty of directly comparing data from different modalities, thereby improving the accuracy of similar object identification.

[0201] In some embodiments, when calculating the similarity between a benchmark object and a platform object, the electronic device can also assign different levels of importance to different feature dimensions by incorporating business semantic awareness weights. This is because, in influencer marketing scenarios, the importance of each feature dimension varies significantly across different business types. For example, in a beauty business scenario, the importance of fan gender distribution characteristics may be higher than that of fan geographic distribution characteristics; in a local life services scenario, the importance of fan geographic distribution characteristics may be higher than that of content tag characteristics; and in a high-end consumer goods promotion scenario, fan age distribution characteristics may have a higher level of importance. Therefore, the electronic device can set corresponding weights for different modal features and different dimensional features according to the current business scenario, project type, or task requirements, and calculate the business semantic similarity between the benchmark object and the platform object based on these weights. In this way, the similarity calculation results can not only reflect the mathematical similarity relationship between objects, but also better reflect the similarity relationship between objects at the level of business requirements.

[0202] In some embodiments, the electronic device can determine objects similar to the benchmark object as candidate objects based on the multimodal feature fusion results and business semantic awareness weights. Furthermore, the electronic device can output similarity information between the similar objects and the benchmark object across various feature dimensions as explanatory information for the similar object determination results, thereby improving the understandability of the similar object recommendation results.

[0203] Through the above methods, electronic devices can more accurately identify candidate objects similar to the benchmark object in influencer marketing scenarios and improve the matching degree between the expanded candidate object results and actual business needs.

[0204] In the above embodiments, the electronic device needs to obtain candidate object data corresponding to the candidate object from the target data platform. The specific process of the electronic device obtaining candidate object data corresponding to the candidate object from the target data platform will be described in detail below.

[0205] In one possible embodiment, the method steps shown in S101 can be implemented by Sd1, which will be described in detail below.

[0206] Sd1: Perform data scraping on the details page corresponding to the candidate object on the target data platform to obtain the candidate object data.

[0207] It should be noted that the details page corresponding to the candidate object is used to display multi-dimensional information about the candidate object. Therefore, by scraping data from the details page, the raw data required for subsequent feature extraction and matching calculation can be obtained.

[0208] In one implementation, the electronic device can automatically access the details page corresponding to the candidate object based on the page link, object identifier, or details entry information, and perform data scraping operations after the page loads. Data scraping can target the data content displayed in different areas of the details page, such as the object's basic information area, data metric display area, audience profile display area, and historical content display area. By scraping data from these page areas, the electronic device can obtain multiple types of data about the candidate object.

[0209] Data scraping includes scraping at least one of the following: text data, numerical data, graphical interface data, and historical content data.

[0210] It should be noted that textual data may include information such as object name, object description, object tags, and object category; numerical data may include data such as number of followers, number of views, number of interactions, interaction rate, and price; graphical interface data may include audience structure-related data displayed in chart form on the target data platform page; historical content data may include a list of historical content published by the candidate object and the corresponding data indicators for each historical content.

[0211] In one implementation, the electronic device can extract text and numerical data from a details page based on field identifiers, display locations, or page node structures. For example, it can extract corresponding text or numerical content by identifying page elements such as object name fields, follower count fields, and interaction metric fields. This method allows for relatively direct extraction of structured or semi-structured data from details pages.

[0212] In another implementation, electronic devices can also obtain graphical interface data or historical content data by parsing or traversing specific areas of the details page. For example, they can locate the audience profile display area and obtain the graphical interface data corresponding to that area, or they can locate the historical content list area and obtain the historical content entries and their corresponding data metrics. In this way, electronic devices can obtain supplementary data beyond the direct fields on the details page, thereby enriching the content of the candidate object data.

[0213] In some embodiments, the electronic device may also perform preliminary processing on the captured data during data capture, such as unifying field names, standardizing numerical formats, and deduplicating duplicate data, so that it can be directly used in the candidate object feature extraction process later.

[0214] In this embodiment, by obtaining relatively complete candidate object data from the details page corresponding to the candidate object on the target data platform, the original data foundation can be provided for subsequent candidate object feature extraction, matching score calculation and screening judgment.

[0215] In the above embodiments, when candidate objects are obtained from the target data platform based on user-input filtering parameters, the electronic device needs to perform object retrieval according to the filtering parameters to obtain candidate objects. To enable candidate object filtering to adapt to different business scenarios and filtering requirements, the filtering parameters can include parameters with multiple dimensions. The specific content of the filtering parameters will be described in detail below.

