Personalized advertisement intelligent recommendation system based on big data and control method

Abstract

The invention relates to the technical field of computers, and provides a personalized intelligent advertisement recommendation system based on big data and a control method, the system comprises an intelligent advertisement recommendation middle table, an acquisition module, a scene recognition module, an advertisement matching module, an advertisement sorting module and an advertisement recommendation module; the acquisition module acquires a user interest label of a target user; the scene recognition module performs scene recognition based on the current use scene information of the target user to obtain a user scene; the advertisement matching module matches the user interest label and the user scene with advertisement feature information extracted from the advertisement to obtain a to-be-recommended advertisement; the advertisement sorting module sorts the to-be-recommended advertisements based on the matching degree of the user interest labels and the user scenes to obtain target recommended advertisements; and the advertisement recommendation module displays the target recommended advertisement on a display interface of the terminal equipment of the target user. According to the embodiment of the invention, the recommended advertisement accurately hits the demand of the user, and the conversion rate of the advertisement and the user experience are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a personalized advertising intelligent recommendation system based on big data and a control method. Background Art

[0002] In today's digital advertising era, intelligent advertising recommendation has become an important means for companies to promote their products and services. An existing common intelligent advertising recommendation method mainly makes recommendations based on the user's basic information (such as age, gender, region, etc.) and historical browsing behavior. The system will pre-set some rules to match advertisements with users based on these user characteristics. For example, if the user is a young woman and often browses fashion web pages, the system will recommend fashion clothing, beauty and other related advertisements to her.

[0003] However, the recommendation accuracy of existing methods is limited. They only rely on basic information and historical browsing behavior to make recommendations, and cannot deeply understand the user's real interests and potential needs. Because the user's browsing behavior may be affected by many factors, such as accidental clicks, recommendations from others, etc., it may not fully reflect their true preferences. Moreover, the division of basic information is relatively broad, and users in the same category may have great differences in interests, resulting in the recommended ads often not accurately hitting the user's needs, thereby reducing the conversion rate of ads and user experience. Summary of the invention

[0004] The present invention provides a personalized advertising intelligent recommendation system and control method based on big data, aiming to achieve accurate hit of recommended advertisements to meet user needs and improve advertisement conversion rate and user experience.

[0005] In a first aspect, the present invention provides a personalized advertising intelligent recommendation system based on big data, including an advertising intelligent recommendation middle station, an acquisition module, a scene recognition module, an advertising matching module, an advertising sorting module and an advertising recommendation module; the advertising intelligent recommendation middle station is respectively connected to the acquisition module, the scene recognition module, the advertising matching module, the advertising sorting module and the advertising recommendation module to manage the data of each module;

[0006] An acquisition module is used to respond to a platform login instruction of a target user and acquire a user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user;

[0007] A scene recognition module, used to perform scene recognition based on the current usage scene information of the target user to obtain the user scene in which the target user is currently located;

[0008] An advertisement matching module is used to match user interest tags and user scenarios with advertisement feature information extracted from advertisements to obtain advertisements to be recommended to the target user;

[0009] An advertisement sorting module is used to sort the advertisements to be recommended based on the matching degree between the user interest tags and the user scenarios to obtain target recommended advertisements;

[0010] The advertisement recommendation module is used to display the target recommended advertisement on the display interface of the terminal device of the target user.

[0011] In a second aspect, the present invention further provides a method for controlling a personalized advertising intelligent recommendation based on big data, which is implemented based on the personalized advertising intelligent recommendation system based on big data described in the first aspect. The method for controlling a personalized advertising intelligent recommendation based on big data includes:

[0012] Responding to a platform login instruction of a target user, obtaining a user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user;

[0013] Performing scene recognition based on the current usage scene information of the target user to obtain the current user scene of the target user;

[0014] Matching the user interest tags and user scenarios with the advertisement feature information extracted from the advertisement to obtain the advertisement to be recommended for the target user;

[0015] Sort the recommended ads based on the matching degree between the user interest tags and the user scenarios to obtain the target recommended ads;

[0016] The target recommended advertisement is displayed on the display interface of the terminal device of the target user.

[0017] In a third aspect, the present invention further provides an electronic device, comprising: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby implementing any of the above-mentioned methods for intelligent recommendation control of personalized advertising based on big data.

[0018] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, wherein a computer software program is stored in the storage medium, and when the computer software program is executed by a processor, the method for controlling personalized advertising intelligent recommendation based on big data as described above is implemented.

[0019] In a fifth aspect, the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements any one of the above-mentioned personalized advertising intelligent recommendation control methods based on big data.

[0020] The personalized advertising intelligent recommendation system based on big data provided by the embodiment of the present invention matches the advertising feature information extracted from the advertisement according to the user interest tags and user scenarios of the users, and not only takes into account the user's interests, but also takes into account the impact of different scenarios on user needs, so that the matched advertisements can more accurately fit the user's preferences and needs in specific scenarios, and further sort the advertisements according to the degree of matching between the user's user interest tags and the user scenarios, so that the final advertisements can more accurately fit the current scenario needs. Therefore, the advertisements can not only accurately fit the user's preference needs, but also accurately fit the user's scenario needs, so that the recommended advertisements can accurately hit the user's needs, thereby improving the conversion rate of the advertisements and the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a structural diagram of a personalized advertising intelligent recommendation system based on big data provided by the present invention;

[0022] Figure 2 It is a flow chart of a method for intelligent recommendation of personalized advertisements based on big data provided by the present invention;

[0023] Figure 3 An embodiment diagram of an electronic device provided by an embodiment of the present invention;

[0024] Figure 4 An embodiment diagram of a computer-readable storage medium provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0026] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0027] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0028] Optional, see Figure 1 As shown, Figure 1 It is a structural diagram of the personalized advertising intelligent recommendation system based on big data provided by the present invention. The personalized advertising intelligent recommendation system based on big data includes an advertising intelligent recommendation middle station, an acquisition module, a scene recognition module, an advertising matching module, an advertising sorting module and an advertising recommendation module, wherein the advertising intelligent recommendation middle station is respectively connected to the acquisition module, the scene recognition module, the advertising matching module, the advertising sorting module and the advertising recommendation module to manage the data of each module.

[0029] Optionally, when the target user performs a login operation on the platform associated with the advertising recommendation system, the system will receive the login instruction. The acquisition module will extract the user interest tags belonging to the target user from the user interest tag library that is pre-analyzed and generated based on user behavior data (such as user search behavior, browsing behavior, purchase behavior, and sharing behavior, etc.), where the interest tags can reflect the user's interest tendencies in different fields, such as the degree of interest in electronic products, fashion clothing, food, etc.

[0030] Optionally, the scene recognition module will collect the target user's current usage scene information, where the usage scene information may include the time when the user logs into the platform (such as daytime, nighttime), the type of device used by the user (such as mobile phone, computer, tablet), the user's geographical location (obtained through device positioning), the type of page the user is currently browsing (such as homepage, product details page, shopping cart page), etc. Further, based on the collected usage scene information, the scene recognition algorithm and rules are used to determine what specific usage scene the target user is currently in, such as shopping scene, leisure and entertainment scene, work scene, etc.

