Method, device, equipment, medium and program product for analyzing recommended content

Abstract

The application discloses an analysis method and device for recommended content, equipment, a medium and a program product, and relates to the technical field of computers. The method comprises the following steps: acquiring operation data and effect data corresponding to a plurality of recommended contents; based on a time sequence correlation relationship between the operation data and the effect data, the operation data and the effect data are interleaved and arranged to generate an event sequence corresponding to the recommended contents; the event sequences corresponding to the plurality of recommended contents are clustered to obtain a content clustering result, and the content clustering result is used to indicate a distribution rule of the recommended effect of the plurality of recommended contents under a delivery strategy; and based on the content clustering result, an analysis report corresponding to the plurality of recommended contents is generated. Recommended contents with similar distribution rules can be clustered in the same category, and the analysis efficiency of the recommended content related data can be improved. The embodiments of the application can be applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and auxiliary driving.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, device, medium, and program product for analyzing recommended content. Background Technology

[0002] Internet advertising, as an emerging form of advertising media, is diverse in form and can directly or indirectly promote goods or services. Advertisers can place these internet ads through advertising platforms, or participate in bidding through these platforms to gain more exposure for their ads.

[0003] Depending on their needs, advertisers can use ad placement platforms to set different placement strategies for their ads, such as adjusting budget based on ad importance. The ad placement platform also records the advertiser's historical activity and displays this information according to the advertiser's requirements. Simultaneously, the platform records and presents data such as click-through rate and return on investment (ROI) for each ad to the advertiser.

[0004] However, advertising platforms only present advertisers with the recorded historical data, requiring advertisers to perform data analysis themselves, resulting in low data analysis efficiency. Summary of the Invention

[0005] This application provides a method, apparatus, device, medium, and program product for analyzing recommended content, which can improve the efficiency of analyzing data related to recommended content in advertising scenarios. The technical solution is as follows:

[0006] On the one hand, a method for analyzing recommended content is provided, the method comprising:

[0007] The system acquires operation data and effect data corresponding to multiple recommended content items, where the multiple recommended content items are content published by content delivery accounts. The operation data is used to indicate the content delivery account's settings for the delivery strategy of the recommended content items, and the effect data is used to indicate the recommendation effect of the recommended content items during the delivery process.

[0008] Based on the temporal correlation between the operation data and the effect data, the operation data and the effect data are interleaved to generate an event sequence corresponding to the recommended content;

[0009] Clustering is performed on the event sequences corresponding to the multiple recommended contents to obtain content clustering results. The content clustering results are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the delivery strategy.

[0010] An analysis report is generated based on the content clustering results, corresponding to the multiple recommended contents.

[0011] On the other hand, a method for analyzing recommended content is provided, the method comprising:

[0012] The content delivery platform displays the platform interface provided by the platform, which offers functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery platform is logged into with a content delivery account.

[0013] The platform interface receives a delivery analysis operation from the content delivery account. The delivery analysis operation instructs the analysis of multiple recommended contents based on the historical delivery strategies set by the delivery strategy setting function. The multiple recommended contents are the content published by the content delivery account.

[0014] Based on the aforementioned delivery analysis operation, an analysis report corresponding to the multiple recommended contents is displayed. The analysis report includes content clustering results, which are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the historical delivery strategy.

[0015] The content clustering result is obtained by clustering the event sequences corresponding to the multiple recommended contents. The event sequence is obtained by interleaving the operation data and the effect data according to the temporal correlation between the operation data and the effect data corresponding to the recommended content. The event sequence is used to indicate the temporal correlation between the operation data and the effect data. The operation data is used to indicate the settings of the historical delivery strategy of the content delivery account. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

[0016] On the other hand, an analysis device for recommended content, the device comprising:

[0017] The acquisition module is used to acquire operation data and effect data corresponding to multiple recommended content. The multiple recommended content are content published by content delivery accounts. The operation data is used to indicate the setting of the delivery strategy of the content delivery account for the recommended content. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

[0018] The first generation module is used to interleave the operation data and the effect data based on the temporal correlation between them, and generate an event sequence corresponding to the recommended content.

[0019] The clustering module is used to cluster the event sequences corresponding to the multiple recommended contents to obtain content clustering results. The content clustering results are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the delivery strategy.

[0020] The second generation module is used to generate an analysis report corresponding to the multiple recommended contents based on the content clustering results.

[0021] On the other hand, an analysis device for recommended content, the device comprising:

[0022] The display module is used to display the platform interface provided by the content delivery platform. The platform interface provides functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery platform is logged in with a content delivery account.

[0023] The receiving module is used to receive the delivery analysis operation indicated by the content delivery account in the platform interface. The delivery analysis operation indicates that multiple recommended contents are analyzed according to the historical delivery strategy set by the delivery strategy setting function. The multiple recommended contents are the content published by the content delivery account.

[0024] The display module is also used to display an analysis report corresponding to the multiple recommended contents based on the delivery analysis operation. The analysis report includes content clustering results, which are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the historical delivery strategy.

[0025] The content clustering result is obtained by clustering the event sequences corresponding to the multiple recommended contents. The event sequence is obtained by interleaving the operation data and the effect data according to the temporal correlation between the operation data and the effect data corresponding to the recommended content. The event sequence is used to indicate the temporal correlation between the operation data and the effect data. The operation data is used to indicate the settings of the historical delivery strategy of the content delivery account. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

[0026] On the other hand, a computer device is provided, the terminal including a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, the at least one program, the code set or instruction set being loaded and executed by the processor to implement the analysis method of recommended content as described in any of the embodiments of this application.

[0027] On the other hand, a computer-readable storage medium is provided, wherein at least one piece of program code is stored in the computer-readable storage medium, the program code being loaded and executed by a processor to implement the method for analyzing recommended content as described in any of the embodiments of this application.

[0028] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the analysis method for the recommended content described in any of the above embodiments.

[0029] The technical solution provided in this application includes at least the following beneficial effects:

[0030] By acquiring operational and performance data of recommended content during the delivery phase, and generating corresponding event sequences based on the correlation between the operational and performance data, the event sequences are used as clustering targets. The clustering results are then used to generate analysis reports. In other words, the content clustering results indicate the distribution pattern of recommended content performance under changes in delivery strategies. Recommended content with similar distribution patterns can be clustered into the same category, which can improve the efficiency of analyzing data related to recommended content. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of a computer system provided in an exemplary embodiment of this application;

[0033] Figure 2 This is a flowchart of a recommended content analysis method provided in an exemplary embodiment of this application;

[0034] Figure 3 This is a flowchart of a method for analyzing recommended content provided in another exemplary embodiment of this application;

[0035] Figure 4 This is a schematic diagram of the clustering process of recommended content provided in an exemplary embodiment of this application;

[0036] Figure 5 This is a flowchart of a recommended content analysis method provided in an exemplary embodiment of this application;

[0037] Figure 6 This is a content projection diagram provided in an exemplary embodiment of this application;

[0038] Figure 7 This is a schematic diagram showing the content clustering results provided in an exemplary embodiment of this application;

[0039] Figure 8 This is a schematic diagram of the platform interface provided in an exemplary embodiment of this application;

[0040] Figure 9 This is a structural block diagram of an analysis device for recommended content provided in an exemplary embodiment of this application;

[0041] Figure 10 This is a structural block diagram of an analysis apparatus for recommended content provided in another exemplary embodiment of this application;

[0042] Figure 11 This is a structural block diagram of an analysis apparatus for recommended content provided in another exemplary embodiment of this application;

[0043] Figure 12 This is a schematic diagram of the structure of a server provided in an exemplary embodiment of this application;

[0044] Figure 13 This is a structural block diagram of a terminal provided in an exemplary embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0046] First, a brief introduction to the terms used in the embodiments of this application:

[0047] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0048] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0049] Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0050] Clustering: The process of dividing a collection of physical or abstract objects into multiple classes composed of similar objects is called clustering. A cluster generated by clustering is a set of data objects that are similar to objects in the same cluster and different from objects in other clusters.

[0051] Based on the above definitions, the application scenarios of the analysis method for the recommended content provided in the embodiments of this application are illustrated.

