CN-122022914-A - DNN-based information flow advertisement planning and delivering method

Abstract

The invention discloses a method for intelligently planning information flow advertisements by using a DNN (deep neural network) model and accurately putting the information flow advertisements on different platforms. Belongs to the technical field of artificial intelligence and advertisement planning science and technology. The system comprises a DNN (deep neural network) model, an advertisement data end, a media platform, a user terminal, media software preinstalled in the user terminal and an advertisement plug-in. Firstly, the deep neural network is utilized to collect user portraits, product characteristics and advertisement characteristics of different platforms for reinforcement learning. And automatically planning out advertisements of the same product aiming at different durations and contents of different platforms. And after negotiating with the consumer, uploading the finally determined advertisement plug-in to the advertisement terminal. When a mobile phone user opens media software, according to corresponding programming, advertisement plug-ins designed for different platforms are selected for accurate delivery. For the commercial trend in the advertising industry, accurate delivery relies on a targeting engine. The method utilizes deep learning to enhance the understanding capability targets of advertisements and users, adopts the hard orientation of regions, sexes and the like, and also adopts the soft orientation method of similar crowd, user relationship and the like to find the favorite advertisements of the users. The method has the advantages of short generation time, high generation quality, low generation cost, less hardware resource consumption, less training data quantity and the like.

Inventors

• AN RAN
• DONG HUAJUN

Assignees

• 大连沣毅电力科技有限公司

Dates

Publication Date

20260512

Application Date

20251225

Claims (10)

1. 1. The method is characterized by comprising a DNN model data end, an advertisement data end, a media platform, a user terminal, media software preinstalled in the user terminal and an advertisement plug-in, and the method realizes intelligent planning and accurate delivery of advertisements through the DNN model, and comprises the following specific steps: (1) Inputting various advertising speech techniques and advertising background music into a DNN model for training, and completing training of the model and saving parameters of the model when the total loss function converges; (2) Inputting the usage groups of the product, namely user portraits, advertisement features expected by current product consumers and product features to be planned; (3) The countermeasure product analyzes data according to the model in the field and provides an advertising strategy; (4) Generating planned product advertisement languages and advertisement pictures aiming at different platforms, and determining a final advertisement plug-in after being selected by consumers and uploading the final advertisement plug-in to an advertisement terminal; (5) When the user terminal opens the media software, the advertisement plug-in is started and runs in the background at the same time, and the advertisement plug-in analyzes and judges the advertisement to be put according to the preset standard according to the type of the media platform to which the current media software belongs and the putting record in the advertisement plug-in; (6) When the user clicks to play the video, the advertisement plug-in unit puts the corresponding advertisement stored in the user terminal into the current media software and inserts the corresponding advertisement into the video, and simultaneously makes a put record.
2. 2. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the user portraits comprise consumer gender, age, occupation, territory, personal interests and uninteresting content.
3. 3. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the advertisement characteristics expected by the consumer include preference for picture content, style, and text.
4. 4. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the planning product features comprise product name and one or more of price, efficacy, composition, picture.
5. 5. The method for planning and delivering a DNN-based information stream advertisement according to claim 1, wherein the step (3) of analyzing data of the domain to which the countermeasure product belongs comprises counting the price of the product in the domain to which the countermeasure product belongs and generating a price profile, and generating a customer portrait and a brand image profile, and wherein the step of providing the advertisement policy comprises providing advertisement policies in a plurality of competing policy directions.
6. 6. The method for planning and delivering the DNN-based information flow advertisement according to claim 1, wherein the advertisement generated in the step (4) comprises a short advertisement and a long advertisement, the short advertisement time t1 is 5-15s, the long advertisement time t2 is 30-60s, the advertisement data end divides the advertisement into the short advertisement and the long advertisement according to the time length of the advertisement, and the short advertisement and the long advertisement which belong to the same brand or commodity after division are respectively assigned with matched ID numbers for correlation.
7. 7. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the media platforms comprise a short video media platform and a long video media platform, the short video media platform delivering short advertisements and the long video media platform delivering long advertisements.
8. 8. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the advertisement delivery time is selected to be delivered at a time when the consumer has a consumption habit after evaluating the click rate, the completion rate and the consumption habit of the advertisement by the consumer of each platform through the deep neural network.
9. 9. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the advertisement delivery request sent by the advertisement plug-in to the advertisement data terminal includes a device identification code of the user terminal, a user sex, a user age, a region, a personal interest, and a non-interesting content.
10. 10. The DNN-based information flow advertisement planning and delivery method according to claim 1, wherein the user terminal is provided with an editing unit for editing user personal interests, uninteresting contents, the editing unit being associated with an advertisement plug-in for updating user information in time.

