Research And Implementation Of Customized Audience System Based On Social Relations

In recent years,with the development of the Internet,the emergence of mobile online sociality provides a large number of opportunities for merchants and third-industry service providers,and also brings negative effects such as “information overload”.Precision marketing has always been the most important business part of service providers and operators.Faced with a huge number of users and hundreds of millions of user data,traditional tag-type recommendation systems are difficult to find potential target users quickly and accurately.With the rise of machine learning,the concept of “audience discovery” has broadly appeared in the field of recommendation.The meaning of “audience discovery” is that when the platform has a certain size of users,the platform recommendation system can automatically discover potential audience users for the merchants in need,even if the user has not had actual contact with the merchant before.The core idea of audience discovery is look-alike service.Its positioning is the thinking mode of “seeking people by people”.The advertiser provides the existing user group as the algorithm seed sample to the platform,and the audience discovery system automatically finds the potential target customer group in the platform.A complete audience discovery system usually consists of three objects: user,advertiser,and the user data.The main research content of this paper is based on the potential user extension of social relationship.Under the basic theory of recommendation algorithm,this paper proposes an audience discovery algorithm model based on social relationship.The audience discovery model mainly studied in this paper lets enterprises no longer need to label users based on attributes or interests,and using the user's social relationships can quickly mine potential users.The model uses the word2 vec algorithm to complete the vectorization of user attribute information and social text information.The node2 vec algorithm is used to calculate the neighbor nodes.The convolutional neural network is used to calculate the user feature vector set.According to the similarity rule,the models find the specific population with the highest similarity to the seed user from a large amount of data.Compared with the collaborative filtering algorithm,the model has higher accuracy and achieves the goal of optimization.Based on the model,the article combines the actual situation of a studio and the user data and commodity data accumulated in the previous operation.According to the business needs of the studio and its corresponding application scenarios,the customized audience system based on social relationship is designed and implemented.The system is equipped with micro-blog platform,which is divided into five modules: data collection,data processing,calculation of micro-blog propagation index,user audience discovery and result analysis.The system integrates data capture,data processing,user discovery,and data analysis.Users can obtain micro-blog propagation based on micro-blog interaction and adjust product promotion plans based on user feedback.The recommendation system designed and implemented in the paper considers the influence of social relationship on the recommendation result in many aspects.In the design of the algorithm,the influence weight of the secondary communication brought by the praise and forwarding in the social relationship is aggravated.Among the recommended results of the system simulation,the conversion rate is high,and the user's expansion is initially achieved.At the same time,the parameters are optimized according to the data analysis,in order to improve the accuracy and conversion rate of user expansion continuously,and achieve the purpose of better user expansion.