Study On Accurate Marketing Algorithm For Unicom Big Data

With the rapid development of mobile Internet technology and the widespread use of applications,the analysis of user behavior and the implementation of accurate marketing recommendations have become new demands in the industry.However,the user's portrait is more difficult to describe accurately,the user beha'vior data has confidentiality requirements,and the operator's data volume is large,which affects the processing effect.Therefore,the accurate marketing based on the operator's big data has become a hot research direction.Aiming at the needs of enterprises and academic challenges,this paper conducts theoretical research and experimental analysis on user portrait causal model and user behavior analysis algorithm based on this model.Label and user portrait have been redefined,and the concept of label operator has been introduced,which converts the user behavior association into a label expression representation,which can directly process the operator's data by label expression representation.Label detection algorithm and user behavior prediction algorithm are proposed.On this basis,the accurate marketing algorithm for China Unicom big data with the above two algorithms as the core is obtained.The algorithm firstly filters out a large number of invalid labels quickly and effectively,and then uses the Bayesian classification model to calculate the probability of user consumption,so as to give the arrangement of the user's consumption possibility,and achieve the purpose of accurately obtaining potential customers.Taking the precise marketing of automobile industry as the application background,this paper chooses the log of mobile Internet users as the data source,and carries on the experimental analysis to the algorithm proposed in this paper.The experiments show that the accurate marketing algorithm for China Unicom big data has higher processing efficiency and prediction accuracy.The feasibility and validity of user portrait causal model and behavior analysis algorithm based on this model are verified.