[0052] Advertising platforms provide advertisers with advertising placement services, allowing them to target their ads across various internet scenarios to increase ad exposure. In one example, ads can be placed on relevant web pages provided by the platform. For instance, if the platform provider is a video platform, the ads can be displayed within the video platform's application interface. Users browsing the video platform's application will also see the displayed ads, thus promoting the ads. Another example combines search engines with ad placement. Users search for keywords using a search engine, and the webpage displays multiple search results for those keywords. Advertisers can bid on relevant keywords, with the highest bidder placing ads for those keywords. While the search results are displayed, users may click on the links to the ads, promoting the ads and increasing the advertiser's business exposure.

[0053] After placing ads through an advertising platform, advertisers can also manage their ads through the platform, such as pausing ad placement, adjusting bids, budgets, and changing ad delivery times. All these actions are recorded by the advertising platform, and advertisers can query this data when needed, i.e., view their historical activity records.

[0054] Meanwhile, the advertising platform also records the performance data generated during the ad delivery process. This performance data can include ad spending, impressions, initial conversions, deep conversions, and clicks. The performance data recorded by the advertising platform is also available for advertisers to query.

[0055] In this embodiment, the aforementioned advertising platform also provides advertisers with an advertising performance analysis function. This function not only provides advertisers with a clear view of performance data but also displays the clustering results obtained after clustering operational and performance data. For example, the advertiser instructs the backend server to perform clustering based on operational and performance data through the advertising performance analysis function. The analysis module in the server performs the corresponding processing to obtain the clustering results, which are then sent to the terminal device. The terminal device displays these results, allowing advertisers to intuitively determine the impact of their operations on advertising performance, thereby improving the efficiency of data analysis and enhancing the user experience.

[0056] Based on the above explanations of terms and application scenarios, the implementation environment of the embodiments of this application will be described. For example... Figure 1 As shown, the computer system of this implementation environment includes: terminal device 110, server 120 and communication network 130.

[0057] Terminal device 110 includes various forms of devices such as mobile phones, tablets, desktop computers, laptops, vehicle terminals, and aircraft. Illustratively, terminal device 110 runs a target application provided by a content delivery platform, which provides services for delivering recommended content. Optionally, the aforementioned target application includes various forms of applications such as standalone applications, web applications, and mini-programs within a host application; no limitation is made here.

[0058] Server 120 is used to provide backend support for the aforementioned target applications.

[0059] In a schematic manner, a user sends a report retrieval request to the server 120 through a target application provided in the terminal device 110. This report retrieval request is used to request the server 120 to cluster a target number of recommended content. After receiving the report retrieval request, the server 120 retrieves the operation data and effect data corresponding to the target number of recommended content from the database, inputs the operation data and effect data into the content clustering module, outputs the content clustering result, and sends the content clustering result to the terminal device 110. The terminal device 110 displays the content clustering result.

[0060] In some embodiments, if the computing power of the terminal device 110 meets the overall operation requirements of the above-mentioned recommendation effect analysis function logic, the above-mentioned recommendation effect analysis function can also be implemented by the terminal device 110 alone.

[0061] It is worth noting that the aforementioned server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0062] Cloud technology refers to a hosting technology that unifies hardware, software, and network resources within a wide area network (WAN) or local area network (LAN) to achieve data computation, storage, processing, and sharing. Based on the cloud computing business model, cloud technology encompasses network technology, information technology, integration technology, management platform technology, and application technology. It can form resource pools, providing flexible and convenient on-demand access. Cloud computing technology will become a crucial support. Backend services of technical network systems require substantial computing and storage resources, such as video websites, image websites, and many portal websites. With the rapid development and application of the internet industry, every item may have its own identification mark in the future, requiring transmission to backend systems for logical processing. Data at different levels will be processed separately, and various industry data will require robust system support, which can only be achieved through cloud computing.

[0063] In some embodiments, the server 120 described above can also be implemented as a node in a blockchain system.

[0064] Indicatively, terminal device 110 and server 120 are connected via communication network 130, which can be a wired network or a wireless network, and is not limited here.

[0065] Please refer to Figure 2 This illustrates an analysis method for recommended content as shown in one embodiment of this application. In this embodiment, the method is applied to, for example... Figure 1 In the server shown, the method includes:

[0066] Step 201: Obtain operation data and effect data corresponding to multiple recommended content items.

[0067] For illustrative purposes only, the above recommended content refers to content placed by the content provider on the content delivery platform; that is, the above multiple recommended contents are content published by the content delivery account.

[0068] The above operational data indicates the settings of the content delivery account's delivery strategy for recommended content. For illustrative purposes, the content delivery platform provides a function to set delivery strategies for recommended content; this function allows users to configure the delivery strategy for recommended content.

[0069] In some embodiments, the recommended content can be configured with attribute information to adjust the delivery strategy. Optionally, the attribute information of the recommended content includes at least one of the following: content information, title information, tag information, keyword information, budget information, bidding information, and delivery time information.

[0070] The recommended content information refers to the specific content displayed to the content recipient during the delivery process. Optionally, the recommended content information can be at least one of the following formats: video, image, audio, and text. The recommended content title information is text information used to summarize and express the recommended content information. Optionally, the title information can be custom-set or extracted from the content information. The recommended content tag information is used to mark the content category of the recommended content. The content delivery platform can allocate the recommended content to appropriate recommendation sections based on the tag information. The recommended content keyword information indicates the words associated with the recommended content during delivery. In one example, taking content delivery in a search engine scenario, when the content recipient searches through a search engine, if the search results contain the keywords included in the recommended content, and the recommended content is in a delivery state, the search engine will display the recommended content. Illustratively, the operation types of the above-mentioned data operations can include creating, modifying, and deleting the aforementioned content information, title information, tag information, and keyword information.

[0071] The budget information for the recommended content indicates the amount of virtual resources the content provider plans to invest during the delivery phase of the recommended content. The bidding information for the recommended content indicates the amount of virtual resources the recommended content will spend competing with other content for delivery resources; that is, the bid set by the content provider to acquire delivery resources for the recommended content. In one example, multiple recommended content providers bid for the same search keyword, and the recommended content with the highest bid acquires the delivery resources associated with that search keyword. The delivery time information indicates the settings for the delivery time of the recommended content. Optionally, the content provider can set the time period for pushing the recommended content to content recipients, including the start and end times. Based on the settings of the delivery time information, the recommended content can be controlled to be in a delivery state or a non-delivery state. Illustratively, the above operational data may also include modifications to the budget information, bidding information, and delivery time information.

[0072] In some embodiments, when the content provider's operation of setting recommended content is recorded by the content delivery platform, the operation type corresponding to the operation will be recorded. The operation data includes the operation type and the corresponding operation content. The operation type is used to indicate what kind of operation the content provider performed, such as creating content, pausing content, restarting content, scheduled restart, increasing / decreasing budget, increasing / decreasing bid, modifying delivery time, etc. The operation content is the descriptive information of the specific operation performed by the content provider. Optionally, the operation content can be the relative proportion of the value change caused by the operation, such as increasing the budget by 100% or -50%, etc. Alternatively, the operation content can also be the absolute value of the value change caused by the operation, which is defined separately according to the range of each indicator. For example, for the budget indicator, if the base budget is 1000 and the budget is increased by 500, the recorded operation content is +500.

[0073] As an illustration, when storing and recording the above-mentioned operation data, the content delivery platform also stores the corresponding relationship between the operation data and the recommended content, as well as the operation time information corresponding to the operation execution time. In one example, when the content provider performs relevant setting operations on the recommended content, the content delivery platform records the operation identifier, operation type, operation content, operation time information, and content identifier of the recommended content in the operation database. The operation identifier is used to uniquely identify the operation data of the record in the database, the operation time information is used to indicate the time point when the operation was generated, and the content identifier is used to uniquely identify the recommended content in the database.

[0074] Performance data is used to indicate the effectiveness of recommended content during the delivery process. Indicatively, the content delivery platform records performance data for recommended content during the delivery phase. Optionally, the performance data recorded by the content delivery platform may include at least one of the following: resource consumption, impressions, clicks, shallow conversions, and deep conversions.