[0216] In one possible embodiment, the filtering parameters include at least one of the following: basic information parameters, object tag parameters, object data performance parameters, audience structure parameters, and price range parameters. By setting different types of filtering parameters, electronic devices can constrain and filter candidate objects from multiple dimensions such as object basic attributes, object content attributes, object dissemination effect, object audience characteristics, and object cooperation cost.

[0217] Specifically, the basic information parameters may include at least one of the following: platform account, project identifier, task name, and number of tasks.

[0218] Among them, the platform account indicates the business platform account or operating entity corresponding to the current screening task; the project identifier indicates the business project or marketing project to which the current screening task belongs; the task name identifies the specific task information of the current screening task; and the task quantity indicates the number of candidate objects that the current screening task hopes to obtain. By setting basic information parameters, electronic devices can limit the operating environment, task affiliation, and task objectives of the screening task.

[0219] The object tag parameter can include at least one of the following: object tag, content category, and keyword.

[0220] Among them, object tags are used to characterize the content style, account positioning, or business attributes of candidate objects; content categories are used to limit the content field to which candidate objects belong, such as beauty, maternal and infant, digital products, automobiles, education, etc.; keywords are used to further describe the core content characteristics or focus of candidate objects. By setting object tag parameters, electronic devices can prioritize the acquisition of candidate objects that match the target needs in terms of content attributes or field attributes.

[0221] The performance parameters of the object data may include at least one of the following: number of reads, interaction rate, viral article rate, and completion rate.

[0222] Among these metrics, readership reflects the reach and impact of a candidate's content; interaction rate reflects the level of interaction between the candidate's content and its audience; viral article rate reflects the candidate's ability to generate highly effective content; and completion rate reflects the audience's consumption of the candidate's content. By setting these performance parameters, electronic devices can constrain candidates in terms of content dissemination and user interaction, thereby selecting candidates whose content performance better aligns with the target needs.

[0223] Audience structure parameters may include at least one of the following: fan number range, fan gender ratio, fan age distribution, and fan geographic distribution.

[0224] Among these parameters, the fan count range limits the fan base of the candidates; the fan gender ratio limits the proportion of different gender groups within the candidate's audience; the fan age distribution limits the age range of the candidate's audience; and the fan geographic distribution limits the distribution of the candidate's audience across different regions. By setting these audience structure parameters, electronic devices can filter out candidates whose audience profiles better match their business promotion goals.

[0225] The pricing range parameter is used to limit the cooperation cost range of candidate partners. For example, pricing ranges can be set for text and image content, video content, or comprehensive cooperation based on budget requirements. By setting pricing range parameters, electronic devices can constrain the cooperation costs of candidate partners while meeting content attribute and data performance requirements, thereby improving the match between candidate screening results and business budget.

[0226] In one implementation, the electronic device can perform object retrieval based on at least one type of filtering parameters input by the user. For example, when the user only inputs object tag parameters and object data performance parameters, the electronic device can retrieve candidate objects based on two dimensions: content attributes and content performance. In another implementation, the electronic device can also perform object retrieval by combining multiple types of filtering parameters. For example, when object tag parameters, object data performance parameters, audience structure parameters, and price range parameters are input simultaneously, the electronic device can simultaneously limit candidate objects from multiple dimensions, thereby improving the matching degree between the candidate object retrieval results and the user's needs.

[0227] In some embodiments, different types of filtering parameters can be expressed in different ways. For example, some filtering parameters can be expressed as ranges, such as a range for the number of followers or a range for pricing; some filtering parameters can be expressed as proportions, such as the gender ratio of followers or the geographic ratio of the audience; some filtering parameters can be expressed as categories, such as content categories or object tags; and some filtering parameters can also be expressed as keywords, so as to provide more flexible descriptions and retrieval of candidate objects. This approach can improve the flexibility and adaptability of filtering parameter configuration.

[0228] In this embodiment, the electronic device can acquire candidate objects from the target data platform based on multi-dimensional filtering parameters, thereby making the source of candidate objects more in line with actual business needs and providing a more targeted data foundation for subsequent automatic screening of candidate objects.

[0229] In the above embodiments, the electronic device can perform persistent storage and export processing on the screening results.

[0230] In some embodiments, when an electronic device persists the captured candidate object-related data, it may employ a metadata-based flexible data persistence mapping mechanism.

