Search

CN-121998717-A - Advertisement intelligent recommendation system based on artificial intelligence

CN121998717ACN 121998717 ACN121998717 ACN 121998717ACN-121998717-A

Abstract

The invention relates to the technical field of advertisement delivery, in particular to an intelligent advertisement recommendation system based on artificial intelligence, the system comprises a user interaction monitoring module, an interaction behavior analysis module, a pushing direction adjustment module, a track data acquisition module and a behavior consistency judging module. According to the method, the interactive behaviors between the user and the advertisement are collected in real time, the click type and the intention strength are identified, and the advertisement pushing sequence is dynamically adjusted under the condition of no click; the method comprises the steps of combining a short-term motion track after user response with a historical path model, constructing a path mutation index, identifying whether behavior deflection occurs towards an advertisement target area, when the user deviates from the advertisement target path, calling historical track data to extract periodic characteristics, evaluating long-term interest preference, dynamically correcting advertisement content pushing, and improving matching precision and response conversion effect of advertisement recommendation through a multidimensional path change parameter and interest weight fusion strategy, wherein the method is suitable for personalized advertisement putting scenes of mobile terminals.

Inventors

  • GAO JIANJUN

Assignees

  • 北京万汇互联科技有限公司

Dates

Publication Date
20260508
Application Date
20260313

Claims (10)