[0075] The aforementioned resource consumption indicates the virtual resources consumed during the delivery phase of the recommended content. These virtual resources are the resources that the advertiser needs to expend when delivering the recommended content; that is, the opportunity to deliver the recommended content is obtained through the exchange of virtual resources. For example, these virtual resources can be virtual diamonds, virtual points, or the electronic currency corresponding to real currency on the internet. The aforementioned exposure indicates the number of times the recommended content is received by the content recipient during the delivery phase. For example, the exposure can be determined by the number of times the recommended content is viewed. The aforementioned click count indicates the number of times the recommended content is actively clicked by the content recipient during the delivery phase. For example, in a search engine scenario, when the content recipient enters search terms into a search engine, the search engine displays search results based on the search terms. These search results include recommended content delivered through keyword bidding. At this time, the recommended content is exposed to the content recipient, meaning the exposure of the recommended content increases. When the content recipient clicks on the recommended content and enters the corresponding interface, the click count is determined to have increased. When the aforementioned shallow conversion instruction recommended content is exposed to the content recipient, it is the ratio between the number of operations received by the recommended content for the first type of operation and the number of exposures. When the aforementioned deep conversion instruction recommended content is exposed to the content recipient, it is the ratio between the number of operations received by the recommended content for the second type of operation and the number of exposures. The content recommendation effect of the aforementioned first type of operation is lower than that of the content recommendation effect of the second type of operation. For example, the first type of operation is at least one of click operation, favorite operation, share operation, etc., and the second type of operation is at least one of purchase operation, contact information authorization operation, manual service request operation, etc.

[0076] As an illustration, when the recommended content is in the delivery state, the performance data of the recommended content may change in real time. If every change in performance data is recorded, there may be a problem of excessive data volume. Therefore, the performance data is recorded in a timed manner. In one example, the performance data of the recommended content is recorded according to a preset time interval.

[0077] When recording and storing effect data, the effect data can be stored in a correspondence with the corresponding collection time. That is, the effect database stores the correspondence between effect data and collection time. In some embodiments, to reduce storage resource consumption, when the collection interval is fixed, only the collection time of the first collection can be recorded when storing effect data, and then the effect data can be stored sequentially. That is, the time corresponding to subsequent effect data can be determined based on the collection time of the first collection and the collection interval. In one example, the effect database stores the effect identifier, effect type, effect content, effect time information, and content identifier corresponding to the effect data. The effect identifier is used to uniquely identify the collection of the effect data, and the effect time information is the time period corresponding to the recommended effect. In some embodiments, a piece of effect data recorded at a point in time may include data corresponding to multiple effect types.

[0078] Step 202: Based on the temporal correlation between operational data and effect data, the operational data and effect data are interleaved and arranged to generate an event sequence corresponding to the recommended content.

[0079] Indicatively, event sequences are used to indicate the temporal relationship between operational data and effect data.

[0080] In some embodiments, the above-mentioned event sequence is a one-dimensional sequence generated by interleaving operation data and effect data after aligning them according to time order. That is, the operation data and effect data are aligned according to the correspondence between the first time information corresponding to the operation data and the second time information corresponding to the effect data to generate the event sequence.

[0081] Indicatively, the operation data undergoes a first preprocessing step to obtain an operation sequence. The first preprocessing step includes extracting the operation data corresponding to the target recommended content based on the content identifier in the operation data, and arranging the operation type and operation content in chronological order according to the operation time information in the operation data to obtain the above-mentioned operation sequence. That is, the operation sequence indicates the order in which the operation data records the operation of modifying the delivery strategy, and the above-mentioned target recommended content is any one of the above-mentioned multiple recommended contents.

[0082] In some embodiments, the first preprocessing step described above may further include filtering the operation content according to requirements, that is, selectively extracting operation data corresponding to a portion of the operation content for clustering. Optionally, the filtering requirement for the operation content may be indicated by the terminal device. In one example, the content delivery platform records operation data corresponding to operation content of N operation types. The terminal device instructs that operation data corresponding to operation content of K operation types be used as operation data for event sequence generation, where N is a positive integer and K is a positive integer less than or equal to N. Then, the operation data of the above K operation types are extracted from the operation data corresponding to the target recommended content, and an operation sequence is generated.

[0083] To illustrate, the effect data undergoes a second preprocessing step to obtain an effect sequence. The second preprocessing step includes extracting the effect data corresponding to the target recommended content based on the content identifier in the effect data, and arranging the effect identifier and effect content in chronological order based on the effect time information in the effect data, thereby obtaining the aforementioned effect sequence.

[0084] Step 203: Cluster the event sequences corresponding to multiple recommended contents to obtain content clustering results.

[0085] The above clustering results are used to indicate the distribution pattern of recommendation effects of multiple recommended content under the delivery strategy.

[0086] In a schematic way, the event sequence corresponding to each recommended content can be generated by using operational data and effect data. That is, the event sequence corresponding to each recommended content is obtained, and clustering is performed through the event sequence to obtain the content clustering result.

[0087] In some embodiments, the above content clustering results include multiple event categories obtained through clustering. That is, multiple recommended contents are divided into different event categories in the content clustering results. For different event categories, corresponding operation patterns can be summarized manually, thereby improving the analysis efficiency when jointly analyzing operation data and effect data.

[0088] To illustrate, before clustering event sequences, it is necessary to convert the generated event sequences into vector representations, that is, to perform vector transformation on the event sequences to obtain event vector representations.

[0089] In some embodiments, by inputting the event sequence into a pre-trained sequence vectorization model, the output is a vector representation corresponding to the event sequence. Optionally, the above-mentioned sequence vectorization model can be at least one of the following models that can be used to convert text into vector representation: one-hot encoding model, word to vector (Word2vec) model, paragraph vector (Doc2vec) model, convolutional neural network (CNN) model, autoencoder (AE) model, etc.

[0090] In a schematic manner, after obtaining the event vector representation, clustering is performed based on the event vector representations corresponding to multiple recommended contents to obtain the content clustering results. Optionally, the clustering method used when clustering the event vector representations includes at least one of partition-based clustering methods, density-based clustering methods, and hierarchical clustering methods. Specifically, partitioning clustering methods include k-means clustering, k-means++, bisecting k-means clustering, and kernel k-means clustering; density-based clustering methods include density-based spatial clustering of applications with noise (DBSCAN) and ordering points to identify the clustering structure (OPTICS); and hierarchical clustering methods include bottom-up agglomerative clustering and top-down divisive clustering.

[0091] In some embodiments, the aforementioned multiple recommended content items may come from different content delivery accounts. That is, with full authorization, the server determines multiple recommended content items from the set of recommended content items delivered by multiple accounts, and obtains the operation data and effect data of the aforementioned multiple recommended content items for the clustering process of recommended content items, thereby obtaining content clustering results. These content clustering results can be used in the overall analysis process of the content delivery platform for the recommended content delivery process, and can also be applied to other scenarios. For example, a prediction model can be trained by using the historical delivery status of recommended content items and the aforementioned content clustering results, and the trained prediction model can be applied to the operation guidance of newly created recommended content items, thereby realizing the application of features from the historical operations and effects of content recommendation to the subsequent operation guidance of recommended content items that are still being delivered or newly added.

[0092] In other embodiments, the multiple recommended contents mentioned above may come from a single content delivery account. That is, under the instruction of the content delivery account, the server performs clustering based on the operation data and effect data corresponding to the recommended contents to obtain content clustering results, and returns the content clustering results to the terminal corresponding to the content delivery account for display.

[0093] Step 204: Generate analysis reports corresponding to multiple recommended content based on the content clustering results.

[0094] For illustrative purposes, the above analysis report is data used by the server to indicate the content clustering results to the terminal device.

[0095] In some embodiments, a content projection map is generated based on the content clustering results. The content projection map is used to indicate the density of multiple recommended contents after clustering. That is, the content clustering results are displayed in the form of a view, and the content projection map is transmitted to the terminal device as the report content of the analysis report.

