Research On Data Mining And Optimization Strategy Of Short Video Users

Because of short duration,fast dissemination speed,large amount of information and low threshold for creation,short videos meet the needs of people to use fragmented time for learning and entertainment.The steady and rapid development of the short video industry has attracted the attention of the advertising industry.If advertisers can screen out potential partners with great influence from a large base of short video users,it can save a lot of time and cost.From this perspective,this study designs a method that can efficiently mine short video user data and combine multidimensional user information to select cooperative users.The main work of this thesis is as follows:Acquisition and preprocessing of short video user data.Intercept and parse the interactive data packets between the mobile terminal and the server terminal of the short video platform,and encapsulate them into a data set after preprocessing.The Word2vec model is used to extract the features of the attention relationship between users.The user features are clustered using an improved CK-means algorithm.Pre-screening is performed within each cluster to reduce the data size for subsequent studies.Research on short video user data mining strategy.Some short video user data cannot be used directly before being processed.Therefore,this study uses data mining technology to conduct in-depth analysis of three parts of user information.1)Aiming at the problem that the number of users of the short video platform grows too fast,which leads to the lack of user authentication,the XGBoost model is used to learn the short video authentication tags to supplement the lack of authentication tags.2)Use an improved PageRank algorithm to analyze the complex network topology relationship constructed by the attention relationship,and obtain the fan attention of short video users.3)This study proposes a text sentiment classification model that combines generalized autoregressive pre-trained language understanding and self-attention mechanism to mine the sentimental tendencies of user comments.The simulation results show that the classification effect of the method proposed in this study is better than the comparison method.Research on short video user preference strategy.In order to screen out potential cooperative users from a large number of users,this study proposes a structural equation model-based optimization strategy and a multi-attribute decision model-based optimization strategy.The optimization strategy based on the structural equation model uses the input data to continuously modify the model parameters,and uses the path coefficients of the model to obtain the comprehensive ranking of users.The optimization strategy based on the structural equation model,a bidirectional projection mixed interval number optimization method combined with the principle of maximum entropy is proposed.The projection weight is obtained by solving the nonlinear projection model,and the degree of closeness is defined to sort users.The simulation results show that the optimization effect of the bidirectional projection mixed interval number optimization method combined with the principle of maximum entropy is ideal,and it can effectively avoid accidental errors.