Score Prediction And Student's Psychological Condition Based On University BBS Data

Tool's improvement is an important symbol of human society development.From the stone to the bronze,from steam power to electricity power,they marked the progresses that human have made on the ability of changing the nature.University is the cradle of training high-quality talent,how to training the talent efficiently is the focus of the education reform in the world."Promoting the information technology level in education field" is listed as an important direction of education reform in 2016 according to the Ministry of Education's the JiaoGao[2016]-2 guide file.Colleges and universities have lots of accesses to huge student behavior data with the information technology develop quickly today.We can know student which is lost in time from the data,understand the students' social relationships through the data and even get the student's final grade qualitative prediction in the mid-term from the data.Many valuable research results were applied to explore the education reform when the study of big data in the field of education goes deeper.We mine student BBS data of the University of Electronic Science and Tchnology of China deep,make every effort to find the correlation between students' psychological condition and students' achievement with the students' BBS activities and give the deep reason of students' academic performance change as well as the psychological change from the perspective of students' BBS data.We will find out the valuable things for university's education reform from making use of data mining technology.The research content of this article is divided into two parts:1)students' mental modelWe collect more than 27,000,000 data records from campus BBS of the University of Electronic Science and Technology of China.Then we build the students' mental model from students' emotion tendency,students' interest and students' social relations.In terms of sentiment analysis,we use traditional sentiment analysis method which is based on dictionary matching to extract the emotional expression of text in the BBS.Besides,we take advantage of the SO-PMI algorithm to build student's BBS domain emotional dictionary which contains 3000 words;In the student interest analysis,we use the text which containing a specific theme as the training data and classify 183 different forum topic to 43 clear theme.We combine the chi-square method and the LDA topic extraction method to reduce the feature dimension.After that,we train a text classification model to classify more than 600,000 campus BBS text for their own subject.Finally,we use psychological questionnaire survey data to test the students' mental model.2)student score performance predictionStudents' course score data contains the score of all the student in the UESTC.We predict student's score rank with students' behavior characteristics by using learn-to-rank model and get good result on two different grade in different semester.Then we put the students BBS activity feature and some indicators of mental models into the student performance prediction model,found that the result of the prediction increased by 6.59% and 4.83%.Respectively it proves that the students' BBS behavior is valid for student performance prediction.From the two points above,our research explored how to mine the value of the college BBS data in the direction of education informatization efficiently and had positive impact on how to using campus BBS data for mining students' features by different views.We find a new way to student performance prediction and have a good result in student performance prediction.Exploring the relationship between student behavior in campus BBS and student's psychological condition is a way which little researchers try.It has promoting effect on the construction of college students' mental and psychological intervention and university's informatization construction.