Association between community psychological label and user portrait model based on multimodal neural network

Abstract

By analyzing traditional deep learning multimode retrieval methods, an optimized multimode retrieval model based on convolutional neural network is established. This article proposes an innovative semi-supervised social network user portrait analysis model (UPAM) based on user portrait model, which integrates users’ social information and some known user attribute information (such as educational background and residence) into a unified topic model framework. Finally, a semi-supervised user portrait analysis method based on user social information and partial known user attribute information is proposed. According to the correlation of user attributes, the cross-validation method is used to train model prediction task and improve the prediction effect. In the first-level model, using a different model to extract the features in the user query, the basis of the second hierarchy model, Stacking is used to further integrate characteristics, finally realizing the attribute population forecast, and experimental verification showing the proposed model’s effectiveness in various attributes of a population.