[0231] Specifically, electronic devices can employ a hybrid storage architecture combining core fixed fields and dynamically extended fields to store candidate object data. For relatively fixed data fields such as object identifiers and follower counts, data can be directly written to preset database fields; for data attributes newly added or dynamically changing on the target data platform page, data can be written to the extended field area. Furthermore, electronic devices can set up a metadata dictionary to record the mapping relationship between dynamically extended fields and actual data attributes. When the electronic device detects a new data attribute, it can automatically register the new data attribute to the metadata dictionary, thus achieving persistent storage of new fields without manual modification of the database table structure definition. This approach improves the system's adaptability to dynamic fields and unstructured web page data.

[0232] In some embodiments, when responding to a user's request to export data, the electronic device may also employ an intelligent dynamic query construction and multi-scope routing mechanism.

[0233] Specifically, when the front-end submits field selection results to the electronic device, the electronic device can extract corresponding attributes from fixed or extended field areas based on the user-selected field set and field mapping relationships, and dynamically construct query statements to avoid the additional data transmission overhead caused by retrieving a large number of irrelevant fields. Furthermore, when the user selects different export ranges, the electronic device can also use different query methods based on the export range. For example, when the user selects to export data corresponding to the current task, a query method targeting a single task range can be used; when the user selects to export all data corresponding to a certain project, the electronic device can establish a range view or execute a related query based on the project identifier and task association relationship to improve the query efficiency of exporting large-scale data. This approach improves data reading efficiency and scope control capabilities during the export query process.

[0234] In some embodiments, when generating exported files, electronic devices may employ a non-blocking streaming file generation method based on a producer-consumer model.

[0235] Specifically, electronic devices can stream query result sets using database cursors and acquire data in preset batches. Simultaneously, they can utilize file processing components supporting low-memory writes, employing a sliding window buffering mechanism to progressively write the read data batches to the exported file, releasing the memory resources occupied by each batch upon completion. Furthermore, the electronic devices can employ a simultaneous query-write-output streaming approach during data reading and file writing, ensuring the exported file is continuously output to the target storage location or network response channel during the generation process. This reduces memory consumption during large-scale data export, avoids the risk of memory overflow when the exported data volume increases, and improves the user experience during export scenarios.

[0236] In this embodiment, the electronic device can achieve flexible storage of dynamic field data, adaptive querying of different export ranges and field sets, and low-memory streaming export of large-scale result data in the candidate automatic screening system, thereby improving the stability, scalability and response efficiency of the system in dynamic web page data processing and batch result export scenarios.

[0237] In this embodiment, to address the issues of excessive load on execution nodes, decreased task processing efficiency, and insufficient task execution stability when multiple screening tasks are submitted concurrently, the system containing the electronic device can also introduce an intelligent queue scheduling mechanism for automated task execution. Unlike general message queues that schedule tasks solely based on their arrival order, the intelligent queue scheduling mechanism can adaptively schedule screening tasks by considering differences in resource consumption, session dependencies, and long-running characteristics.

[0238] In some embodiments, the intelligent queue scheduling mechanism may include multi-level priority queues and a scheduling strategy based on resource profiles.

[0239] Specifically, the system can preset at least two task queues, such as a high-priority queue and a batch processing queue. Tasks with high response time requirements and low resource consumption can enter the high-priority queue, while tasks with long processing times or large data volumes can enter the batch processing queue.

[0240] When a task is submitted, the system can generate a resource consumption prediction profile based on the task parameters. This profile includes at least one of the following: estimated CPU time, memory usage, and network bandwidth. The scheduling center can then select tasks from the task queue whose resource requirements better match the currently available resources, based on the real-time remaining resource status of each execution node, rather than being limited to a simple first-in, first-out (FIFO) scheduling method. This approach reduces head-of-queue blocking caused by large tasks blocking smaller tasks and improves overall resource utilization.

[0241] In some embodiments, the intelligent queue scheduling mechanism may also include a browser affinity-based task sticky scheduling method.

[0242] Specifically, when an execution node first executes a filtering task corresponding to a specific platform and account, the system can save the browser session information corresponding to that execution node. This browser session information includes at least one of the following: browser fingerprint, cookies, and local storage information. Subsequently, when filtering tasks corresponding to the same platform and account exist, the scheduling center can prioritize assigning the filtering task to the execution node holding the browser session information, thereby reusing existing session states and reducing repeated login operations. This approach reduces the time consumed by browser cold starts and repeated logins, and lowers the probability of triggering risk controls on the target data platform.

[0243] In some embodiments, the intelligent queue scheduling mechanism may also include a fault-tolerant handling method based on heartbeat detection and task state rollback.

