Research On Early Warning Model Of Depressive Tendency Of College Students Based On Social Platform Data

In recent years,the prevalence and seriousness of college students’ psychological problems have emerged,and in view of the high correlation between college students’ mental health and their learning efficiency,learning effect,and comprehensive physical and mental development,it is necessary for colleges and universities and even society to further enhance their attention to college students’ mental health.In the social network platform,the psychological crisis early warning mechanism can take the lead in discovering students’ psychological problems,prioritize matching educational resources for counseling,improve students’ learning environment,and enable students to achieve comprehensive physical and mental development.Based on this,this paper takes college student users in social network platforms as research objects,studies mental health prediction technology and intervention strategies,and specifically applies them to the direction of depression tendency prediction.Taking the text data processing method in the social network platform as the starting point,the existing literature is sorted out and summarized,and it is found that there are still the following problems in this field: 1)Theoretically,because the research field belongs to the interdisciplinary disciplines of psychology,education,artificial intelligence and other interdisciplinary disciplines,the construction of early warning system indicators is not complete in terms of theoretical integration between various disciplines,and there is a lack of Chinese text datasets for detecting psychological states;2)Technically,the early warning system is still not deep enough in the selection and optimization of models,and there is a problem of low generalization ability;3)From the perspective of cross-integration,it is difficult to integrate theoretical knowledge and emerging technologies.In view of the above three deficiencies,the following research is carried out:1.A Chinese text dataset for the prediction of depressive tendency of college students was constructed.A total of 1,029 Weibo users were read,and 400 valid Weibo users were obtained after screening.The crawler technology was used to obtain 104,670 pieces of personal information and Weibo text data of 400 users,and 50,431 valid microblogs were obtained through data cleaning,word segmentation,and stop word removal and other data preprocessing technologies.By constructing Chinese text dataset,it helps solve the scarcity problem of Chinese text training data sample database,and provides a data foundation for the training of mental health warning models.2.An effective feature extraction method based on depression knowledge graph is proposed.By analyzing the characteristics and in-depth knowledge relationships of college students in terms of gender,language,behavior,emotion,etc.,the depression knowledge map was constructed,and the characteristic information with the degree of depressive tendency discrimination was extracted.In order to verify the validity of the extracted feature information,it is trained as the input of the SVM classifier prediction model,and the results of the SVM prediction model in the reference literature are compared experimentally.The results show that the feature information extracted in this paper improves the accuracy of depression classification by 8% to 93%,indicating that the processing method of feature extraction based on knowledge graph is more effective under the premise of using the same SVM classifier for text classification.3.A deep learning model CNN-GRU that is more suitable for classification is proposed.In order to achieve better prediction effect,an improved classification model is constructed under the premise of keeping the feature extraction method unchanged,and the advantages of CNN and GRU are combined to further mine text time series information and capture information such as users’ emotional changes and topic evolution in Weibo.Compared with the prediction data of different deep learning methods in the published literature,the CNN-GRU model has a 10% performance improvement and can achieve an accuracy of 96%,indicating that the improved model has stronger adaptability and better classification performance in the prediction problem of depressive tendency.All in all,the two classification model algorithms constructed in this paper have obtained good results in the verification set,and the feature extraction method based on knowledge graph and the classification model based on deep learning can effectively improve the classification performance of the model.This provides valuable reference and enlightenment for future research directions,and also provides new ideas and methods for the improvement and optimization of mental health early warning system.In addition,in order to help the psychological crisis early warning model achieve better application effect,this paper designs intervention strategies from the perspective of universities,teachers and technical support theory,which provides new strategies for the field of mental health education and helps to better solve the mental health problems of college students.