Research And Implementation Of Student Psychological Warning System Under Comprehensive Indicators Horizon

Student mental health issues have attracted increasing attention.The warning of student psychological problems is one of the important means to ensure their mental health.Traditional psychological warning methods mainly rely on psychological testing and case analysis,which cannot comprehensively and timely reflect students’psychological status.Relying on computer technology to achieve student mental warning is an innovative attempt to eliminate the disadvantages of traditional methods,which is beneficial to improving the comprehensiveness and effectiveness of warning work.It has profound significance for the reform,innovation,and deep optimization of psychological education in universities.The thesis aims to meet the requirements of student psychological warning from a comprehensive indicators horizon,and establishes four warning dimensions based on the experience of psychological construction in universities.Based on the experience of psychological construction in universities,four early warning dimensions are established,among which this thesis focuses on text sentiment analysis algorithm and grade ranking prediction algorithm related to the early warning dimensions of"emotional state" and "academic status".After analyzing the current state of research in related fields,it is concluded that the current state has the following shortcomings:First,there are obvious bottlenecks in the quality and sample size of datasets in the task domain,which are difficult to meet the needs of practical application scenarios.Second,there are limitations in the generalization performance of generic models and single models,resulting in the model performance advantages cannot be fully demonstrated.In this thesis,corresponding solutions are proposed to address the above problems,and the main work contents and innovations are as follows:1.A text sentiment analysis algorithm suitable for the social network domain was proposed,which further pre-trains the model using text data from the social network task domain in the pre-training stage and uses a dynamic-static word embedding aggregation method to enrich the semantic representation information of the text,and adds a trust domain smoothing control adversarial regular optimization method to adjust the loss function in the fine-tuning stage.Experimental comparison was conducted on a social network sentiment analysis dataset,and the findings indicated that the algorithm effectively enhances sentiment analysis for social network text classification.2.A grade ranking prediction algorithm based on multi-learner fusion was proposed,which improves the experimental dataset with the data enhancement method of sample overlay,makes the traditional regression problem of grade rank prediction into a binary classification problem,and constructs multiple heterogeneous learners in the form of base learners and meta learners to obtain a two-layer multi-learner fusion classification prediction model.Experimental comparisons are conducted on the improved student dataset,and the results demonstrated that the algorithm is effective in improving the prediction performance of student grade ranking.Based on the above research content,this thesis designs and implements a student psychological warning system under comprehensive indicators horizon,using the comprehensive index system of psychological warning as psychological evaluation criteria,and applying the proposed text sentiment analysis algorithm and grade ranking prediction algorithm to the system.After system testing,the system can help university educators conduct psychological crisis analysis of student-related data and report the psychological status of the students to university educators in different warning levels,achieving real-time and dynamic system support for student psychological warning work.