[0244] Specifically, when assigning a task to an execution node, the scheduling center can first mark the task status as "in execution" and set a lease time. During task processing, the execution node periodically sends heartbeat information to the scheduling center and sends confirmation information upon completion of the task. If the scheduling center does not receive heartbeat or confirmation information from the corresponding execution node within the lease period, it can determine that the execution node is out of contact and roll back the corresponding task status to "pending processing" so that other available execution nodes can take over the execution.

[0245] Furthermore, execution nodes can periodically record processing progress information during task execution, so that new execution nodes can continue execution based on this progress information after taking over the task. This approach improves the fault tolerance and system availability of filtering tasks in long-running scenarios.

[0246] In some embodiments, the intelligent queue scheduling mechanism may further include a queuing theory-based elastic scaling prediction strategy.

[0247] Specifically, the system can continuously monitor the task arrival rate and average task processing time in the task queue, and predict the queue backlog trend within a preset time range based on the task arrival rate and average task processing time. When the prediction indicates that the future queue length will exceed the current system processing capacity, the system can start new execution nodes in advance; when the task arrival rate decreases and idle execution nodes continue to exceed a preset time, the system can also automatically reclaim some execution nodes. In this way, the scaling up and down of execution resources can be transformed from passive response to proactive prediction, thereby improving the system's responsiveness in sudden task scenarios.

[0248] In some embodiments, to further enhance the isolation when multiple tasks are executed in parallel, the execution instances of the filtering tasks can also run in mutually isolated execution environments. The execution environment can be a containerized runtime environment, a virtualized runtime environment, or other isolated runtime environments. This approach reduces resource interference between different filtering tasks and improves stability and security during execution.

[0249] In this way, the system containing the electronic device can adaptively schedule the screening tasks based on task resource requirements, session status, and system load when multiple screening tasks are executed concurrently. Combined with fault tolerance and elastic scaling mechanisms, it can improve the stability and resource utilization during task execution, thereby enhancing the overall processing capability of the automatic candidate screening system in high-concurrency and long-term operation scenarios.

[0250] Figure 5 This is a schematic diagram of the structure of an automatic candidate screening device provided in an embodiment of this application. Figure 5 As shown, the candidate object automatic screening device 500 provided in this embodiment includes an acquisition module 501, a feature extraction module 502, a matching module 503, a first determination module 504, a second determination module 505, and a repetition execution module 506.

[0251] The acquisition module 501 is used to acquire candidate object data corresponding to the candidate object from the target data platform; wherein, the candidate object is acquired from the target data platform based on the filtering parameters input by the user, or the candidate object is determined based on a benchmark object, wherein the benchmark object is the object specified by the user.

[0252] The feature extraction module 502 is used to extract features from the candidate object data to obtain the feature information of the candidate object.

[0253] The matching module 503 is used to match the feature information of the candidate object with the target information to obtain a matching score; wherein, when the candidate object is determined based on the filtering parameters input by the user, the target information is the filtering parameters; when the candidate object is determined based on the benchmark object, the target information is the feature information of the benchmark object.

[0254] The first determining module 504 is used to determine whether the candidate object meets the target requirements based on the matching degree score.

[0255] The second determining module 505 is used to determine the candidate object as a valid object and add the valid object to the valid object set when the candidate object meets the target requirements.

[0256] The repeat execution module 506 is used to repeatedly execute the step of obtaining candidate object data corresponding to the candidate object from the target data platform when the number of valid objects in the valid object set does not reach a preset number, until the number of valid objects in the valid object set reaches the preset number.

[0257] It should be understood that the corresponding processes performed by each module have been described in detail in the above method embodiments, and will not be repeated here for the sake of brevity.

[0258] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 600 provided in this embodiment includes a memory 601 and a processor 602.

[0259] The memory 601 can be a separate physical unit, connected to the processor 602 via a bus 603. Alternatively, the memory 601 and processor 602 can be integrated and implemented in hardware. The memory 601 stores program instructions, which the processor 602 calls to execute the operations performed by the electronic device in any of the above method embodiments.

[0260] Optionally, when some or all of the methods in the above embodiments are implemented by software, the electronic device 600 may also include only the processor 602. A memory 601 for storing programs is located outside the electronic device 600, and the processor 602 is connected to the memory via circuits / wires to read and execute the programs stored in the memory. The processor 602 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. The processor 602 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0261] The memory 601 may include volatile memory, such as random-access memory (RAM); the memory may also include non-volatile memory, such as flash memory, hard disk drive (HDD) or solid-state drive (SSD); the memory may also include a combination of the above types of memory.