[0031] Optionally, the advertisement matching module analyzes various advertisements on the platform in advance, extracts characteristic information of the advertisements, such as the product categories promoted by the advertisements (such as clothing, electronic products, food, etc.), the applicable scenarios of the advertisements (such as sports scenarios, office scenarios, family scenarios, etc.), and the characteristics of the target audience of the advertisements (such as age range, gender, hobbies, etc.), and further compares and matches the user interest tags of the target users and the identified user scenarios with the characteristic information of the advertisements one by one. If the characteristic information of an advertisement matches or partially matches the user interest tags and user scenarios, then the advertisement will be screened out as a to-be-recommended advertisement for the target user.

[0032] Optionally, the advertisement sorting module calculates the degree of match between the characteristic information of each advertisement and the user interest tags and user scenarios of the target users, wherein the degree of match can be based on some specific rules and algorithms, such as the overlap of interest tags, the similarity between the user scenarios and the applicable scenarios of the advertisements, etc., and further sorts these advertisements to be recommended in order from high to low according to the degree of match. The advertisements at the front become the target recommended advertisements, which are considered to be the most in line with the current interests and scenarios of the target users.

[0033] Optionally, the advertising recommendation module sends the relevant content of the target recommended advertisement (such as advertising images, text descriptions, links, etc.) to the terminal device (such as a mobile phone, computer, tablet, etc.) used by the target user. After receiving the advertising content, the terminal device will display the target recommended advertisement on the display interface of the device according to the display rules set by the system, such as on the homepage of the e-commerce platform, in a specific advertising position on the product details page, etc., so that the target user can see these recommended advertisements.

[0034] The embodiments of the present invention match the advertisement feature information extracted from the advertisement according to the user interest tags and user scenarios of the users, and not only consider the user interests, but also consider the impact of different scenarios on the user needs, so that the matched advertisements can more accurately fit the user's preferences in specific scenarios, and further sort the advertisements according to the degree of matching between the user's user interest tags and the user scenarios, so that the final advertisements can more accurately fit the current scenario needs. Therefore, the advertisements can not only accurately fit the user's preferences, but also accurately fit the user's scenario needs, so that the recommended advertisements can accurately hit the user's needs, thereby improving the conversion rate of the advertisements and the user experience.

[0035] Optional, see Figure 2 , Figure 2The flowchart of the personalized advertising intelligent recommendation control method based on big data provided by the present invention is shown in FIG. The execution subject of the personalized advertising intelligent recommendation control method based on big data in the embodiment of the present invention is an advertising recommendation system. Therefore, the personalized advertising intelligent recommendation control method based on big data includes:

[0036] Step 10, responding to the platform login instruction of the target user, and obtaining the user interest tags of the target user.

[0037] Optionally, when the target user performs a login operation on the platform associated with the advertising recommendation system, the system will receive the login instruction. The system will then extract the user interest tags belonging to the target user from the user interest tag library that is pre-analyzed and generated based on user behavior data (such as the user's search behavior, browsing behavior, purchase behavior, and sharing behavior, etc.), where the interest tags can reflect the user's interest tendencies in different fields, such as the degree of interest in electronic products, fashion clothing, food, etc. The specific steps of generating the user interest tags of the target user based on the in-depth analysis of the user behavior data of the target user are described in steps 101 to 104.

[0038] In one embodiment, user A logs in on an e-commerce platform, and the advertising recommendation system receives the login instruction from user A. Previously, user A has browsed smartphone and tablet related products on the platform many times, clicked on the introduction pages of some new smartphones, and purchased smartphone accessories once. Based on the behavioral data, the system analyzes and generates user interest tags for user A, such as "smartphone".

[0039] "Tablet computer", "Mobile phone accessories", etc. When user A logs in to the platform, the advertising recommendation system will obtain these user interest tags belonging to user A.

[0040] Step 20: Perform scene recognition based on the target user's current usage scene information to obtain the target user's current user scene.

[0041] Furthermore, the advertising recommendation system will collect the target user's current usage scenario information, where the usage scenario information may include the time when the user logs into the platform (such as daytime or night), the type of device the user uses (such as mobile phone, computer, tablet), the user's geographic location (obtained through device positioning), the type of page the user is currently browsing (such as homepage, product details page, shopping cart page), etc.

[0042] Therefore, the advertising recommendation system uses the scene recognition algorithm and rules based on the collected usage scenario information to determine what specific usage scenario the target user is currently in, such as shopping scenario, leisure and entertainment scenario, work scenario, etc. In one embodiment, user A uses his mobile phone to log in to the e-commerce platform at 8 pm and is currently browsing the product details page of sportswear. The advertising recommendation system obtains the information that user A logged in at night, the device is a mobile phone, and he is browsing the product details page of sportswear. According to the scene recognition rules set by the system, it is determined that user A is currently in a leisure shopping scenario and may have the intention to buy sportswear.

[0043] Step 30, matching the user interest tag and the user scenario with the advertisement feature information extracted from the advertisement to obtain the advertisement to be recommended to the target user.

[0044] Furthermore, the advertising recommendation system will analyze various types of advertisements on the platform in advance and extract characteristic information of the advertisements, such as the product categories promoted by the advertisements (such as clothing, electronic products, food, etc.), the applicable scenarios of the advertisements (such as sports scenarios, office scenarios, family scenarios, etc.), and the characteristics of the target audience of the advertisements (such as age range, gender, interests and hobbies, etc.).

[0045] Furthermore, the advertisement recommendation system compares and matches the user interest tags and the identified user scenarios of the target users with the advertisement feature information one by one. If the feature information of an advertisement matches or partially matches the user interest tags and the user scenarios, then the advertisement will be screened out as the advertisement to be recommended for the target user, as described in steps 301 to 304.

[0046] In one embodiment, for user A, his user interest tag is "sportswear", the user scenario is a leisure shopping scenario at night, and he may have the intention to buy sportswear. The advertising recommendation system analyzes the advertisements on the platform. There is an advertisement promoting a certain brand of sports T-shirts. The product category in its feature information is "sportswear", the applicable scenario is "leisure sports scenario", and the target audience includes people interested in sportswear. The system matches the interest tag and user scenario of user A with the feature information of this advertisement and finds that the degree of conformity is high, so the advertisement promoting sports T-shirts is selected as the advertisement to be recommended for user A.

[0047] Step 40: sort the advertisements to be recommended based on the matching degree between the user interest tags and the user scenarios to obtain target recommended advertisements.

[0048] Furthermore, the advertising recommendation system calculates the degree of matching between the characteristic information of each advertisement and the user interest tags and user scenarios of the target users, wherein the degree of matching can be based on some specific rules and algorithms, such as the overlap of interest tags, the similarity between the user scenarios and the applicable scenarios of the advertisements, etc.

