Research On Advertising Recommendation Algorithm And Application Based On "Face Portrait"

In recent years,artificial intelligence has been developed rapidly under the impetus of big data and depth learning,face recognition technology is one of the most popular research fields.In the lab environment,the accuracy of face recognition technology has exceeded 99.5%.On this basis,the parties are also seeking applications for face recognition,and due to the lack of targeted personalized recommendations,advertising machines have affected the effectiveness of advertising.This paper combines face recognition technology with offline advertising player playback scenarios,focusing on the advertising recommendation algorithm based on "face portrait".This is an extension of the application area of face recognition technology,and has great significance for the intelligent development of offline advertising recommendation system.In the scene of advertising machines with fixed locations,such as advertising machines next to elevators and bus stations,face recognition technology can be used to create user portrait for users viewing ads,such as age,gender,etc.The same characteristic user group has the same preference for advertisements.For example,women may be more interested in cosmetics advertisements.According to this situation,this paper proposes an advertisement recommendation probability model based on"face portraits".It identifies facial features of users through face recognition technology and recommends advertisements that may be of interest to users.Firstly,A semantic similarity recommendation model for "face portraits" and advertisements is proposed.According to the correlation between face attributes and advertising features,cold start recommendation is performed,and the user-interested advertising features are learned through the content-based recommendation algorithm,and the recommendation of the model is continuously improved.Secondly,user feedback recommendation model based on user viewing duration is proposed.The longer the user watches the advertisement,the higher the possibility of interest in the advertisement,and the user's historical viewing record is analyzed through the collaborative filtering algorithm to perform advertisement recommendation.Thirdly,the time density recommendation model is proposed by analyzing the history of advertisement playback big data.The DBSCAN algorithm is used to analyze the playing history of the advertisement and cluster the advertisement playback time to predict the time period during which the current advertisement is most likely to be viewed.Finally based on the different application scenarios of the above three models,this paper fuses the models by probability and proposes a probabilistic recommendation model based on "face portraits".In experiments,the average daily viewing rate of the recommended advertising player system was 1.59%higher than that of the non-recommended advertising player system,and the average daily viewing time ratio increased by 0.33%,the effectiveness of the recommendation algorithm based on "face portrait" is tested through practical application.