  1. 1. An artificial intelligence based advertising intelligent recommendation system, comprising: the advertisement delivery module is used for receiving a plurality of advertisements to be delivered, which are input by the terminal, forming a push queue according to an initial recommendation sequence, and selecting any advertisement to be delivered to a user to deliver the advertisement to form a delivered advertisement; the user interaction monitoring module is connected with the advertisement putting module and is used for determining whether interaction behaviors exist between a user and the advertisement putting module according to the collected interaction data; The interactive behavior analysis module is connected with the user interactive monitoring module and is used for analyzing the category of the interactive behavior when the interactive behavior exists to obtain a first interactive behavior category and a third interactive behavior category; The pushing direction adjusting module is connected with the interaction behavior analyzing module and is used for selecting the next advertisement to be put in the pushing queue for putting when the first interaction behavior category is obtained; The track data acquisition module is connected with the interaction behavior analysis module and is used for acquiring the current geographic position and the actual motion track of the user when the third interaction behavior category is obtained so as to determine whether the moving direction of the user deviates from the advertisement target area or not and analyzing the recommendation effect of the put advertisement; The behavior consistency judging module is connected with the track data acquisition module and is used for acquiring historical track data of the user when the user temporarily deviates from the advertisement target area so as to evaluate the long-term interest of the user according to the historical track data and correct the advertisement direction.
  2. 2. The intelligent advertisement recommendation system based on artificial intelligence according to claim 1, wherein the user interaction monitoring module comprises an interaction behavior acquisition unit, a false touch elimination unit and an advertisement area interaction judgment unit; The interactive behavior acquisition unit is used for acquiring interactive data between a user and a display interface area for advertising in real time, wherein the interactive data comprises touch events, sliding tracks and stay time; the false touch rejection unit is used for rejecting false touch behaviors according to the set touch sensitivity threshold; The advertisement area interaction judging unit is used for analyzing whether the touch event of the user is in the display interface area for putting advertisements based on the interaction data so as to judge whether the touch event is effective interaction.
  3. 3. The artificial intelligence based advertisement intelligent recommendation system according to claim 1, wherein the interactive behavior analysis module comprises an acceptance behavior analysis unit and a click behavior pattern determination unit; The acceptance behavior analysis unit is used for judging whether the user executes advertisement clicking operation when effective interaction exists between the user and the advertisement, and obtaining a first interaction behavior result when the advertisement clicking operation is not executed; The click behavior mode judging unit is used for judging the click intention intensity of the user based on the click delay time parameter after the user executes the advertisement click operation, and obtaining the corresponding interaction behavior category based on the category to which the click intention intensity belongs.
  4. 4. The artificial intelligence based advertisement intelligent recommendation system according to claim 3, wherein the click behavior pattern determination unit comprises a click delay time calculation subunit and a click intention strength evaluation subunit; the click delay time calculating subunit is used for obtaining a delay time parameter from the first time of advertisement contact to the click of advertisement; The click intention strength evaluation subunit is configured to determine the click intention strength of the user based on a comparison result of the delay time parameter and a preset standard delay time threshold.
  5. 5. The artificial intelligence based advertisement intelligent recommendation system according to claim 1, wherein the trajectory data acquisition module comprises an interactive response positioning unit, a path mutation detection unit, a behavior deflection intention recognition unit, and a trajectory effectiveness screening unit; The interactive response positioning unit is used for acquiring geographic position information of the user at preset time window intervals when a third interactive behavior of the user is obtained, and constructing a short-term motion track; The path mutation detection unit is used for calculating a path mutation index based on a historical conventional path model and a short-term motion trail of a user; The behavior deflection intention recognition unit is used for recognizing whether a user deflects a behavior path towards the advertisement target area or not according to the geographic position of the advertisement target area when the path mutation index exceeds a deflection threshold value, and judging the advertisement recommendation effect according to the behavior path deflection intention recognition unit; The track effectiveness screening unit is used for eliminating fuzzy data caused by sensor errors and path jumps related to non-put advertisements in the short-term track.
  6. 6. The artificial intelligence based advertisement intelligent recommendation system according to claim 5, wherein the path mutation detection unit comprises a trajectory comparison subunit, an offset parameter extraction subunit, and a mutation index calculation subunit; The track comparison subunit is used for constructing a track comparison structure between a short-term motion track after the user puts advertisement response and a historical conventional path model; the offset parameter extraction subunit is used for extracting parameters representing path behavior change from the track comparison structure, wherein the parameters comprise a direction jump rate, a displacement amplitude difference value and a stay point dense-scattered ratio; And the mutation index calculating subunit is used for weighting and fusing the parameters according to preset weights to generate a path mutation index.
  7. 7. The artificial intelligence based advertising intelligent recommendation system according to claim 5, wherein the behavior deflection intent recognition unit comprises a deflection trigger determination subunit and a geographic matching analysis subunit; The deflection triggering judging subunit is used for triggering geographic matching analysis when the path mutation index exceeds a preset deflection threshold value; The geographic matching analysis subunit is used for identifying whether the user track deflects towards the advertising target area or not based on the spatial position relation between the user mutation path segment and the advertising target area.
  8. 8. The intelligent advertisement recommendation system based on artificial intelligence according to claim 1, wherein the behavior consistency discriminating module comprises a history track retrieving unit, a periodic feature extracting unit and a content correction control unit; The periodic characteristic extraction unit is used for carrying out periodic behavior analysis on the historical track data, extracting a long-term path and calculating a path deviation parameter between a current path and the long-term path; the content correction control unit is used for carrying out content correction of the advertisement recommendation direction based on the historical behavior data characteristics of the user under the condition that the path deviation degree parameter is detected to exceed a preset threshold value.
  9. 9. The artificial intelligence based advertisement intelligent recommendation system according to claim 8, wherein the periodic feature extraction unit comprises a high frequency path clustering subunit, a behavior deviation quantum unit, and an analysis decision subunit; the high-frequency path clustering subunit is used for aggregating the historical track data of the user through a space-time clustering algorithm to generate a high-frequency path set with a time period label; The behavior deviation quantum unit is used for calculating a track coincidence degree parameter of a current path and a high-frequency path of a corresponding period based on the French distance and generating a period loyalty parameter based on historical frequency statistics; The analysis and judgment subunit is configured to output a path deviation parameter overrun signal to the content correction control subunit when the track overlap ratio parameter is lower than an overlap threshold or the period loyalty parameter is lower than a preset standard loyalty threshold.
  10. 10. The artificial intelligence based advertising intelligent recommendation system according to claim 8, wherein the content modification control unit comprises a geographic weight modification subunit and a priority reconstruction subunit; the geographic weight correction subunit is used for dynamically calculating and correcting geographic orientation weight coefficients based on geographic orientation information preset by advertisement delivery and the real-time geographic position of a user; The priority reconstruction subunit is used for identifying the advertisement in the push queue meeting the preset long-term interest judgment condition, calculating the similarity between the user interest vector and the advertisement content, calculating and obtaining the total weight of the advertisement, and carrying out priority reconstruction sequencing on the advertisement in the push queue.