[0096] In other embodiments, a category division list is generated based on the content clustering results. The category division list is used to indicate the content set after multiple recommended contents are divided into at least two event categories. That is, the content clustering results are displayed in the form of a list, and the category division list is transmitted to the terminal device as the report content of the analysis report.

[0097] In summary, the recommended content analysis method provided in this application obtains operational and performance data of the recommended content during the delivery phase, generates corresponding event sequences based on the correlation between the operational and performance data, clusters the event sequences as clustering targets, and generates an analysis report based on the content clustering results. That is, the content clustering results indicate the distribution pattern of the recommended content's performance under changes in delivery strategies. Recommended content with similar distribution patterns can be clustered into the same category, which can improve the efficiency of analyzing related data of recommended content.

[0098] Please refer to Figure 3 This illustration shows a method for analyzing recommended content according to an embodiment of this application. Indicatively, this method is applied to a server of a content delivery platform. The server performs a clustering process of recommended content upon receiving a request from a terminal device. The method includes:

[0099] Step 301: Receive a report retrieval request from the content delivery account. The report retrieval request includes clustering criteria.

[0100] The aforementioned report request is used to request analysis of a target number of recommended content items. Illustratively, the report request includes clustering criteria, which indicate the compositional structure of the event sequence. Specifically, the content delivery account can instruct the generation of the event sequence based on all or part of the operational and performance data, and the aforementioned clustering criteria indicate which data from the operational and performance data are used.

[0101] In this embodiment, the clustering process of the recommended content requires clustering based on data indicated by clustering criteria. Illustratively, the event sequence is determined by the clustering criteria to be a sequence composed of a first type of data and a second type of data, wherein the first type of data comes from operational data and the second type of data comes from effect data.

[0102] In some embodiments, the clustering criteria can be implemented as at least one of the following combinations: a, operation + effect; b, operation + effect + time; c, operation + effect + time + operation magnitude; d, operation + effect + time + effect category; e, operation + effect + time + operation magnitude + effect category.

[0103] For illustrative purposes, the target number of recommended content can be all recommended content placed by the content placement account on the content placement platform, or it can be recommended content selected from all recommended content. For example, based on the closing time of the recommended content, recommended content with a relatively long closing time can be filtered out, and recommended content with a recent placement period can be selected as the target number of recommended content.

[0104] In other embodiments, the request to obtain the report may also be issued by a management account with specified management permissions in the content delivery platform.

[0105] In some embodiments, the report retrieval request also includes content identifiers corresponding to the target number of recommended content items. That is, the server can determine the recommended content items that need to be clustered by parsing the report retrieval request. In some embodiments, when the target number of recommended content items is large, if the report retrieval request needs to carry the content identifiers of all recommended content items, then if the content identifiers are consecutive, the target number of recommended content items is indicated by indicating the first content identifier and the number of recommended content items.

[0106] Step 302: Obtain the operation data and effect data corresponding to the target number of recommended content based on the report acquisition request.

[0107] As an illustration, upon receiving a report retrieval request, the server first authenticates the request. If the request is deemed legitimate, it determines the target number of recommended content items indicated in the request and retrieves the corresponding operational and performance data from the database. The operational data indicates the settings for the recommended content delivery strategy, while the performance data indicates the effectiveness of the recommended content during the delivery process.

[0108] Step 303: Extract the first type of data from the operational data based on the clustering criteria.

[0109] The first type of data consists of operational data that meets the clustering criteria.

[0110] In some embodiments, after acquiring the operation data, the operation data undergoes a first preprocessing step to obtain an operation sequence. In one example, the operation sequence op_seq = <e1,e2,…,e n > where n is a positive integer, and each op_seq represents an operation sequence corresponding to a recommended content, with operation item e in the operation sequence. i = <t i o i c i >, where t i Indicates operation time information, o i Indicates the operation type, c i This represents the operation content, where i is a positive integer. Optionally, the above t... i It can be used to indicate the timestamp of an operation received, i.e., the actual operation time corresponding to the operation; or, t i It can be used to indicate the relative time between the creation timestamp of the recommended content and the timestamp of the operation; or, t i It can be used to indicate the interval between the current operation and the previous operation; or, t i It can be used to indicate the time difference of the timestamp of the operation relative to 0.

[0111] Schematic illustration: After processing operational data into an operational sequence, the first category of data can be extracted from the operational sequence based on clustering criteria. In one example, if the report request indicates clustering criterion a (operation + effect), then the extracted first category of data is derived from the operation sequence o. i The data is composed of...; in another example, when the report retrieves requests indicating clustering criteria b (operation + effect + time) or d (operation + effect + time + effect category), the extracted first category of data is composed of o in the operation sequence. i and t i The data is composed of the following: In another example, when the report request indicates clustering criteria c (operation + effect + time + operation magnitude) or e (operation + effect + time + operation magnitude + effect category), the operation sequence is directly used as the first type of data.

[0112] Step 304: Extract the second type of data from the effect data based on the clustering criteria.

[0113] The second type of data consists of data from the effect data that meets the clustering criteria.

[0114] In some embodiments, after obtaining the effect data, a second preprocessing step is performed on the effect data to obtain an effect sequence. In one example, the effect sequence effect_seq = <r1,r2,…,r n > where n is a positive integer, and one effect_seq represents the effect sequence corresponding to a recommended content, and the effect item r in the effect sequence. j = <t j d j e j >, where t j Indicates effect time information, d j Indicates the effect type, e j This represents the effect content, where j is a positive integer.

[0115] In some embodiments, the above e j Alternatively, the effect category can be obtained by categorizing the effect content. That is, the recommended content is categorized according to the effect content to obtain the effect category corresponding to the recommended content. Here, the effect category is related to the effect item r. j Correspondingly, a single recommended piece of content can correspond to different performance categories at different times.

[0116] In some embodiments, the effect categories described above can be obtained by inputting the effect content into a pre-trained effect classifier. Optionally, the effect classifier can be any model that can be used for text classification, such as a decision tree classifier, a Bayesian classifier, or a support vector machine (SVM).

[0117] In other embodiments, to enhance the classification effect of effect categories, the creation time of the recommended content can be introduced to further refine the resulting effect categories when classifying based on effect content. In one example, recommended content can be divided into the following categories based on effect content and creation time: a) Cold start content: created within 3 days and with fewer than 10 deep conversions; b) Stable performance content: deep conversions greater than 10 and daily clicks greater than or equal to 70% of the previous day's clicks; c) Declining performance content: deep conversions greater than 10 and daily clicks less than 70% of the previous day's clicks; d) Declining performance content: successful cold start, with fewer than 10 conversions in the last three days; e) No growth content: created more than 4 days ago and with fewer than 10 conversions. It is worth noting that the above effect categories are only illustrative and can be implemented using other category limitation methods depending on specific needs; these are not limited here.

[0118] Schematic illustration: After processing the effect data into an effect sequence, a second type of data can be extracted from the effect sequence based on clustering criteria. In one example, if the report retrieval request indicates clustering criterion a (operation + effect), then the extracted second type of data is d from the effect sequence. j The data is composed of...; in another example, when the report request indicates clustering criteria b (operation + effect + time) or c (operation + effect + time + operation magnitude), the extracted second type of data is composed of d from the effect sequence. j and t j The data is composed of two parts; in another example, when the report request indicates clustering criteria d (operation + effect + time + effect category) or e (operation + effect + time + operation magnitude + effect category), the effect sequence is directly used as the second type of data.

[0119] Optionally, steps 303 and 304 above can be executed in parallel or sequentially, and this is not limited here.

[0120] Step 305: Based on the temporal correlation between the first type of data and the second type of data on the time axis, the first type of data and the second type of data are interleaved to generate an event sequence.

[0121] In some embodiments, an event sequence is obtained by aligning operation time information and effect time information, thereby aligning the first type of data and the second type of data on the time axis. In one example, the process of generating the event sequence includes: extracting an operation sequence from the operation data; classifying recommended content based on the recommendation effect indicated by the effect data to obtain the effect category corresponding to the recommended content; extracting a first correspondence between operations and operation time information from the operation sequence; determining a second correspondence between effect category and effect time information based on the effect data; and interleaving the first type of data and the second type of data based on the first and second correspondences to generate the event sequence.