[0262] For example, this application provides a computer-readable storage medium having computer program instructions stored thereon, which are executed by the processor of an electronic device to cause the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0263] For example, this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the operations performed by the electronic device in the above method embodiments.

[0264] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for automatically selecting candidate objects, characterized in that, The method includes: The candidate object data corresponding to the candidate object is obtained from the target data platform; wherein the candidate object is obtained from the target data platform based on the filtering parameters input by the user, or the candidate object is determined based on the benchmark object, which is the object specified by the user; Feature extraction is performed on the candidate object data to obtain the feature information of the candidate object; The feature information of the candidate object is matched with the target information to obtain a matching score; wherein, when the candidate object is determined based on the filtering parameters input by the user, the target information is the filtering parameters; when the candidate object is determined based on the benchmark object, the target information is the feature information of the benchmark object; Based on the matching score, it is determined whether the candidate object meets the target requirements; When the candidate object meets the target requirement, the candidate object is determined as a valid object and added to the valid object set; When the number of valid objects in the set of valid objects does not reach the preset number, the step of obtaining the candidate object data corresponding to the candidate object from the target data platform is repeated until the number of valid objects in the set of valid objects reaches the preset number.

2. The method according to claim 1, characterized in that, The process of matching the feature information of the candidate object with the target information to obtain a matching score includes: Construct a candidate object feature vector based on the feature information of the candidate object; Based on the target information, construct a target information vector; Based on the candidate object feature vector and the target information vector, a matching score is calculated; wherein, the matching score is calculated by a weighted function, and the weight parameters in the weighted function are used to characterize the importance of different matching dimensions.

3. The method according to claim 1, characterized in that, Determining whether the candidate object meets the target requirements based on the matching score includes: Determine whether the matching score is less than a preset scoring threshold; When the matching score is greater than or equal to the preset score threshold, it is determined that the candidate object meets the target requirement; When the matching score is less than the preset score threshold, it is determined that the candidate does not meet the target requirements.

4. The method according to claim 1, characterized in that, The step of extracting features from the candidate object data to obtain the feature information of the candidate object includes: Extract the basic attribute data, audience structure data, and content performance data of the candidate objects from the candidate object data; Based on the candidate object's basic attribute data, audience structure data, and content performance data, the feature information of the candidate object is generated. The basic attribute data includes at least one of object identification information, object profile information, and object tag information; the audience structure data includes at least one of fan number data, gender distribution data, age distribution data, and geographic distribution data; and the content performance data includes at least one of readership data, interaction data, interaction rate data, viral article rate data, and completion rate data.

5. The method according to claim 4, characterized in that, Extracting the audience structure data and content performance data from the candidate object data includes: The graphical interface elements in the target data platform page are parsed to obtain the audience structure data of the candidate object; wherein, the graphical interface elements include Canvas elements or SVG elements, and the audience structure data includes at least one of gender distribution data, age distribution data, and geographic distribution data; Obtain the historical content data of the candidate object; Based on the historical content data, the content performance data of the candidate objects are determined.

6. The method according to claim 1, characterized in that, When the candidate objects are obtained from the target data platform based on user-input filtering parameters, the method further includes: Access the target data platform via a browser-automated execution program; The search criteria are automatically set based on the filtering parameters, and the candidate objects are retrieved based on the search criteria. The step of obtaining candidate object data corresponding to the candidate object from the target data platform includes: Data scraping is performed on the details page corresponding to the candidate object on the target data platform to obtain the candidate object data; wherein, the data scraping includes scraping at least one of text data, numerical data, graphical interface data, and historical content data.

7. The method according to claim 1, characterized in that, When the candidate object is determined based on the benchmark object, the method further includes: Obtain the baseline object data of the baseline object; Feature extraction is performed on the reference object data to obtain the feature information of the reference object; Based on the feature information of the benchmark object, objects similar to the benchmark object are identified in the target data platform as candidate objects.

8. The method according to claim 1, characterized in that, The filtering parameters include at least one of the following: basic information parameters, object tag parameters, object data performance parameters, audience structure parameters, and price range parameters. The basic information parameters include at least one of the following: platform account, project identifier, task name, and number of tasks. The object tag parameters include at least one of object tags, content categories, and keywords; The object data performance parameters include at least one of the following: readership, interaction rate, viral article rate, and completion rate; The audience structure parameters include at least one of the following: fan number range, fan gender ratio, fan age distribution, and fan geographic distribution.

9. An electronic device, characterized in that, include: Memory and at least one processor; The memory stores computer-executed instructions; The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when executed, implement the method as described in any one of claims 1 to 8.

Images