Description

Advertisement intelligent recommendation system based on artificial intelligence Technical Field The invention relates to the technical field of advertisement delivery, in particular to an intelligent advertisement recommendation system based on artificial intelligence. Background With the rapid popularization of the mobile Internet, advertisements based on geographic positions are gradually raised, from early single-dependent SMS short messages and coarse-grained GPS pushing to the realization of fine positioning by utilizing Wi-Fi, bluetooth beacons (such as iBeacon) and geofences (Geofencing) technologies, advertisement recommendation gradually evolves to real-time perception and dynamic response, and 2009 Fourd provides APIs for platforms such as Fourdeure and the like, so that the geographic data integration capability of a first party and a third party is obviously enhanced, and a data basis is provided for accurate advertisements. In recent years, advertisement systems further introduce artificial intelligence and machine learning models, shift from static tag matching to dynamic behavior modeling, predict user intent usually by means of collaborative filtering, deep learning recommendation models (such as DLRM), but also face a number of challenges such as cold start, lack of historical behavior of new users, resulting in reduced recommendation effect, increase of recommendation system data sparsity and algorithm complexity, influence on real-time response and model expandability, and recommendation preference is too concentrated on hot items to ignore long-term interests and cold content. Although the prior art can make recommendations according to the clicking, purchasing history and other behaviors of users, the prior art mainly focuses on explicit behaviors, is difficult to sense and read for path deviation, hesitation or exploratory behaviors which are not clicked by users but are shown in short-term tracks, when users do not directly click advertisements, the travelling paths of the advertisements are deviated or stopped, the 'weak feedback' is usually ignored, and a dynamic feedback mechanism based on path change is absent in a system, so that advertisement recommendation still cannot accurately capture short-term intention fluctuation and the opportunity of adjusting the throwing direction is missed. CN116664204a discloses an artificial intelligent advertisement putting marketing recommendation system based on big data, which comprises a client management platform and an administrator auditing end, wherein the output end of the client management platform is electrically connected with a processor. The artificial intelligent advertisement putting marketing recommendation system based on big data uploads advertisements through an advertisement management platform and can automatically classify the advertisements according to labels of providers, so that the advertisements are transmitted to a data storage center, an administrator auditing end extracts the uploaded advertisements from the data storage center for auditing, illegal advertisements are deleted, delivery of the illegal advertisements is prevented, information of a delivering user is collected and analyzed through a client management platform, advertisements conforming to the user are processed through a processor, and the corresponding advertisements are pushed through an advertisement pushing module. Meanwhile, the prior art belongs to a classification matching engine based on static rule driving, and an advertisement pushing strategy of the classification matching engine mainly depends on a fixed mapping relation of a user tag corresponding to the advertisement tag, lacks intelligent dynamic regulation and control capability, and is difficult to accurately respond to complex behavior intention changes of a user in an advertisement contact process, such as click hesitation, path orientation deflection and other key behavior features due to lack of a deep analysis mechanism of user interaction behavior, lack of real-time perception capability on user track changes and lack of construction of a dynamic weight correction model based on feedback parameters. In addition, the system lacks robustness when processing external interference factors such as false touch behaviors and track noise, and cannot realize accurate judgment and self-adaptive optimization correction of advertisement content based on behavior consistency, so that the recommendation effect is limited to a basic audience screening level, and intelligent recommendation application requirements of high precision and high response requirements are difficult to meet. Disclosure of Invention Therefore, the invention provides an artificial intelligence based advertisement intelligent recommendation system, which is used for solving the problems that the prior art lacks in acquiring the click behavior of a user advertisement in real time and judging the short-term advertisement