| In recent years,due to the impact of the new coronavirus epidemic,academic employment pressure and family environment,the incidence of mental health problem is increasing year by year,and the number of disappearances and suicides caused by mental health problem is also increasing.In order to reduce the impact of mental health problem,it is necessary to detect mental health status in advance.In this thesis,big data mining and analysis techniques are used to predict user’s mental health status.At the same time,considering the risk of data leakage,privacy protection technology is used to strengthen the security of data.The improved storage method for frequent pattern growth(ISM-FP-Growth)algorithm is used to analyze the characteristics of mental health status,construct a mental health status prediction model,and design a mental health status prediction system.Achieve high-precision prediction of user mental health development trend.The main work of this thesis is as follows:(1)Aiming at the problem of personal privacy protection of evaluation data in the field of mental health,this thesis designs a privacy protection scheme based on k-anonymity to protect the privacy of mental health evaluation data.In this thesis,k-anonymity algorithm is used to summarize or desensitize the confirmed sensitive fields in the evaluation data.The experimental results show that in the processed evaluation data,each record is completely similar to the relevant attribute values of other k-1 records,thus effectively solving the security problem of mental health evaluation data.(2)Aiming at the problem of low time and space efficiency of association rule mining algorithm,this thesis proposes an improved storage frequent pattern growth(ISM-FP-Growth)algorithm,which improves the algorithm by establishing a two-dimensional matrix stored in the scanning database,arranging and grouping data items,and constructing a new FP-tree structure.The experimental results show that the improved algorithm has higher spatio-temporal mining efficiency.Based on the data in this thesis,the factors affecting user’s mental health status and their correlation degree are excavated.It is concluded that the correlation degree between anxiety symptoms and depression symptoms reaches 0.77,the correlation degree between anxiety symptoms and obsessive-compulsive symptoms reaches0.71,which lays a foundation for the feature construction of the mental health state prediction model of anxiety symptoms.(3)Aiming at the problems of high spatial complexity and low accuracy of prediction results,this thesis designs an improved XGBoost algorithm to predict mental health status,and uses Covariance Matrix Adaptation Evolution Strategy(CMA-ES)to optimize the algorithm.By searching the optimal parameter combination,setting a more scientific parameter range and space,and combining fuzzy logic to reduce the misjudgment of the prediction results of the binary classification model,the overall performance of the algorithm model is significantly improved.After several rounds of iterative training,the hyperparameter value of the model at the best performance is finally calculated.The results of comparative experiments show that the accuracy of the mental health status prediction model based on XGBoost algorithm is 93.85%without optimization.After optimization,the accuracy of the prediction model can be improved to 96.03%.This shows that the optimization method proposed in this thesis has achieved remarkable results in improving the performance of the algorithm.(4)Aiming at the functional defects of the current mental health assessment system,such as the inability to predict mental health status in advance,this thesis designs and develops a set of mental health status prediction system.Based on the Django framework,the system has three modules: evaluation management module,prediction management module and system management module.It involves online evaluation,privacy calculation,data analysis,prediction result visualization and other functions.It realizes the function of predicting user’s mental health status and presenting visual results,and achieves the expected goal of the system. |
