Research On Group User Portrait Of Online Video Community Based On Danmaku

With the expansion of online entertainment,users pay more attention to online interactive behavior.User portraits provide support for the personalized service to solve the "pain point" of the contradiction between the growing unique needs of users and the supply of video resources.In view of this,this paper constructs the user portrait model of online video community from the perspective of danmaku,realizes the feature recognition and user behavior visualization of user group,and achieves the purpose of personalized service and accurate recommendation.Firstly,the model of danmaku emotion analysis is proposed.The danmaku contains users' negative,neutral and positive information.Mining the emotional information in danmaku is conducive to in-depth understanding of users' state.This paper introduces the preprocessing content of danmaku text mining,and proposes a model for sentiment analysis based on convolutional neural network and recurrent neural network.Through the process of model training and parameter optimization,the classification accuracy of the model on the training set reaches has improved.Secondly,the model of online video community group user portrait based on danmaku is established,which makes full use of the multi-dimensional information published by users to form a comprehensive and three-dimensional user portrait.In order to achieve this goal,the stratified and hierarchical label system of user portrait in online video community has been built;The crawler technology is used in the user portrait data acquisition and preprocessing stage;In the stage of user portrait model proposed and improved,RMS(Recently-Monetary-Sentiment)model is proposed as three indicators of user clustering.Each user is divided into specific groups by clustering algorithm,and the typical user features in each group are extracted;In the visualization stage of user portraits,python programming technology is used to draw the word clouds of each group of users.Finally,the user portrait model is applied to Bili Bili video community to construct the user portrait of the documentary "The College Entrance Examination" video community.According to the analysis of danmaku emotion,the number of neutral emotion is less,while the number of positive and negative emotion is more.The clustering results show that users can be clustered into three categories.Different user groups have different performance in publishing time,text length and emotional intensity.According to the user portrait results,suggestions are put forward for community accurate recommendation and personalized service.