[0049] Furthermore, the advertisement recommendation system sorts these advertisements to be recommended in order of the degree of matching from high to low, and the advertisements in the front become the target recommended advertisements, which are considered to be the most suitable for the current interests and scenarios of the target users. Continuing with the above embodiment, in addition to the advertisements promoting sports T-shirts, there are two other advertisements to be recommended, one is an advertisement promoting sports running shoes, and the other is an advertisement promoting sports backpacks. The system calculates that the degree of matching between the advertisement promoting sports T-shirts and the user interest tag "sportswear" of user A and the leisure shopping scenario is 80%; the degree of matching between the advertisement promoting sports running shoes is 70%; and the degree of matching between the advertisement promoting sports backpacks is 60%. According to the sorting from high to low according to the degree of matching, the advertisement promoting sports T-shirts is ranked first, followed by the advertisement promoting sports running shoes, and finally the advertisement promoting sports backpacks. Then, the advertisement promoting sports T-shirts becomes the target recommended advertisement for user A.

[0050] Step 50: Display the target recommended advertisement on the display interface of the terminal device of the target user.

[0051] Furthermore, the advertising recommendation system sends the relevant content of the target recommended advertisement (such as advertising pictures, text descriptions, links, etc.) to the terminal device (such as mobile phones, computers, tablets, etc.) used by the target user. After receiving the advertising content, the terminal device will display the target recommended advertisement on the display interface of the device according to the display rules set by the system, such as on the homepage of the e-commerce platform, in a specific advertising position on the product details page, etc., so that the target user can see these recommended advertisements.

[0052] The embodiments of the present invention match the advertisement feature information extracted from the advertisement according to the user interest tags and user scenarios of the users, and not only consider the user interests, but also consider the impact of different scenarios on the user needs, so that the matched advertisements can more accurately fit the user's preferences in specific scenarios, and further sort the advertisements according to the degree of matching between the user's user interest tags and the user scenarios, so that the final advertisements can more accurately fit the current scenario needs. Therefore, the advertisements can not only accurately fit the user's preferences, but also accurately fit the user's scenario needs, so that the recommended advertisements can accurately hit the user's needs, thereby improving the conversion rate of the advertisements and the user experience.

[0053] In one embodiment, the description of steps 101 to 104 is as follows:

[0054] Step 101, starting from the search behavior, mining the browsing behavior, purchase behavior and / or sharing behavior occurring at different times after the search behavior, and obtaining a behavior sequence.

[0055] Optionally, the user behavior data in the embodiment of the present invention includes search behavior, browsing behavior, purchase behavior and sharing behavior.

[0056] Therefore, the ad recommendation system will monitor the target user's activities on the platform. Once a user has performed a search, the system will begin to record the target user's subsequent browsing behavior (which product pages, information pages, etc. have been browsed), purchasing behavior (which products have been purchased), and sharing behavior (which content on the platform has been shared) at different time points. Furthermore, the ad recommendation system organizes these behaviors into a behavior sequence in chronological order. Each behavior sequence reflects a series of related activities carried out by the user based on the search topic after completing a search.

[0057] In one embodiment, user A searches for "sports shoes" on an e-commerce platform. Within half an hour, user A browses the product detail pages of brand X and brand Y sports shoes. Two days later, user A buys a pair of brand X sports shoes. A week later, user A shares the purchase page of the pair of brand X sports shoes on a social platform. The ad recommendation system captures these behaviors and organizes them into a behavior sequence: [search for "sports shoes", browse the product detail page of brand X sports shoes, browse the product detail page of brand Y sports shoes, buy brand X sports shoes, share the purchase page of brand X sports shoes].

[0058] Step 102: cluster similar behavior sequences based on the behaviors in each behavior sequence and the sequence length of each behavior sequence to obtain behavior clusters.

[0059] Furthermore, the advertising recommendation system analyzes the numerous collected behavior sequences. First, it compares the specific behaviors in each behavior sequence, for example, whether they all include browsing and purchasing behaviors of the same type of goods; second, it considers the length of the behavior sequence, because the sequence length can reflect, to a certain extent, the complexity or duration of the user's activities around a certain point of interest.

[0060] Furthermore, the advertising recommendation system classifies behavior sequences with similar characteristics into one category based on similarity to form behavior clusters, where each behavior cluster represents a potential user interest pattern, meaning that users with the cluster characteristics have certain commonalities in their interest behavior performance.

[0061] In one embodiment, the behavior sequences of multiple users after sports-related searches are collected. In addition to the above-mentioned behavior sequence of user A, there is also the behavior sequence of user B: [search for "sports backpack", browse the product details page of brand Z sports backpack, and purchase brand Z sports backpack]. The behavior sequence of user C: [search for "sports shoes", browse the product details page of brand N sports shoes, and purchase brand N sports shoes]. The system finds that the behavior sequences of user A and user C both start with the search for "sports shoes", and both browse the product details page of sports shoes and purchase sports shoes. Although the brands are different, the behavior types and sequence lengths are relatively similar. Therefore, the behavior sequences of user A and user C are clustered together to form a behavior cluster, which reflects the user's interest pattern in sports shoes from search to purchase; while the behavior sequence of user B revolves around sports backpacks, which is different from the previous two and is clustered separately, reflecting the user's interest pattern in sports backpacks from search to purchase.

[0062] Step 103 : for each behavior cluster, determine the interest pattern strength of each behavior cluster based on the time interval from the starting behavior to the last behavior in the behavior cluster.

[0063] Furthermore, for each behavior cluster, the advertising recommendation system calculates the time interval experienced by the users in the cluster from the initial behavior (usually search behavior) to the last behavior (browsing, purchasing or sharing behavior). The shorter the time interval, the faster and more coherent the user's actions on this point of interest are, reflecting the higher intensity of the user's interest pattern; conversely, the longer the time interval, the slower the user's decision-making process on this point of interest is or the interest is not persistent, and the intensity of the interest pattern is lower.

[0064] Continuing with the above embodiment, in the sports shoe-related behavior clusters of user A and user C, user A experienced a week from searching for sports shoes to sharing the purchase page; user C experienced a day from searching for sports shoes to purchasing. In contrast, the time interval from the starting behavior to the last behavior in the behavior sequence of user C is shorter, indicating that the intensity of user C's interest mode in sports shoe purchase is relatively high; although user A eventually completed the purchase and sharing, the whole process took a long time, and the intensity of his interest mode is slightly lower than that of user C.

[0065] Step 104 performs in-depth analysis based on the interest pattern strength of each behavior cluster combined with the knowledge graph to generate user interest tags for the target user.