Description

DNN-based information flow advertisement planning and delivering method Technical Field The invention relates to a method for generating and putting information flow advertisements, in particular to a method for generating and putting advertisements based on DNN (deep neural network) collected data. Background The advertisement plan is a design activity or process for expressing advertisement objects and intentions by combining the use characteristics of advertisement media on a computer through related design software to express advertisement elements such as images, characters, colors, layouts, graphics and the like. The advertisement plan is a combination arrangement of five elements of advertisement theme, creative, language words, image and setting off, and the final purpose is to attract public attention through the advertisement so as to achieve the propaganda effect. At present, most of enterprises' advertisement planning is completed by customs planning, creative or design departments, which can put in a great deal of manpower and financial resources to carry out advertisement planning work, and has high manpower cost, long decision flow and unpredictable finished products. Meanwhile, in recent years, video media has been rapidly developed, and various types of video media platforms have appeared. The same user may download multiple media software simultaneously on the same terminal to view video on different video media platforms through different media software, for example, short video is brushed through short video media platforms such as tremble during the time of day, and a drama is traced through video media platforms such as the aide art at night at the time of the sky. Different advertisement types need to be planned for different platforms, so the required quantity and the required quality are required, which causes many advertisement planners to be unable to complete tasks required by clients at specified time. Because the work is busy, more and more users consume more time on the short video, and the short video is usually only a few minutes, even tens of seconds, and the time length of the existing advertisement put on the long video is too long, if the existing advertisement put on the long video is put on the short video, the user dislike is easily caused, so that most of the current short video has no advertisement or only GIF (graphic information) advertisements with a few seconds, and the advertising effect is poor. Therefore, in order to achieve the maximum advertising effect, the advertiser should make advertisements with different durations for the same brand or commodity according to different media, so as to put the advertisements on different media. However, the number of advertisements to be put by the existing advertisers is large, the advertisements are related to a plurality of the same brands or commodities, the putting time and the probability between different advertisements of the same brands or commodities put on different media platforms are mutually independent, the linkage effect is not achieved, the audience is difficult to be impressed, and the advertising effect is seriously reduced. In addition, the traditional advertisement putting mode is that when a user opens a video, media software sends an advertisement request to an advertisement data end, the advertisement data end sends and inserts corresponding advertisements into the video after analyzing user information according to a pre-standard, the operation speed of the advertisement data end can be influenced when a large amount of data is analyzed, the speed of feeding back the advertisements is reduced, in addition, when the advertisements are inserted, the user terminal needs to download and play the advertisements through a network at the same time, the video and the advertisements need to be loaded through the network at the same time, the video is downloaded when the advertisements are played, the loading speed of the video and the advertisements can be greatly delayed, and the occupied memory is more, so that the normal operation of terminal equipment is influenced. Disclosure of Invention The invention aims to provide a method for automatically generating information flow advertisements based on DNN and accurately putting the information flow advertisements on different platforms. First, small advertisers lack specialized creative designers. The creative material is high in manufacturing difficulty, high in threshold, time-consuming and labor-consuming, and high in requirements on writing of advertisement texts and selection of pictures. Second, advertisers have a limited budget. How to build accounts and optimize advertisement delivery according to different media and people is still another problem facing advertisers under limited budget. Facing both problems, what are conventional solutions and the problems they have? 1. Creative production is achieved by self production of clients or application