[0122] As an illustration, when aligning operation time information and effect time information, the operation time information and effect time information can be processed at the same level. That is, since the operation time information records the moment corresponding to the operation, while the effect time information records the time period corresponding to the effect, the same granularity of division is needed to align the operation time information and effect time information. In some embodiments, the time information is divided with the target duration as the granularity of division, and then the first type of data is processed according to the operation time information, and / or the second type of data is processed according to the effect information time, so that the granularity of time division between the data is the same. In one example, a target duration of 2 hours is used for division, that is, the first type of data belonging to the same time period is statistically analyzed to generate a new first type of data. For example, the operation with the operation timestamp of 12:30 is divided into the time period of 12:00 to 14:00. The processing of the second type of data is the same as that of the first type of data.

[0123] Then, the newly obtained first and second category data are used to generate an event sequence based on the aligned time. In one example, the event sequence event =<op,effect> Specifically, this can be implemented as event= <z1,z2,…,z n >, where, taking clustering criterion e (operation + effect + time + operation magnitude + effect category) as an example, event item zij = <o i d j , t ij c i e j >

[0124] Step 306: Cluster the event sequences of the target number to obtain the content clustering results.

[0125] To illustrate, before clustering event sequences, it is necessary to convert the generated event sequences into vector representations. That is, to perform vector transformation on the target number of event sequences to obtain the target number of event vector representations.

[0126] In a schematic manner, after obtaining the event vector representation of the target number of events, clustering is performed based on the event vector representation of the target number of events to obtain the content clustering results. Optionally, the clustering method used when clustering the event vector representation of the target number of events can be at least one of the following clustering methods: K-Means, DBSCAN, Hierarchical Methods, etc.

[0127] In some embodiments, clustering using the K-Means algorithm is used as an example for illustration. Optionally, when clustering event vector representations, the number of categories corresponding to the content clustering results can be indicated by the terminal device, preset by the system, or determined by the clustering profile coefficient. The clustering profile coefficient is used to indicate the density between event vector representations within each event category under the number of candidate categories.

[0128] In one example, when using cluster silhouette coefficients to determine the number of categories, the clustering process includes: S1, clustering based on the event vector representations of the target number to obtain the cluster distribution data of the event vectors; S2, obtaining the cluster silhouette coefficients corresponding to the number of n candidate categories under the cluster distribution data, where n is a positive integer; S3, dividing the cluster distribution data according to the number of candidate categories with the largest cluster silhouette coefficients to obtain the content clustering results. That is, the cluster silhouette coefficients indicate the density of the event vector representations within each cluster corresponding to each event category. A larger cluster silhouette coefficient indicates a tighter relationship between the event vector representations within the cluster, and vice versa.

[0129] In another example, when the number of categories is as indicated by the terminal device or preset by the system, the clustering process of the event sequence includes: S1, converting the event sequence into an event vector representation using a sequence vectorization model; S2, randomly creating k points as initial cluster centers, where k is a positive integer and k is the number of categories obtained; S3, for any recommended content's event vector representation, calculating its distance to the k cluster centers, and classifying the event vector representation into the cluster with the smallest distance, iterating n times; S4, during each iteration, updating the cluster centers of each cluster using methods such as the mean; S5, for the k cluster centers, after iterative updates using steps S3 and S4, if the position changes very little (which can be judged based on a preset position threshold), it is considered to have reached a stable state, and the iteration ends.

[0130] Indicative, such as Figure 4As shown, it illustrates a schematic diagram of the clustering process of recommended content provided by an exemplary embodiment of this application, wherein operation data 411 and effect data 421 are processed to obtain operation sequence 412 and effect sequence 422, event sequence extraction is performed on operation sequence 412 and effect sequence 422 to generate event sequence 431, event sequence 431 is processed to obtain event vector representation 432, and event vector representation 432 is clustered to obtain content clustering result 433.

[0131] Optionally, the content clustering results can be presented in the analysis report as either a content projection map or a category partitioning list. The content projection map uses the recommended content as projection nodes, displaying the clustering results based on the density of different projection nodes. The category partitioning list categorizes the recommended content into different event categories and indicates the recommended content included in each event category.

[0132] Step 307: Generate an analysis report based on the content clustering results and send the analysis report to the terminal device.

[0133] As an illustration, after receiving the content clustering results, the terminal device displays the content clustering results in the target application corresponding to the content delivery platform.

[0134] In some embodiments, the analysis report also includes textual descriptions. Besides providing clustering functionality for recommended content, the content delivery platform can also generate textual descriptions corresponding to the event sequences of recommended content upon request. That is, in response to receiving a request to generate a description of target content, it obtains the target event sequence corresponding to the target content, generates textual descriptions corresponding to the target event sequence, where the target content is any content from the target number of recommended content items, and the textual descriptions express the target event sequence in natural language. For example, recommended content A, which was initially launched with a budget increase at 9:00 AM, then operated at a normal, stable volume. Then, at 12:00 PM, the budget for recommended content A was decreased, and then recommended content A experienced a drop in volume.

[0135] In summary, the recommended content analysis method provided in this application obtains operational and performance data of a target number of recommended content during the delivery phase, generates corresponding event sequences based on the correlation between the operational and performance data, clusters these event sequences as clustering targets, and generates analysis reports based on the content clustering results. In other words, the content clustering results indicate the distribution pattern of recommended content performance under changing delivery strategies. Recommended content with similar distribution patterns can be clustered into the same category, thus improving the efficiency of analyzing recommended content-related data.

[0136] In this embodiment of the application, event sequences with different data combinations are obtained by using clustering criteria indicated by the terminal device, thereby enriching the diversity of content clustering results and meeting the needs of content providers for different analytical data.

[0137] In other embodiments, the clustering criteria indicated by the terminal device may be clustering criteria only for operational data or effect data. In one example, the clustering criteria indicate clustering by operation type. The server generates an operation sequence based on the operation type in the operation data, and performs clustering based on the operation sequence to obtain the content clustering result. In another example, the clustering criteria indicate clustering by operation type and operation time. The server generates an operation sequence based on the operation type and operation time information in the operation data, and performs clustering based on the operation sequence. This is illustrative; the above process of clustering using one type of data is only an example of operational data and can also be implemented for clustering other types of data. The specific settings can be configured according to actual needs and are not limited here.

[0138] Please refer to Figure 5 This illustration shows a method for analyzing recommended content according to an embodiment of this application. Indicatively, this method is applied to a terminal device running a target application corresponding to a content delivery platform. The method includes:

[0139] Step 501: Display the platform interface provided by the content delivery platform.

[0140] For illustrative purposes, the above platform interface is the application interface of the target application corresponding to the content delivery platform. The platform interface provides functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery line number is logged in on the above content delivery platform.

[0141] Content providers can modify the delivery strategy for recommended content through the delivery strategy settings. For example, the delivery strategy can be adjusted by setting the attribute information of the recommended content. Optionally, the attribute information of the recommended content includes at least one of the following: content information, title information, tag information, keyword information, budget information, bidding information, and delivery time information.

[0142] Optionally, the delivery strategy used for recommended content can be customized by the content provider, that is, the content provider modifies the attribute information of the recommended content to modify the delivery strategy; or it can be selected from candidate delivery strategies, that is, the content delivery platform provides at least two candidate delivery strategies, and in the process of setting the delivery strategy for the target recommended content, in response to receiving the selection operation of the candidate delivery strategy, the candidate delivery strategy is determined as the delivery strategy corresponding to the target recommended content.

[0143] As an illustration, modifications made by content providers to the recommendation strategy are recorded by the content delivery platform, thus obtaining the operation data. At the same time, the content delivery platform also records the corresponding performance data of the recommended content during the delivery process.

[0144] Step 502: Receive the content delivery account's instructions for delivery analysis on the platform interface.

[0145] The delivery analysis operation instruction analyzes multiple recommended content items based on the historical delivery strategies set in the delivery strategy settings function. These multiple recommended content items are the content published by the content delivery account.