[0066] Furthermore, the advertising recommendation system assigns weights to different interest patterns according to the strength of the interest patterns determined by the behavior clustering. Furthermore, the advertising recommendation system combines the knowledge graph, which contains a wealth of information such as commodities, brands, categories, user attributes, and the associations between them, and generates user interest tags for target users by analyzing the connection between the interest patterns reflected by the behavior clustering and the elements in the knowledge graph, such as the categories and brand characteristics of the commodities involved in the interest patterns, as described in steps 1041 to 1045.

[0067] The embodiments of the present invention start from the basic behavior of users, and after layer-by-layer analysis and refinement, provide high-quality user interest tags for advertising recommendations, which helps to more accurately recommend advertisements that meet the interests of users, improve the pertinence and effectiveness of advertising recommendations, thereby achieving the recommended advertisements accurately meeting the needs of users, and improving the conversion rate of advertisements and user experience.

[0068] In one embodiment, the description of steps 1041 to 1045 is as follows:

[0069] Step 1041, based on the mapping relationship between the behaviors in the target behavior cluster whose interest pattern intensity is greater than the intensity threshold and the entities in the knowledge graph, semantic expansion is performed to determine the extended information of the target behavior cluster in the knowledge graph.

[0070] Optionally, the advertising recommendation system first sets a strength threshold to filter out target behavior clusters with higher interest pattern strength. For these target behavior clusters, the advertising recommendation system will establish a mapping relationship between the behaviors (such as the commodities, brands, and other elements involved in the search, browsing, and purchase behaviors) and the entities in the knowledge graph. Through this mapping, the advertising recommendation system performs semantic expansion in the knowledge graph, mines other entities associated with these entities and the attributes of these entities, and thus obtains the extended information of the target behavior cluster in the knowledge graph. Therefore, the extended information includes the extended entity and the attributes of the extended entity.

[0071] Continuing with the above embodiment, the strength threshold is set to 60%. Regarding the behavioral clustering of sports shoes, the strength of user C's interest pattern in purchasing sports shoes is 70%, which is greater than the threshold and belongs to the target behavioral clustering. In the knowledge graph, the entity "Brand N Sports Shoes" is associated with entities such as "Sports Shoes", "Sports Brands", and "Fashion Trends", and "Brand N Sports Shoes" has attributes such as "Air Cushion Technology" and "Multiple Colors Available". Through the mapping relationship, the associated entities and attributes are used as extended information of the target behavioral clustering.

[0072] Step 1042, adding the extended information of the target behavior cluster to the user interest pattern corresponding to the target behavior cluster, to obtain the semantically extended interest pattern of the target behavior cluster.

[0073] Furthermore, the advertising recommendation system integrates the extended information into the user interest pattern corresponding to the original target behavior cluster, and obtains the semantically extended interest pattern of the target behavior cluster, so that the original interest pattern formed based on user behavior is enriched and refined, and transformed from a simple behavior description to an interest pattern containing more semantic information, which can reflect the user's interests more comprehensively and deeply.

[0074] In one embodiment, the original user interest pattern of user C regarding sports shoes is a series of behaviors from searching for "sports shoes" to purchasing sports shoes of brand N. After adding extended information, the semantically extended interest pattern becomes: starting from searching for "sports shoes", browsing sports shoes product detail pages of brand N, brand Y, etc., and purchasing sports shoes of brand N (sports shoes of brand N are products of sports brand N, with air cushion technology, multiple colors to choose from, and related sports brands, fashion trends, etc.).

[0075] Step 1043, constructing a semantic feature graph based on the behavioral feature description information and extended information in the interest pattern after semantic expansion.

[0076] Furthermore, the advertising recommendation system uses the behavior feature description information (such as search behavior, browsing behavior, purchase behavior, etc.) and extended information (extended entities and attributes) in the interest model after semantic expansion as nodes. According to the actual semantic relationship between the feature description information and the extended information (such as "Brand X sports shoes" and "sports brand" are affiliation relationships, and "Brand X sports shoes" and "air cushion technology" are attribute association relationships), edges are established between nodes to construct a semantic feature graph.

[0077] In one embodiment, a semantic feature graph is constructed based on the semantically expanded interest pattern of user C. The nodes include "search for sports shoes", "browse the product details page of brand X sports shoes", "buy brand X sports shoes", "brand X sports shoes", "sports brand", "air cushion technology", "fashion trend", etc. "Search for sports shoes" and "browse the product details page of brand X sports shoes" are connected by the edge of "subsequent behavior";

[0078] "Brand X sports shoes" and "sports brand" are connected through the edge of "belonging to the brand"; "Brand X sports shoes" and "air cushion technology" are connected through the edge of "having technology", and so on to construct the entire semantic feature graph.

[0079] Step 1044, taking the target nodes in the semantic feature graph whose label potential values ​​are greater than a preset threshold as interest label candidates, performing relationship modeling based on the interest label candidates, and determining the relationships between the interest label candidates.

[0080] Furthermore, the advertising recommendation system sets a label potential value for each node in the semantic feature graph, which reflects the possibility of the node being an interest label.

[0081] Optionally, the advertising recommendation system sets a preset threshold to filter out target nodes whose tag potential values ​​are greater than the threshold, and these nodes become interest tag candidates.

[0082] Furthermore, the advertising recommendation system performs relationship modeling on these interest tag candidates and analyzes the mutual relationships between them, such as which candidate tags are in parallel relationship and which are in inclusion relationship. In one embodiment, the preset threshold is 0.5. In the semantic feature graph, the label potential value of "Brand X Sports Shoes" is 0.8, the label potential value of "Sports Brand" is 0.6, and the label potential value of "Air Cushion Technology" is 0.4. Then "Brand X Sports Shoes" and "Sports Brand" become interest tag candidates. Through analysis, it is found that "Brand X Sports Shoes" and "Sports Brand" are in a belonging relationship, that is, "Brand X Sports Shoes" belongs to the category of "Sports Brand".

[0083] Step 1045 , determining user interest tags based on the relationships between interest tag candidates.

[0084] Furthermore, the advertising recommendation system integrates and refines the candidate interest tags based on the relationships between the candidate interest tags, comprehensively considers their relationships and importance in the semantic feature graph, and finally determines the user interest tags that can accurately describe the interests of the target users.

[0085] In one embodiment, based on the two interest tag candidates "Brand X sports shoes" and "Sports brand" and the relationship between them, the user interest tag for user C is determined to be "Sports brand enthusiast - Brand X sports shoes preference".

[0086] The embodiments of the present invention mine user interests in depth and breadth, making the generated user interest tags more accurate, comprehensive, and detailed, providing high-quality user interest tags for advertising recommendations, and helping to more accurately recommend advertisements that match users' interests, thereby improving the pertinence and effectiveness of advertising recommendations, thereby achieving accurate targeting of recommended advertisements to meet user needs, and improving advertising conversion rates and user experience.

[0087] In one embodiment, the description of steps 301 to 304 is as follows:

[0088] Step 301, construct a directed graph based on user interest tags and user scenarios.

