| As an institution of higher learning,the core is to cultivate high-level talents.At present,educational informationization of each college has the great development and accumulates lots of educational data,which offer the foundation on data mining to study high-level undergraduate talents.However,scholars seldom give the in-depth analysis on the educational data.The results obtained by data in undergraduate talent cultivation process through data mining are applied in daily talent cultivation management to make a decision and management,professional construction and layout based on data and continuously track and evaluate cultivation quality of teachers and students.For this reason,this paper does a study as follows:1.The employment questionnaire investigation for graduates has been enacted.Afterwards,online questionnaire management system has been used to conduct questionnaire investigation to the undergraduates of grade 2012.Finally,this paper gives the data pretreatment on data of graduates’ course performance,employment and internship,compares with different pretreatment methods,and uses the association mining algorithm for mining.To discuss the correlation between courses in the professional cultivation scheme,as well as the correlation between courses and students’ employment and quality.2.Outlier algorithm analysis is used to analyze the students with abnormal learning states in School of Computing in grade 2013.Then,the performance of students in grade 2012 will be used to to set up the multilayer neural network model.The performances of 15 core subjects in the first five semesters for junior students will be utilized to predict the performances of the target subjects of the sixth semester.compares with accuracy of different gradient descent algorithms in neural network training process,and confirms the optimal gradient descent algorithm.Finally,the students with abnormal learning status in grade 2013 will be predicted of their performances in the sixth semester,so as to guide the students to have emphasized studying to the subjects with failing risks. |