An Analysis On User’s Behavior Of Campus Network Based On Hadoop

The network has become an essential part in everyone’s work and life. At present, this situation is especially evident in college. As a basic platform of teaching, research and information services, the campus network has become one of the important indicators of informationconstructionatcollege and the wisdom campus. With the increased use of the network on campus, these bring about a massive online users’ behavior data in the form of logs. At the same time, with the growing number of users and increasing the size of the campus network, it gives rise to many problems for campus network to optimize and upgrade the management, daily operations and maintenances. In addition, student users overuse the network, which bring great harm to their own learning, life and health. Thus the new requirements should be put forward for the College Students Education Management. Therefore, network behavior analysis of campus network users will help university network management departments to develop and improve a more reasonable and effective network management system and daily operation and maintenance strategies, and help teachers and students to improve the safe, fast and reliable network environment. It will contribute to understand and find students’ ideological trend, learning status and other useful information in a timely manner for the university student management department.The thesis is based on the campus network of the Northwest University for Nationalities, the study objects of users’ network behavior are clickstream logs of accessing to the network from Campus network core switch mirror port H3 C 12508 and user login data from Shen Lan gateway Srun3000.The research of this thesis are mainly as follows: 1. Distributed Hadoop experimental environment is constructed which includes Hive and Sqoop subitems. 2. The data of the campus network user behaviors are preprocessed and classified by User Category in Hadoop cluster. 3. According to Hive QL query technology, group work for campus network users are conducted from five aspects, that is the analysis of the number of online users in different periods, the analysis of user’s online time length, the analysis of the destination address of user ’s access, user’s online traffic analysis and analysis of students’ abnormal behaviors. In the analysis of the destination address of user’s access, Linux Shell scripting language is written for sort and statistics of destination address in order to speed up the destination statistics and sorting speed, which achives a good operating results. Through the comparison and research of the daily online time, the average daily use of network traffic and the number of daily online from the users of different grades, different level of training and different schools, the causes are discussed and analysed preliminary why students’ network abnormal behavior have abnormity, and some specific pinions and suggestions are put forward. 4. Group behavior of students are analyzed by cluster analysis. Firstly, the implementation method of this algorithm is found by the ideas of K-means clustering algorithm parallelization. Secondly, Mapper function and Reducer function are coded to realize K-means algorithm. Finally, from the aspect of the users’ online time, flow downward and upward flow, network behaviors of campus network users are researched by clustering. Users will be divided into five clusters in the result, and the causes of each cluster and its characteristics are analyzed in detail.In short, the related campus network users’ behavior research have important reference value and guiding significance for the university network management and student management.