[0089] Optionally, the advertising recommendation system constructs a directed graph with user interest tags and user scenarios as nodes, wherein for each user interest tag, the connection between it and different user scenarios is analyzed. If there is a connection, a directed edge is established from the user interest tag to the user scenario. This edge is accompanied by an attribute, namely the association trigger condition, which is used to clarify under what conditions the user interest tag will be associated with the corresponding user scenario. Such a graph can clearly show the potential connection between user interest tags and user scenarios and the conditions that trigger these connections. Therefore, it can be understood that the nodes in the directed graph represent user interest tags and user scenarios; for each user interest tag, a directed edge connection is established between it and the relevant user scenario, and the direction of the edge is from the user interest tag to the user scenario; the attribute of the edge is the association trigger condition, which indicates under what circumstances the user interest tag is associated with the user scenario.

[0090] In one embodiment, user B's interest tags include "outdoor sports" and "photography", and their common user scenarios include "weekend leisure time" and "holiday travel". For the interest tag "outdoor sports", in the "weekend leisure time" scenario, when user B goes to the park for running, mountain climbing, etc., the two are associated, so in the directed graph, a directed edge is established from the "outdoor sports" node to the "weekend leisure time" node, and the attribute association trigger condition of the edge is "outdoor fitness activities on weekends". For the interest tag "photography", in the "holiday travel" scenario, when user B goes out to travel and uses a camera to take pictures of scenery, people, etc., the two are associated, so a directed edge is established from the "photography" node to the "holiday travel" node, and the attribute association trigger condition of the edge is "taking pictures while traveling on vacation".

[0091] Step 302: for each user interest tag, determine the target user scenario associated with the user interest tag based on the directed graph.

[0092] Furthermore, for each user interest tag, the advertising recommendation system will filter out user scenarios that are actually associated with the user interest tag in the current situation based on the connection relationship and associated trigger conditions in the graph. These filtered user scenarios are the target user scenarios.

[0093] Furthermore, the advertising recommendation system determines whether the associated trigger conditions are met by analyzing the user's current behavior, time and other information, thereby determining the target user scenario.

[0094] In one embodiment, on a certain weekend, user B plans to go running in the park. At this time, the association trigger condition of the "outdoor sports" interest tag and the "weekend leisure time" scene is met, "there are outdoor fitness activities on weekends". Based on the directed graph and the current situation, "weekend leisure time" is determined as the target user scene associated with the "outdoor sports" interest tag. Similarly, if user B is planning a holiday trip and is going to take a camera to shoot, the association trigger condition of the "photography" interest tag and the "holiday travel" scene is met, then "holiday travel" is the target user scene associated with the "photography" interest tag.

[0095] Step 303: Match the target advertisement features in the advertisement feature information according to the compatibility between the scene characteristics of the user scene and the semantic structure of the advertisement features in the advertisement feature information.

[0096] Furthermore, the ad recommendation system will first analyze the scene characteristics of each user scene, such as the time, location, user behavior tendency, etc. At the same time, the ad feature information is parsed. Each ad feature information is composed of one or more semantic components. These semantic components form a specific semantic structure. The root node represents the entire ad feature information, and the child nodes are the semantic descriptions of each semantic component.

[0097] Furthermore, the advertising recommendation system compares the scenario characteristics of the user scenario with the semantic structure of the advertising feature information to determine the compatibility between the two, thereby screening out target advertising features that are compatible with the target user scenario from among a large amount of advertising feature information.

[0098] In one embodiment, for the target user scenario of user B, "Weekend leisure time - outdoor sports", the scenario characteristics are weekends, outdoor sports and fitness. There is an advertisement about sports drinks in the advertising feature information, and the root node in its semantic structure is "sports drink advertisement", and the sub-node semantic descriptions include semantic components such as "suitable for replenishing energy during outdoor sports" and "weekend promotions". Through comparison, the advertising recommendation system finds that the semantic structure of the advertisement is compatible with the target user scenario characteristics of user B. "Suitable for replenishing energy during outdoor sports" corresponds to the outdoor sports scene, and "weekend promotions" correspond to weekend leisure time, so the advertising feature information of this sports drink advertisement is used as the target advertising feature.

[0099] Step 304: Match the user interest tags with the target advertisement features to obtain the advertisement to be recommended.

[0100] Furthermore, the advertisement recommendation system matches the user interest tag with the target advertisement feature. If the product or service promoted by the target advertisement feature matches the user interest reflected by the user interest tag, the corresponding advertisement is determined as the advertisement to be recommended, wherein the matching can be performed by analyzing the consistency between the product category, brand positioning, target audience and other elements in the advertisement feature information and the user interest tag, as described in steps 3041 to 3044.

[0101] The embodiments of the present invention deeply associate and match user interests, user scenarios and advertising features, so that it is possible to more accurately screen out recommended advertisements that match the user's current interests and scenarios. Therefore, the advertisements can not only accurately meet the user's preference needs, but also accurately meet the user's scenario needs, thereby achieving recommended advertisements that accurately meet user needs and improving advertising conversion rates and user experience.

[0102] In one embodiment, the description of steps 3041 to 3044 is as follows:

[0103] Step 3041, based on the semantic association relationship between the user interest tag and each target advertisement feature, the user interest tag is mapped with each target advertisement feature to obtain an initial matching value between the user interest tag and each target advertisement feature.

[0104] Optionally, the ad recommendation system will conduct an in-depth analysis of the semantic connection between the user's interest tags and each target ad feature. By comparing the meaning of the interest tags, such as the field of interest and product type, with the characteristics of the products and services promoted in the ad features, the system can determine the degree of correlation between the two. Based on this degree of correlation, the system assigns an initial matching value between the user's interest tags and each target ad feature, which preliminarily reflects the degree of fit between the user's interest and the ad features.

[0105] In one embodiment, user B’s interest tag is “outdoor sports”. For the target advertising features, there is an advertising feature about sports watches, which emphasizes the functions of sports track recording and heart rate monitoring, which is very suitable for outdoor sports scenes. Through semantic analysis, the advertising recommendation system found that the “outdoor sports” interest tag and the sports watch advertising feature are closely semantically related in terms of functional adaptation to outdoor sports, so the initial matching value between “outdoor sports” and the sports watch advertising feature is 0.8 (the value range is 0-1, the higher the value, the higher the matching degree). For another advertising feature about fashionable casual shoes, although it is also in the category of footwear, it mainly emphasizes the fashionable and casual style, and the semantic association with “outdoor sports” is relatively weak, and the initial matching value is set to 0.3.

[0106] Step 3042, based on the constraints of the user scenario, adjust the initial matching values ​​between the user interest tags and each target advertisement feature to obtain the target matching values ​​between the user interest tags and each target advertisement feature.