[0146] Optionally, the above-mentioned delivery analysis operation can be triggered by the clustering control provided in the platform interface or by a preset shortcut key, and there is no limitation here.

[0147] This is illustrative of a report retrieval request generated based on a delivery analysis operation. The report retrieval request requests the server to cluster a target number of recommended content items. In some embodiments, during the triggering of the delivery analysis operation, candidate recommended content items are displayed on the platform interface. These candidate recommended content items can be content delivered by a content delivery account logged into the target application, or recommended content pre-filtered based on specified filtering criteria (e.g., creation time). In response to selecting a target number of recommended content items from the candidate recommended content, a report retrieval request is sent to the server, instructing the selected recommended content items to be clustered.

[0148] Since the clustering process for a target number of recommended content items requires the use of operational and performance data corresponding to the recommended content, in some embodiments, after receiving the delivery analysis operation, it is also necessary to receive an authorization operation for the operational and performance data. For example, in response to receiving the delivery analysis operation, a data authorization prompt is displayed. This prompt indicates that the content clustering process requires the use of operational and performance data corresponding to the recommended content; that is, it instructs the content provider on the data to be obtained and its specific purpose. In response to receiving a confirmation operation for this authorization prompt, a report retrieval request is sent to the server.

[0149] In other embodiments, the report retrieval request may also be sent automatically to the server based on the application status of the target application. That is, in response to receiving the activation of the recommendation effect analysis function in the target application, a report retrieval request is automatically sent to the server.

[0150] Step 503: Display analysis reports corresponding to multiple recommended content items based on the delivery analysis operation.

[0151] For illustrative purposes, the above analysis report includes content clustering results, which indicate the distribution pattern of the recommendation effect of the target number of recommended content under historical delivery strategies. Specifically, the content clustering results are obtained by clustering the event sequences corresponding to the target number of recommended content. The event sequences are obtained by interleaving the operation data and effect data based on the temporal correlation between the operation data and effect data corresponding to the recommended content. The event sequences indicate the temporal correlation between the operation data and effect data. The operation data indicates the settings of the content delivery account for historical delivery strategies, and the effect data indicates the recommendation effect of the recommended content during the delivery process. For illustrative purposes, the server's clustering process for recommended content is the same as steps 201-203 and steps 301-307, and will not be described in detail here.

[0152] Indicatively, in response to receiving the content clustering results returned by the server based on the report retrieval request, the content clustering results are displayed.

[0153] Optionally, the analysis report can contain either a content projection diagram or a category list. The content projection diagram uses the recommended content as projection nodes, displaying the clustering results of the recommended content based on the density of different projection nodes. In one example, such as... Figure 6 As shown, a content projection diagram 600 provided in an exemplary embodiment of this application is illustrated. Each projection node 601 displayed in the content projection diagram 600 represents a recommended piece of content. In this content projection diagram 600, a target number of recommended pieces of content are divided into categories A 610, B 620, C 630, D 640, E 650, F 660, G 670, H 680, and I 690. The above category division list is used to categorize recommended content into different event categories and indicates the recommended content included in each event category.

[0154] In some embodiments, the content projection map and category division list are displayed in different areas of the platform interface. Illustratively, the content projection map indicating the content clustering results is displayed in a first area of ​​the platform interface, indicating the density of the target number of recommended content after clustering; the category division list indicating the content set after the target number of recommended content is divided into at least two event categories is displayed in a second area of ​​the platform interface. Optionally, the first and second areas are different areas within the same interface.

[0155] In one example, such as Figure 7 As shown, it illustrates a display diagram of the content clustering results provided by an exemplary embodiment of this application. A content projection map is displayed in the first area 710 of the platform interface 700, and a category division list is displayed in the second area 720. The recommended content in the category division list corresponds one-to-one with the projection nodes in the content projection map.

[0156] In some embodiments, the category lists corresponding to different event categories in the second region can be arranged according to the number of recommended content items in the lists. In one example, the category lists corresponding to event categories are sorted in descending order based on the number of recommended content items for each event category.

[0157] In some embodiments, when the number of recommended items included in the category list corresponding to an event category is large, the category list can be displayed in a collapsed manner, and a scrollbar can be set in the category list. That is, a first number of event sequences are displayed in the category list, the remaining event sequences are hidden, and other event sequences are displayed in response to a scrollbar receiving a scroll operation. Optionally, the scrollbar can also be implemented as a collapse control, which is not specifically limited here.

[0158] In some embodiments, the correspondence between recommended content in the content projection map and recommended content in the category classification list can be displayed by selecting a projection node in the content projection map or by selecting recommended content in the category classification list.

[0159] In one example, in response to receiving a selection operation for the target recommended content in the second region, the target event sequence corresponding to the target recommended content is highlighted in the second region, while the projection node corresponding to the target recommended content is highlighted in the first region.

[0160] In some embodiments, when a target recommended content is selected, its content details are displayed in a third area of ​​the platform interface. Optionally, the content details may include at least one of the following: the current delivery status of the target recommended content, historical operation data, performance data, and attribute information.

[0161] like Figure 8 The diagram illustrates a platform interface 800 provided in an exemplary embodiment of this application. In the first area of ​​the platform interface 800, a content projection map 810 is displayed, highlighting the projection node 801 corresponding to the target recommended content. In the second area, a category classification list 820 is displayed, highlighting the event sequence 802 corresponding to the target recommended content. In the third area, content detail information 830 corresponding to the target recommended content is displayed.

[0162] In summary, the recommended content analysis method provided in this application provides visualization of the joint analysis process of operation data and effect data by instructing the delivery analysis operation in the platform interface provided by the content delivery platform and displaying the content clustering results in the platform interface. Content providers can intuitively obtain the content clustering results from the platform interface, thereby improving the efficiency of data analysis.

[0163] It should be noted that all information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in this application have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the operational data and effect data involved in this application were obtained with full authorization.

[0164] Please refer to Figure 9 The diagram illustrates a structural block diagram of an analysis apparatus for recommended content provided in an exemplary embodiment of this application. The apparatus includes the following modules:

[0165] The acquisition module 910 is used to acquire operation data and effect data corresponding to multiple recommended contents. The multiple recommended contents are the contents published by the content delivery account. The operation data is used to indicate the setting of the delivery strategy of the content delivery account for the recommended contents. The effect data is used to indicate the recommendation effect of the recommended contents in the delivery process.

[0166] The first generation module 920 is used to interleave the operation data and the effect data based on the temporal correlation between the operation data and the effect data to generate an event sequence corresponding to the recommended content;

[0167] Clustering module 930 is used to cluster the event sequences corresponding to the multiple recommended contents to obtain content clustering results. The content clustering results are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the delivery strategy.

[0168] The second generation module 940 is used to generate an analysis report corresponding to the multiple recommended contents based on the content clustering results.

[0169] In some alternative embodiments, such as Figure 10 As shown, the clustering module 930 further includes:

[0170] The conversion unit 931 is used to perform vector conversion on the event sequence to obtain an event vector representation;

[0171] Clustering unit 932 is used to perform clustering based on the event vector representations corresponding to the multiple recommended contents to obtain the content clustering results.

[0172] In some optional embodiments, the clustering unit 932 is further configured to perform clustering based on the event vector representations corresponding to the plurality of recommended contents, to obtain clustering distribution data of the event vector representations;

[0173] The acquisition unit 933 is used to acquire the clustering profile coefficients corresponding to the number of n candidate categories under the clustering distribution data. The clustering profile coefficients are used to indicate the density between event vector representations within each event category under the number of candidate categories, where n is a positive integer.

[0174] The determining unit 934 is used to divide the clustering distribution data according to the number of candidate categories with the largest clustering profile coefficients to obtain the content clustering results.

[0175] In some alternative embodiments, the apparatus further includes:

[0176] The receiving module 950 is used to receive a report retrieval request indicated by the content delivery account. The report retrieval request is used to request clustering of a target number of recommended content. The report retrieval request includes clustering criteria, which are used to indicate the composition structure of the event sequence.

