Research On Campus Implicit User Behavior Data Mining

As the development of information system, people produced a large amount of data. Compared with explicit user behavior data, implicit user behavior data are pro-duced by not intended. For example, such data set could be generated in social voting activities, shopping, borrow or return books etc. They can reflect the behavior of indi-viduals. In this paper, we utilize datasets from undergraduate students in University of Science and Technology of China, including behavior data on controllable social net-work experiment, smart campus card record, library, campus email, weibo and renren to complete the following research.1. Biased Voting Behavior Experiment on social network This paper introduces the use of a controlled experiment on social network user behavior data to find how network structure, incentive mechanism and social phe-nomenon can influence the ’vote’ social activity. Analysis the behavior of indi-vidual people such as stubborn.2. Using students in and outside campus behavior data generate "health report" system This paper introduces the use of multiple sources of implicit in and outside cam-pus user behavior data to discover living style of students in University of Science and Technology of China. Improvement recommendations are put forward based on routine irregular students. It defines our sleep, eating time rules and algorithm to generate a health report. The implementation of the system are introduced, in-cluding the system architecture, data crawler, CakePHP website construction for user interaction. Because the system is hosted on Microsoft’s Windows Azure cloud computing platform, the architecture of Azure and features are also intro-duced.3. Use of smart campus card data to warn students’ canteen emergency and abnormal left school Using USTC smart campus card data analysis are introduced in dining room turnover and daily life style of students. Using historical data of the canteen in-come for warning the canteen emergency. Using record of smart campus card data to construct students history time series forecasting model for warning stu-dents’abnormal left school. In "research" section, we introduced the three ex-ponential smoothing algorithm and its evolution process, Box-Jenkins algorithm and ARIMA model. In the experimental analysis section, we prove it’s conform to the requirements of being stationary and the periodicity of the above model through the smart campus card shopping data visualization, and introduces the results of our model fitting and forecast the case example analysis.