[0107] Furthermore, the advertising recommendation system will take into account the constraints brought by the user's current user scenario. Different user scenarios will affect the specific expression and needs of user interests. For example, in a work scenario, the user's demand for electronic products may be more focused on office functions; while in a leisure travel scenario, the demand for the portability and entertainment functions of electronic products may be more prominent. Therefore, the advertising recommendation system adjusts the initial matching value based on these scenario constraints. If the advertising features are well adapted to the user scenario, the initial matching value will be appropriately increased; otherwise, it will be reduced to obtain the target matching value, which more accurately reflects the degree of matching between the user's interests and the advertising features in the current user scenario.

[0108] In one embodiment, if user B is in the "weekend leisure time - outdoor sports" scenario, for the sports watch advertising feature, since its function is very practical in the weekend outdoor sports scenario, the advertising recommendation system adjusts the initial matching value of 0.8 to the target matching value of 0.9 according to this scenario constraint. For the fashion casual shoes advertising feature, its adaptability is poor in this weekend leisure time scenario dominated by outdoor sports, so the initial matching value of 0.3 is adjusted to the target matching value of 0.1.

[0109] Step 3043, construct a matching path from user interest tags and user scenarios to advertisement feature information based on the target matching value.

[0110] Furthermore, the advertisement recommendation system takes user interest tags and user scenarios as starting points, and based on the target matching values, filters out target advertisement features whose target matching values ​​are greater than a preset matching threshold.

[0111] Furthermore, the advertising recommendation system connects these eligible target advertising features and constructs a matching path from user interest tags and user scenarios to advertising feature information, where the matching path shows the association relationship with advertising feature information that has a high degree of user adaptability under the current user interests and scenarios.

[0112] Continuing with the above embodiment, the preset matching threshold is 0.6. In the "Weekend Leisure Time - Outdoor Sports" scenario, the target matching value of the "Outdoor Sports" interest tag and the sports watch advertising feature of user B is 0.9, which is greater than the threshold. Therefore, the advertising recommendation system constructs a matching path from the "Weekend Leisure Time - Outdoor Sports" scenario and the "Outdoor Sports" interest tag to the sports watch advertising feature information. However, the target matching value of the fashion casual shoes advertising feature of 0.1 is less than the threshold and will not appear in the matching path.

[0113] Step 3044: determine the advertisement set corresponding to the matching path as the advertisement to be recommended.

[0114] Furthermore, the advertising recommendation system will match the advertising feature information corresponding to the path, find the corresponding actual advertisement, and form an advertising set. This advertising set is the advertisements to be recommended based on user interest tags, user scenarios, and matching rules. These advertisements are considered to be the advertisements that are most likely to meet user needs and interests under the current circumstances. In one embodiment, there is only one actual advertisement corresponding to the sports watch advertising feature information, so this sports watch advertisement is determined as the advertisement to be recommended for user B in the "weekend leisure time-outdoor sports" scenario. If the matching path also involves other advertising feature information that meets the conditions, such as the advertising feature information of a sports bottle also meets the conditions, then the sports bottle advertisement will also be included in the set of advertisements to be recommended.

[0115] The embodiments of the present invention not only take into account the semantic association between user interests and advertising features, but also combine user scenario constraints, thereby greatly improving the accuracy and personalization of advertising recommendations, and can provide users with advertising recommendations that are more in line with their current interests and scenario requirements, thereby achieving recommended advertisements that accurately meet user needs, thereby improving advertising conversion rates and user experience.

[0116] In one embodiment, the description of steps 401 to 404 is as follows:

[0117] Step 401 : determining the correlation between the user interest tag and the user scenario based on the number of features shared between the user interest tag and the user scenario and the total number of features of the user interest tag and the user scenario itself.

[0118] Optionally, the ad recommendation system first extracts the features of the user interest tag and the user scenario. For example, the user interest tag "outdoor sports" may have features including outdoor, sports, fitness, etc.; the user scenario "weekend park jogging" may have features including weekend, park, jogging, outdoor, etc., and calculate the number of features shared by the two.

[0119] Furthermore, the ad recommendation system combines the user interest tags and the total number of features of the user scenario itself, and determines the correlation between them through certain rules. Generally speaking, the more common features there are and the greater their proportion in the total number of features, the higher the correlation.

[0120] In one embodiment, the user interest tag "food cooking" has features such as food, cooking, and kitchen; the user scenario "weekend family dinner preparation" has features such as weekend, family, dinner, preparation, food, and cooking. The common features are food and cooking, a total of 2. The total number of features of "food cooking" itself is 3, and the total number of features of "weekend family dinner preparation" itself is 6. The advertising recommendation system calculates the correlation degree as 2 / (3+6-2)=0.4 according to the rule (set correlation degree = common feature number / (total number of user interest tag features + total number of user scenario features - common feature number)).

[0121] Step 402: Combine user interest tags to obtain different tag combinations.

[0122] Furthermore, the ad recommendation system will arrange and combine multiple interest tags owned by the user. For example, if the user has the interest tags "movie", "travel", and "food", then possible tag combinations include "movie", "travel", "food", "movie-travel", "movie-food", "travel-food", "movie-travel-food", etc. Through such a combination, the relationship between different interest tag combinations and user scenarios can be explored, and a more comprehensive user interest pattern can be mined.

[0123] Step 403: Determine the matching degree between the tag combination and the user scenario based on the appearance frequency of the user interest tag in its corresponding tag combination and the association between the user interest tag and the user scenario.

[0124] Furthermore, for each tag combination, the ad recommendation system will count the frequency of occurrence of each user interest tag in the combination. At the same time, combined with the correlation between each user interest tag and the user scenario, the higher the frequency of occurrence and the higher the correlation with the user scenario, the higher the matching degree between this tag combination and the user scenario. For example, if the frequency of "outdoor sports" in a tag combination is high, and "outdoor sports" is highly correlated with the current user scenario "weekend mountain climbing activities", then the matching degree between this tag combination and the "weekend mountain climbing activities" scenario is high.

[0125] In one embodiment, for the user scenario "weekend gym workout", the interest tag "fitness" is associated with the scenario at a degree of 0.8, "music" is associated with the scenario at a degree of 0.2, and "reading" is associated with the scenario at a degree of 0.1. In the tag combination "fitness-music", the frequency of "fitness" is 0.5, and the frequency of "music" is 0.5. According to the rule (matching degree = frequency of interest tag 1 * correlation between interest tag 1 and scenario + frequency of interest tag 2 * correlation between interest tag 2 and scenario), the degree of matching between this tag combination and the "weekend gym workout" scenario is 0.5*0.8+0.5*0.2=0.5. In the tag combination "fitness", the frequency of "fitness" is 1, and its degree of matching with the scenario is 1*0.8=0.8.

[0126] Step 404: sort the advertisements to be recommended based on the matching degree between the tag combination and the user scenario to obtain target recommended advertisements.

[0127] Furthermore, the advertisement recommendation system sorts the advertisements to be recommended according to the matching degree between the tag combination and the user scenario to obtain the target recommended advertisements, as specifically described in steps 4041 to 4043 .

