The Analytics Of College Students' Behavior Basced On Machine Lerning And Big Data Technology

With the development of the society, information technology is thriving fast, and here comes to the era of digitalization. The booming technology of Computer science, together with the internet technologies,are producing huge volume of data that not only contain countless fortune,but also challenge the traditional cognition. Big data technology is originated from the eager for mining the principles and secrets of the data,and it has been implemented widely into various fields to find out the information, knowledge and wisdom beneath.Recently, campus data has drawn an increasingly attention upon students' management staff. Through the analysis of all kinds of data collected from students they hope to implement the theory related to machine learning and big data technology, and analyze the students'campus data to reach to the law of growth of students and find out danger in time so that they can create a more scientific and humanized managing strategy and achieve the objective of teaching accordingly.This thesis is aiming at analyzing all kinds of university data that deriving from students' daily life to provide theoretical support for university officers about their students' management work. By researching students' group behavior pattern analyzing technology, we explored how to use the machine learning methodology of K-means and NMF algorithm to classify university students group behavior pattern. We also used the sentimental polarity classification algorithm based on dependence relationship to study how to categorize the students'sentimental comments on the internet. Finally, by using abnormal behavior classification technology, we tried to deal with students'emergency as soon as possible.Apart from the research and introduction of relating theoretical knowledge in this thesis, we accomplished some experiments based on machine learning with K-means and NMF algorithms. We came up with group behavior pattern classification results and characteristics, and we implemented sentimental polarity classification based on dependence relationship methodology to classify some of the real students' comments from a specific university BBS. After decided the student's sentimental tendency, we studied the following steps used SVM algorithm theory.Firstly, we found out the financially difficult collegians according to students' expense condition. Then we predicted whether the student should be paid special attention to or not combining with some student's information, and we compared the prediction performance of single computer with that of Spark platform.