[0177] The first generation module 920 includes:

[0178] Extraction unit 921 is used to extract a first type of data from the operation data based on the clustering criteria, wherein the first type of data is the data in the operation data that conforms to the clustering criteria;

[0179] The extraction unit 921 is further configured to extract a second type of data from the effect data based on the clustering criteria, wherein the second type of data is the data in the effect data that conforms to the clustering criteria;

[0180] The generation unit 922 is used to interleave the first type of data and the second type of data based on the temporal correlation between the first type of data and the second type of data on the time axis to generate the event sequence.

[0181] In some optional embodiments, when the clustering criteria indicate clustering based on the recommendation effect corresponding to the operation of the content delivery account,

[0182] The extraction unit 921 is further configured to extract an operation sequence from the operation data, the operation sequence being used to indicate the order of operations recorded in the operation data for modifying the delivery strategy;

[0183] The extraction unit 921 is further configured to classify the recommended content based on the recommendation effect indicated by the effect data, and obtain the effect category corresponding to the recommended content;

[0184] The generation unit 922 is further configured to extract a first correspondence between the operation and the operation time information from the operation sequence, wherein the operation time information is used to indicate the time point at which the operation was generated;

[0185] The generation unit 922 is further configured to determine a second correspondence between the effect category and the effect time information based on the effect data, wherein the effect time information is the time period corresponding to the recommended effect;

[0186] The generation unit 922 is further configured to interleave the first type of data and the second type of data based on the first correspondence and the second correspondence to generate the event sequence.

[0187] In some optional embodiments, the acquisition module 910 is further configured to, in response to receiving a description generation request for the target content, acquire a target event sequence corresponding to the target content, wherein the target content is any one of the plurality of recommended content;

[0188] The second generation module 940 is further configured to generate text description content corresponding to the target event sequence. The text description content is the content of the target event sequence expressed in natural language. The analysis report includes the text description content.

[0189] In summary, the recommended content analysis device provided in this application obtains operational and performance data of a target number of recommended contents during the delivery phase, generates corresponding event sequences based on the correlation between the operational and performance data, clusters the event sequences as clustering targets, and generates an analysis report based on the content clustering results. That is, the content clustering results indicate the distribution pattern of recommended content performance under changes in delivery strategies. Recommended contents with similar distribution patterns can be clustered into the same category, which can improve the analysis efficiency of recommended content-related data.

[0190] Please refer to Figure 11 The diagram illustrates a structural block diagram of an analysis apparatus for recommended content provided in an exemplary embodiment of this application. The apparatus includes the following modules:

[0191] Display module 1110 is used to display the platform interface provided by the content delivery platform. The platform interface provides functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery platform is logged in with a content delivery account.

[0192] The receiving module 1120 is used to receive the delivery analysis operation indicated by the content delivery account in the platform interface. The delivery analysis operation indicates that multiple recommended contents are analyzed according to the historical delivery strategy set by the delivery strategy setting function. The multiple recommended contents are the content published by the content delivery account.

[0193] The display module 1110 is also used to display an analysis report corresponding to the multiple recommended contents based on the delivery analysis operation. The analysis report includes content clustering results, which are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the historical delivery strategy.

[0194] The content clustering result is obtained by clustering the event sequences corresponding to the multiple recommended contents. The event sequence is obtained by interleaving the operation data and the effect data according to the temporal correlation between the operation data and the effect data corresponding to the recommended content. The event sequence is used to indicate the temporal correlation between the operation data and the effect data. The operation data is used to indicate the settings of the historical delivery strategy of the content delivery account. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

[0195] In some alternative embodiments, the apparatus further includes:

[0196] The sending module (not shown in the figure) is used to send a report retrieval request to the server based on the delivery analysis operation. The report retrieval request is used to request the server to analyze the operation data and the effect data based on the target number of recommended content.

[0197] The display module 1110 is further configured to, in response to receiving the analysis report returned by the server based on the report acquisition request, display the analysis report corresponding to the target number of recommended contents.

[0198] In some optional embodiments, the display module 1110 is further configured to display the content projection map indicated by the content clustering result in a first area of ​​the platform interface, wherein the content projection map is used to indicate the density of the multiple recommended contents after clustering.

[0199] The display module 1110 is further configured to display the category division list indicated by the content clustering results in a second area of ​​the platform interface, wherein the category division list is used to indicate the content set after the multiple recommended contents are divided into at least two event categories.

[0200] In some optional embodiments, the display module 1110 is further configured to, in response to receiving a selection operation for the target recommended content in the second region, highlight the target event sequence corresponding to the target recommended content in the second region;

[0201] The display module 1110 is also used to highlight the projection node corresponding to the target recommended content in the first area.

[0202] In summary, the recommended content analysis device provided in this application provides visualization of the joint analysis process of operation data and effect data by instructing the delivery analysis operation in the platform interface provided by the content delivery platform and displaying the content clustering results in the platform interface. Content providers can intuitively obtain the content clustering results from the platform interface, thereby improving the efficiency of data analysis.

[0203] It should be noted that the recommended content analysis device provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the recommended content analysis device and the recommended content analysis method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0204] Figure 12This illustration shows a schematic diagram of the structure of a server provided in an exemplary embodiment of this application. Specifically, it includes the following structure.

[0205] Server 1200 includes a Central Processing Unit (CPU) 1201, a system memory 1204 including Random Access Memory (RAM) 1202 and Read Only Memory (ROM) 1203, and a system bus 1205 connecting the system memory 1204 and the CPU 1201. Server 1200 also includes a mass storage device 1206 for storing an operating system 1213, application programs 1214, and other program modules 1215.

[0206] Mass storage device 1206 is connected to central processing unit 1201 via a mass storage controller (not shown) connected to system bus 1205. Mass storage device 1206 and its associated computer-readable media provide non-volatile storage for server 1200. That is, mass storage device 1206 may include computer-readable media (not shown) such as hard disk or compact disc read-only memory (CD-ROM) drives.

[0207] Without loss of generality, computer-readable media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes RAM, ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other solid-state memory technologies, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic tape cassettes, magnetic tape, disk storage, or other magnetic storage devices. Of course, those skilled in the art will recognize that computer storage media are not limited to the above-mentioned types. The system memory 1204 and mass storage device 1206 described above can be collectively referred to as memory.

[0208] According to various embodiments of this application, server 1200 can also be connected to a remote computer on a network, such as the Internet. That is, server 1200 can be connected to network 1212 via network interface unit 1211 connected to system bus 1205, or network interface unit 1211 can be used to connect to other types of networks or remote computer systems (not shown).

[0209] The aforementioned memory also includes one or more programs, which are stored in the memory and configured to be executed by the CPU.

[0210] Figure 13 A structural block diagram of a terminal 1300 provided in an exemplary embodiment of this application is shown. The terminal 1300 may be a smartphone, tablet computer, Moving Picture Experts Group Audio Layer III (MP3) player, Moving Picture Experts Group Audio Layer IV (MP4) player, laptop computer, or desktop computer. The terminal 1300 may also be referred to as a user device, portable terminal, laptop terminal, desktop terminal, or other names.

[0211] Typically, terminal 1300 includes a processor 1301 and a memory 1302.

[0212] Processor 1301 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1301 may be implemented using at least one hardware form selected from Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). Processor 1301 may also include a main processor and a coprocessor. The main processor, also known as a central processing unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1301 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 1301 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0213] The memory 1302 may include one or more computer-readable storage media, which may be non-transitory. The memory 1302 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1302 are used to store at least one instruction, which is executed by the processor 1301 to implement the virtual game-based control method provided in the method embodiments of this application.

[0214] This is illustrative; terminal 1300 also includes other components, as those skilled in the art will understand. Figure 13 The structure shown does not constitute a limitation on terminal 1300 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0215] Embodiments of this application also provide a computer device including a processor and a memory. The memory stores at least one instruction, at least one program, code set, or instruction set. The processor loads and executes the at least one instruction, at least one program, code set, or instruction set to implement the analysis method of the recommended content provided in the above-described method embodiments. Optionally, the computer device may be a terminal or a server.

[0216] Embodiments of this application also provide a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the analysis method of the recommended content provided in the above-described method embodiments.

[0217] Embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the analysis method for the recommended content described in any of the above embodiments.