[0128] The embodiments of the present invention can analyze user interests and scenarios in more detail, and prioritize advertisements that best match the user's current interests and scenarios to the user, thereby ensuring that the recommended advertisements accurately meet the user's needs, thereby improving the conversion rate of advertisements and user experience.

[0129] In one embodiment, the description of steps 4041 to 4043 is as follows:

[0130] Step 4041, based on the number of keywords in each advertisement to be recommended that match the keywords in the tag combination and the total number of keywords in each advertisement to be recommended, determine the degree of association between the tag combination and each advertisement to be recommended.

[0131] Optionally, the advertisement recommendation system extracts keywords from each advertisement to be recommended to form an advertisement keyword set. At the same time, each tag combination is also decomposed into keywords.

[0132] Furthermore, the advertisement recommendation system compares the keyword set of each advertisement to be recommended with the keywords of the tag combination one by one, and counts the number of keywords in the advertisement to be recommended that match the keywords of the tag combination. Furthermore, the advertisement recommendation system determines the degree of association between the tag combination and the advertisement to be recommended based on the proportion of the number of matches in the total number of keywords in the advertisement to be recommended. The higher the proportion, the higher the degree of association.

[0133] In one embodiment, there is a tag combination of "outdoor sports-fitness", and the corresponding keywords are "outdoor", "sports", and "fitness". The keywords of the advertisement C to be recommended are "outdoor running shoes", "sports and fitness equipment", and "indoor fitness courses", a total of 3 keywords. Among them, the keywords that match the tag combination are "outdoor", "sports", and "fitness", a total of 3 keywords. The degree of association between the tag combination and the advertisement C to be recommended is 3÷3=1. The keywords of the advertisement D to be recommended are "fashion watches" and "business office supplies", a total of 2 keywords, and the number of matches with the tag combination keywords is 0, so the degree of association is 0÷2=0.

[0134] Step 4042: Determine the comprehensive score of each advertisement to be recommended based on the matching degree between the tag combination and the user scenario and the association degree between the tag combination and each advertisement to be recommended.

[0135] Furthermore, the advertising recommendation system combines the degree of match between the tag combination and the user scenario, and the degree of association between the tag combination and each to-be-recommended advertisement, and determines a comprehensive score for each to-be-recommended advertisement through certain rules (non-weighted summation rules). For example, if the tag combination has a high degree of match with the user scenario, and a high degree of association with a certain to-be-recommended advertisement, then the comprehensive score of the to-be-recommended advertisement will be higher. In one embodiment, for the user scenario "weekend park fitness activities", the tag combination "outdoor sports-fitness" has a degree of match with the scenario of 0.7. The degree of association between the to-be-recommended advertisement C and the "outdoor sports-fitness" tag combination is 1. According to the set rule (comprehensive score = the degree of match between the tag combination and the scenario + the degree of association between the tag combination and the advertisement), the comprehensive score of the to-be-recommended advertisement C is 0.7+1=1.7. The degree of association between the to-be-recommended advertisement D and the "outdoor sports-fitness" tag combination is 0, and its comprehensive score is 0.7+0=0.7.

[0136] Step 4043 sorts each advertisement to be recommended based on its comprehensive score to obtain a target recommended advertisement.

[0137] Furthermore, the advertising recommendation system sorts all the advertisements to be recommended from high to low according to the comprehensive score determined for each advertisement to be recommended. The advertisements to be recommended that are ranked at the front have higher comprehensive scores, which means that they perform better in matching with the user scenario and associating with relevant tag combinations. These advertisements are determined as target recommended advertisements and will be displayed to users preferentially. Continuing with the above embodiment, the comprehensive score of advertisement C to be recommended is 1.7, the comprehensive score of advertisement D to be recommended is 0.7, and the comprehensive score of advertisement E to be recommended is 1.2. The advertising recommendation system sorts the advertisements to be recommended from high to low according to the comprehensive scores as follows: advertisement C to be recommended, advertisement E to be recommended, and advertisement D to be recommended. Then advertisement C to be recommended and advertisement E to be recommended become target recommended advertisements, and are preferentially displayed to users in the "weekend park fitness activities" scenario.

[0138] The embodiments of the present invention not only take into account the relationship between user interests and scenarios, but also delve into the association between advertising content and interest tag combinations, greatly improving the accuracy and pertinence of advertising recommendations, and can provide users with advertising recommendations that are highly consistent with current interests and scenarios, so that recommended advertisements accurately meet user needs, thereby improving advertising conversion rates and user experience.

[0139] See also Figure 3 , Figure 3 FIG. 1 is an embodiment diagram of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0140] Responding to the platform login instruction of the target user, obtaining the user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user;

[0141] Perform scene recognition based on the target user's current usage scene information to obtain the target user's current user scene;

[0142] Match the user interest tags and user scenarios with the ad feature information extracted from the ad to obtain the ad to be recommended for the target user;

[0143] Sort the recommended ads based on the matching degree between the user interest tags and the user scenarios to obtain the target recommended ads;

[0144] The target recommended advertisement is displayed on the display interface of the target user's terminal device.

[0145] See also Figure 4 , Figure 4 Detailed description of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0146] Responding to the platform login instruction of the target user, obtaining the user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user;

[0147] Perform scene recognition based on the target user's current usage scene information to obtain the target user's current user scene;

[0148] Match the user interest tags and user scenarios with the ad feature information extracted from the ad to obtain the ad to be recommended for the target user;

[0149] Sort the recommended ads based on the matching degree between the user interest tags and the user scenarios to obtain the target recommended ads;

[0150] The target recommended advertisement is displayed on the display interface of the target user's terminal device.

[0151] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the personalized advertising intelligent recommendation control method based on big data provided by the above methods. The personalized advertising intelligent recommendation control method based on big data includes:

[0152] Responding to the platform login instruction of the target user, obtaining the user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user;

[0153] Perform scene recognition based on the target user's current usage scene information to obtain the target user's current user scene;

[0154] Match the user interest tags and user scenarios with the ad feature information extracted from the ad to obtain the ad to be recommended for the target user;

[0155] Sort the recommended ads based on the matching degree between the user interest tags and the user scenarios to obtain the target recommended ads;

[0156] The target recommended advertisement is displayed on the display interface of the target user's terminal device.

[0157] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0158] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A personalized advertising intelligent recommendation system based on big data, characterized in that: It includes an advertising intelligent recommendation platform, an acquisition module, a scene recognition module, an advertising matching module, an advertising sorting module, and an advertising recommendation module; the advertising intelligent recommendation platform is connected to the acquisition module, the scene recognition module, the advertising matching module, the advertising sorting module, and the advertising recommendation module respectively to manage the data of each module; An acquisition module, used to respond to a platform login instruction of a target user and acquire user interest tags of the target user; User interest tags are generated based on the user behavior data analysis of the target user; A scene recognition module, used to perform scene recognition based on the current usage scene information of the target user to obtain the user scene in which the target user is currently located; An advertisement matching module is used to match user interest tags and user scenarios with advertisement feature information extracted from advertisements to obtain advertisements to be recommended to the target user; An advertisement sorting module is used to sort the advertisements to be recommended based on the matching degree between the user interest tags and the user scenarios to obtain target recommended advertisements; The advertisement recommendation module is used to display the target recommended advertisement on the display interface of the terminal device of the target user.

