Research And Application Of User Profile Technology With Two-layer Model Fusion In Social Media

Today's social networks have become a window for people to obtain information,show themselves,and promote marketing.Everyone in modern society will leave a lot of digital footprints on social media platforms.How to use the massive data in social networks to provide valuable services has become a research hotspot.User profile technology refers to mining user digital footprint information to classify users,thereby establishing the overall picture of users and providing data support for providing better services.User portraits are the basis of the recommendation system.Most of the existing research is about user portraits in dimensions such as user age,geographic location,and interest preferences.Aiming at the problem of low accuracy and generalization of the traditional user profiling model,this paper proposes a user portrait model with a two-layer fusion algorithm.The personality dimension is introduced into the user portrait to use the portrait to improve the recommendation effect.Heterogeneous data is used to construct user portraits and how to apply user portraits to recommendation systems.The main research contents are as follows:(1)For text data and social network structure data of user on social platforms,it is proposed to extract features from different views,including the use of LIWC psychology dictionary,LDA topic model,etc.to extract text features,and the use of network representation learning to extract user network structure features.For these features,the first layer of the user portrait model uses machine learning algorithms to predict personality attributes.Different modal features have a semantic gap.In order to make full use of the different modal features,a neural network model of multi-view fusion is proposed to generate deep interactions between multi-modalities,which helps to learn the correlation representation between the multi-modal features to predict user personality labels.(2)A user portrait method using an integrated learning algorithm and a stack is proposed,and the prediction result of the first layer is used as the input of the second layer,which further improves the accuracy and generalization of the model.The experiment compares the integrated algorithm with a single model As the effect of the two-layer classifier,the results show that the integrated algorithm has the best prediction effect.(3)Aiming at the problem of data sparsity in collaborative filtering recommendation algorithm,this paper proposes an improved matrix decomposition algorithm,which combines user portrait with matrix decomposition algorithm,and introduces personality attributes from different dimensions into matrix decomposition algorithm.Experiments in relevant scenarios and data sets prove that the improved collaborative filtering method proposed in this paper is an effective means to alleviate data sparseness.