Study On Psychological Stress Assessment Algorithm For College Students Based On Data Mining Technology

Psychological stress is a kind of psychological stress reaction when people are faced with a difficult situation.Research shows that the right amount of stress can lead people to progress and reach their potential.For students,proper pressure can improve their learning efficiency and is conducive to their growth and development.However,too much psychological pressure may bring physical and psychological pain to students,and even lead to suicidal behavior.If students with abnormal psychological pressure can be found in time,the school can provide timely help and intervention to relieve the psychological pressure.At present,psychological stress assessment is mainly conducted by regular questionnaires.However,this method can only reflect the psychological pressure of the participating students at that time,and cannot obtain the pressure state of the non-participating students.In addition,students may fill in the questionnaire perfunctorily or deliberately fill in the wrong information to conceal the real situation.Therefore,this thesis studies the method of using data mining technology to evaluate the psychological stress of college students.Some data(such as students’ personal information and campus daily life data)used in this thesis are related to students’ privacy,and the data representing students’ identity information has been encrypted before analysis.The main work of this thesis includes the following two contents:The first is to propose a method of psychological stress assessment for college students based on the daily data of campus.This thesis firstly identifies the data sources that can be used for psychological evaluation in university campus,including personal information of students,records of course scores,records of campus card consumption,and records of financial aid.Then the calculation method of extracting features from these data sources is given,and finally the features are acquired to train the machine learning model.Secondly,an ensemble sampling neural network(ES-ANN)model is proposed for unbalanced sample data.As the proportion of students with greater psychological pressure is small,serious data imbalance is prone to occur in the training process.To solve this problem,we propose an ensemble sampling neural network(ES-ANN)model.The model is composed of several neural network models,each of which uses an undersampling algorithm and then integrates the classification results of each neural network with the integrated learning technology.At the same time,in order to study the difference between supervised learning and unsupervised learning,LOF algorithm is also introduced for comparison.Experiments show that ES-ANN’s G-mean value is about 8 percentage points higher than the best benchmark algorithm,and F1 value is about 4 percentage points higher.Compared with the traditional method based on questionnaire,the campus daily data has higher authenticity and more real-time,which is helpful for the school to find the students with great psychological pressure in time.The study of this thesis shows that this idea has potential feasibility and is worth further verification and improvement in practice.