[0218] Optionally, the computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. The random access memory may include resistive random access memory (ReRAM) and dynamic random access memory (DRAM). The sequence numbers of the embodiments in this application are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0219] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0220] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for analyzing recommended content, characterized in that, The method includes: The system acquires operation data and effect data corresponding to multiple recommended content items, where the multiple recommended content items are content published by content delivery accounts. The operation data is used to indicate the content delivery account's settings for the delivery strategy of the recommended content items, and the effect data is used to indicate the recommendation effect of the recommended content items during the delivery process. Based on the temporal correlation between the operation data and the effect data, the operation data and the effect data are interleaved to generate an event sequence corresponding to the recommended content; Clustering is performed on the event sequences corresponding to the multiple recommended contents to obtain content clustering results. The content clustering results are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the delivery strategy. An analysis report is generated based on the content clustering results, corresponding to the multiple recommended contents.

2. The method according to claim 1, characterized in that, The step of clustering the event sequences corresponding to the multiple recommended contents to obtain content clustering results includes: The event sequence is transformed into a vector representation to obtain an event vector representation; Clustering is performed based on the event vector representations corresponding to the multiple recommended contents to obtain the content clustering results.

3. The method according to claim 2, characterized in that, The clustering based on the event vector representations corresponding to the multiple recommended contents to obtain the content clustering results includes: Clustering is performed based on the event vector representations corresponding to the multiple recommended contents to obtain the clustering distribution data of the event vector representations; Obtain the clustering profile coefficients corresponding to the number of n candidate categories under the clustering distribution data. The clustering profile coefficients are used to indicate the density between event vector representations within each event category under the number of candidate categories, where n is a positive integer. The clustering distribution data is divided according to the number of candidate categories with the largest clustering profile coefficients to obtain the content clustering results.

4. The method according to any one of claims 1 to 3, characterized in that, Before obtaining the operation data and effect data corresponding to multiple recommended contents, the process also includes: The system receives a report retrieval request from the content delivery account. The report retrieval request is used to request the analysis of a target number of recommended content. The report retrieval request includes clustering criteria, which are used to indicate the compositional structure of the event sequence. The step of generating an event sequence corresponding to the recommended content based on the operation data and the effect data includes: Based on the clustering criteria, a first type of data is extracted from the operational data, wherein the first type of data is the data in the operational data that conforms to the clustering criteria; Based on the clustering criteria, a second type of data is extracted from the effect data, wherein the second type of data is the data in the effect data that conforms to the clustering criteria; Based on the temporal correlation between the first type of data and the second type of data on the time axis, the first type of data and the second type of data are interleaved to generate the event sequence.

5. The method according to claim 4, characterized in that, When the clustering criteria indicate that clustering should be performed based on the recommendation effect corresponding to the operation of the content delivery account, The extraction of the first type of data from the operational data based on the clustering criteria includes: An operation sequence is extracted from the operation data, and the operation sequence is used to indicate the order of operations recorded in the operation data to modify the delivery strategy; The extraction of the second type of data from the effect data based on the clustering criteria includes: Based on the recommendation effect indicated by the effect data, the recommended content is classified to obtain the effect category corresponding to the recommended content; The step of generating the event sequence by interleaving the first type of data and the second type of data based on their temporal correlation on the time axis includes: A first correspondence between operations and operation time information is extracted from the operation sequence, wherein the operation time information is used to indicate the time point in which the operation was generated; Based on the effect data, a second correspondence is determined between the effect category and the effect time information, wherein the effect time information is the time period corresponding to the recommended effect; Based on the first correspondence and the second correspondence, the first type of data and the second type of data are interleaved to generate the event sequence.

6. The method according to any one of claims 1 to 3, characterized in that, The method further includes: In response to receiving a request to generate a description of target content, the system obtains a target event sequence corresponding to the target content, wherein the target content is any one of the multiple recommended contents; Generate text description content corresponding to the target event sequence. The text description content will express the target event sequence in natural language form. The analysis report includes the text description content.

7. A method for analyzing recommended content, characterized in that, The method includes: The content delivery platform displays the platform interface provided by the platform, which offers functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery platform is logged into with a content delivery account. The platform interface receives a delivery analysis operation from the content delivery account. The delivery analysis operation instructs the analysis of multiple recommended contents based on the historical delivery strategies set by the delivery strategy setting function. The multiple recommended contents are the content published by the content delivery account. Based on the aforementioned delivery analysis operation, an analysis report corresponding to the multiple recommended contents is displayed. The analysis report includes content clustering results, which are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the historical delivery strategy. The content clustering result is obtained by clustering the event sequences corresponding to the multiple recommended contents. The event sequence is obtained by interleaving the operation data and the effect data according to the temporal correlation between the operation data and the effect data corresponding to the recommended content. The event sequence is used to indicate the temporal correlation between the operation data and the effect data. The operation data is used to indicate the settings of the historical delivery strategy of the content delivery account. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

8. The method according to claim 7, characterized in that, The analysis report displayed based on the delivery analysis operation includes: Based on the aforementioned delivery analysis operation, a report retrieval request is sent to the server. The report retrieval request is used to request the server to analyze the operation data and the effect data based on a target number of recommended content. In response to receiving the analysis report returned by the server based on the report retrieval request, the analysis report corresponding to the target number of recommended contents is displayed.

9. The method according to claim 7, characterized in that, The analysis report displaying the multiple recommended contents includes: The content projection map indicated by the content clustering results is displayed in the first area of ​​the platform interface. The content projection map is used to indicate the density of the multiple recommended contents after clustering. The category division list indicated by the content clustering results is displayed in the second area of ​​the platform interface. The category division list is used to indicate the content set after the multiple recommended contents are divided into at least two event categories.

10. The method according to claim 9, characterized in that, The method further includes: In response to receiving a selection operation for target recommended content in the second area, the target event sequence corresponding to the target recommended content is highlighted in the second area; The projection node corresponding to the target recommended content is highlighted in the first area.

11. An analysis device for recommended content, characterized in that, The device includes: The acquisition module is used to acquire operation data and effect data corresponding to multiple recommended content. The multiple recommended content are content published by content delivery accounts. The operation data is used to indicate the setting of the delivery strategy of the content delivery account for the recommended content. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process. The first generation module is used to interleave the operation data and the effect data based on the temporal correlation between the operation data and the effect data to generate an event sequence corresponding to the recommended content; The clustering module is used to cluster the event sequences corresponding to the multiple recommended contents to obtain content clustering results. The content clustering results are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the delivery strategy. The second generation module is used to generate an analysis report corresponding to the multiple recommended contents based on the content clustering results.

12. An analysis device for recommended content, characterized in that, The device includes: The display module is used to display the platform interface provided by the content delivery platform. The platform interface provides functions for setting delivery strategies for recommended content and analyzing recommendation effects. The content delivery platform is logged in with a content delivery account. The receiving module is used to receive the delivery analysis operation indicated by the content delivery account in the platform interface. The delivery analysis operation indicates that multiple recommended contents are analyzed according to the historical delivery strategy set by the delivery strategy setting function. The multiple recommended contents are the content published by the content delivery account. The display module is also used to display an analysis report corresponding to the multiple recommended contents based on the delivery analysis operation. The analysis report includes content clustering results, which are used to indicate the distribution pattern of the recommendation effect of the multiple recommended contents under the historical delivery strategy. The content clustering result is obtained by clustering the event sequences corresponding to the multiple recommended contents. The event sequence is obtained by interleaving the operation data and the effect data according to the temporal correlation between the operation data and the effect data corresponding to the recommended content. The event sequence is used to indicate the temporal correlation between the operation data and the effect data. The operation data is used to indicate the settings of the historical delivery strategy of the content delivery account. The effect data is used to indicate the recommendation effect of the recommended content in the delivery process.

13. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one program, which is loaded and executed by the processor to implement the method for analyzing recommended content as described in any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one piece of program code, which is loaded and executed by a processor to implement the method for analyzing recommended content as described in any one of claims 1 to 10.

15. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the method for analyzing recommended content as described in any one of claims 1 to 10.

Images