2. A method for controlling personalized advertising intelligent recommendation based on big data, implemented based on the personalized advertising intelligent recommendation system based on big data as claimed in claim 1, characterized in that: The personalized advertising intelligent recommendation control method based on big data includes: Responding to a platform login instruction of a target user, obtaining a user interest tag of the target user; the user interest tag is generated based on the user behavior data analysis of the target user; Performing scene recognition based on the current usage scene information of the target user to obtain the current user scene of the target user; Matching the user interest tags and user scenarios with the advertisement feature information extracted from the advertisement to obtain the advertisement to be recommended for the target user; Sort the recommended ads based on the matching degree between the user interest tags and the user scenarios to obtain the target recommended ads; The target recommended advertisement is displayed on the display interface of the terminal device of the target user.

3. The method for intelligent recommendation control of personalized advertising based on big data according to claim 2, characterized in that: The step of matching the user interest tag and the user scenario with the advertisement feature information extracted from the advertisement to obtain the advertisement to be recommended to the target user includes: A directed graph is constructed based on user interest tags and user scenarios; the nodes in the directed graph represent user interest tags and user scenarios; for each user interest tag, a directed edge connection is established between it and the relevant user scenario, and the direction of the edge is from the user interest tag to the user scenario; the attribute of the edge is the association trigger condition, which indicates under what circumstances the user interest tag is associated with the user scenario; For each user interest tag, determining a target user scenario associated with the user interest tag based on the directed graph; According to the adaptability between the scene characteristics of the user scene and the semantic structure of the advertisement features in the advertisement feature information, the target advertisement features are matched in the advertisement feature information; each advertisement feature information includes one or more semantic components, the semantic components form a semantic structure, the root node of the semantic structure represents the advertisement feature information, and the child nodes represent the semantic description of the semantic components; The advertisement to be recommended is obtained by matching the user interest tag with the target advertisement feature.

4. The method for intelligent recommendation control of personalized advertising based on big data according to claim 3 is characterized in that: The step of matching the user interest tag with the target advertisement feature to obtain the advertisement to be recommended includes: Based on the semantic association between the user interest tag and each target advertisement feature, the user interest tag is mapped with each target advertisement feature to obtain an initial matching value between the user interest tag and each target advertisement feature; Adjust the initial matching value between the user interest tag and each target advertisement feature based on the constraint conditions of the user scenario to obtain the target matching value between the user interest tag and each target advertisement feature; Constructing a matching path from the user interest tag and the user scenario to the advertisement feature information based on the target matching value; the target matching value corresponding to the target advertisement feature in the matching path is greater than a preset matching threshold; The advertisement set corresponding to the matching path is determined as the advertisement to be recommended.

5. The method for intelligent recommendation control of personalized advertising based on big data according to claim 2, characterized in that: The method of sorting the advertisements to be recommended based on the matching degree between the user interest tags and the user scenarios to obtain target recommended advertisements includes: Determine the correlation between the user interest tag and the user scenario based on the number of common features between the user interest tag and the user scenario and the total number of features of the user interest tag and the user scenario itself; Combine user interest tags to obtain different tag combinations; Determine the matching degree between the tag combination and the user scenario based on the frequency of occurrence of the user interest tag in its corresponding tag combination and the correlation between the user interest tag and the user scenario; The advertisements to be recommended are sorted based on the matching degree between the tag combination and the user scenario to obtain the target recommended advertisement.

6. The method for controlling personalized advertising intelligent recommendation based on big data according to claim 5, characterized in that: The step of sorting the advertisements to be recommended based on the matching degree between the tag combination and the user scenario to obtain the target recommended advertisement includes: Determine the degree of association between the tag combination and each advertisement to be recommended based on the number of keywords in each advertisement to be recommended that match the keywords in the tag combination and the total number of keywords in each advertisement to be recommended; Determine the comprehensive score of each advertisement to be recommended based on the matching degree between the tag combination and the user scenario and the correlation degree between the tag combination and each advertisement to be recommended; Each advertisement to be recommended is sorted based on the comprehensive score of each advertisement to be recommended to obtain the target recommended advertisement.

7. The method for controlling personalized advertising intelligent recommendation based on big data according to any one of claims 2 to 6, characterized in that: The user behavior data includes search behavior, browsing behavior, purchase behavior and sharing behavior; The specific steps of performing in-depth analysis based on the user behavior data of the target user and generating the user interest tag of the target user include: Taking the search behavior as the starting point, mining the browsing behavior, purchase behavior and / or sharing behavior occurring at different times after the search behavior to obtain a behavior sequence; Based on the behaviors in each behavior sequence and the sequence length of each behavior sequence, similar behavior sequences are clustered to obtain behavior clusters; each behavior cluster represents a potential user interest pattern; For each behavior cluster, the intensity of the interest pattern of each behavior cluster is determined based on the time interval from the starting behavior to the last behavior in the behavior cluster; Based on the interest pattern strength of each behavior cluster combined with the knowledge graph, an in-depth analysis is performed to generate user interest tags for the target user.

8. The method for controlling personalized advertising intelligent recommendation based on big data according to claim 7, characterized in that: The interest pattern strength based on each behavior cluster is combined with the knowledge graph to perform in-depth analysis to generate the user interest tag of the target user, including: Based on the mapping relationship between the behaviors in the target behavior cluster whose interest pattern intensity is greater than the intensity threshold and the entities in the knowledge graph, semantic expansion is performed to determine the extended information of the target behavior cluster in the knowledge graph; the extended information includes the extended entity and the attributes of the extended entity; Adding the extended information of the target behavior cluster to the user interest pattern corresponding to the target behavior cluster to obtain the semantically extended interest pattern of the target behavior cluster; Based on the behavior feature description information and the extended information in the semantically expanded interest pattern, a semantic feature graph is constructed; the nodes in the semantic feature graph are the behavior feature description information and the extended information, and the edges between the nodes are the semantic relationships between the behavior feature description information and the extended information; The target nodes in the semantic feature graph whose label potential values ​​are greater than a preset threshold are taken as interest label candidates, and relationship modeling is performed based on the interest label candidates to determine the mutual relationships between the interest label candidates; The user interest tags of the target user are determined based on the mutual relationships between the interest tag candidates.

9. An electronic device, comprising: A memory and a processor, characterized in that a computer software program is stored in the memory, and when the processor reads and executes the computer software program, the personalized advertising intelligent recommendation control method based on big data as described in any one of claims 2 to 8 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, which, when executed by a processor, implements the big data-based personalized advertising intelligent recommendation control method as described in any one of claims 2 to 8.

Images