Research And Implementation Of Personality Feature Prediction Algorithm Based On Semi-superviserd Learning

Personality characteristics information is a high-level summary of human individual characteristics,and it is also a scientific quantitative standard for distinguishing differences between people.Accurate analysis of personality characteristics is of great importance for grasping individual psychological states,and then predicting and analyzing their behavior.key meaning.The personality trait prediction task is playing an increasingly important role in solving social network problems due to its wide range of application scenarios.Such as understanding users’ mental health,providing personalized medicine,online learning,and assisting various types of recommendation systems.Existing research methods usually use each posting document of the user as the boundary,and use deep learning models such as CNN and RNN for forward propagation.However,these methods fail to effectively utilize prior psychological knowledge and fail to target personality.The current situation of small feature data samples is generalized.In this regard,this paper carried out the research and implementation of the following contents:1.A graph neural network-based personality feature prediction algorithm PerGCN is proposed,combined with the LIWC dictionary in psychology to perform multi-level mining of user text information.The graph is composed of three categories of nodes:user,vocabulary,and psychological keyword categories,so that the user node is directly adjacent to the word node,skipping the "document" level representation features and using the LIWC dictionary to extract user text features.The feature representations of user nodes and psychological word category nodes are weighted and summed through the graph convolution module and the attention module.The effectiveness of PerGCN is verified by comparative experiments and ablation studies.2.A semi-supervised learning graph neural network model semiPerGCN is proposed for personality trait prediction.Regularize the learning process of labeled data by making full use of a large amount of unlabeled data to constrain the prediction results of the model.Reducing the impact of data noise also reduces the dependence of model performance on the size of the dataset to a certain extent.On the basis of quasisupervised learning based on personality trait labels,consistent regular expressions are added to make the model train in a semi-supervised learning manner.The anti-noise ability of the model is enhanced by largescale unlabeled data,which makes our model relatively more robust and generalizable.The advanced nature of semi-PerGCN is verified by comparative experiments and ablation experiments.3.Design and develop a personality trait prediction system,which realizes the functions of user management,data storage,personality trait prediction and result visualization,and integrates multiple algorithm components and provides a friendly interactive operation interface to promote the research and development of personality prediction tasks.