Research And Practice Of Achievement Prediction Based On Comprehensive Analysis Of Multi-source Data

Using Data Mining Technology analyses the data of students on campus to make research related to student achievement has become an important topic in universities.It is benefit to schools that develop a personalized teaching plan for different students,standardize bad behavior and improve test scores by reasonable prediction of student's scores through data mining technology.The research work in this thesis is as follows:(1)Student's information data and behavioral data are collected from several data sources,such as from school's relevant departments and Internet.(2)After the preliminary analysis of collected data,the related data in thesis gets a more comprehensive pre-processing work by data cleaning,data selection,data integration.(3)The related data is further analyzed from different dimensions,and then the targeted suggestions are improved to school according to the each dimension of data analysis.Forward the academic performance and behavioral rules of the last semester,improved a new method that using changes in student life in two semesters predict the grades of students in this semester,and finally builds the predictive model by extracting the student features.(4)Through two kinds of classification algorithms which includes Naive Bayes and K-nearest neighbor classification algorithm constructing the performance prediction model,then making the model evaluation and comparison with two prediction models.Based on the data of students,the thesis finally using Data Mining Technology establishes a student performance prediction model.After verification,the model has achieved a good prediction effect,which enables school understand students more clearly and help students improve their scores or prevent their achievements slide down.