Prediction of population distribution on campus based on historical location data

In this paper, we proposed a prediction model to estimate the number of students given the place and the time based on location data generated from Shanghai Jiao Tong University. The location data originated from the records of users consumption in campus and the WIFI data implicitly contain massive spatial and temporal information of human activity and mobility. Prediction of population distribution will surely benefits both the students and the faculty who work or live on campus as well as the university administrators in daily planning, scheduling or policy making. Four key factors are considered in the proposed prediction model to perform the estimation of peoples, i.e., time of the day, day of the week, weather condition and holiday & exam information. Compared with existing work the holiday & exam factor were introduced into the prediction model for the first time considering the context of campus. Our inference engine is based on the Naive Bayesian classifier with the prior knowledge getting from historical data and adjusted using error-based learning. We manage to give a prediction of peoples distribution on campus 24 hours ahead based on the original data collected from over 30000 WiFi and smart card users of Shanghai Jiao Tong University. With holiday & exam information being considered, the proposed method exhibits 18.9% improvement comparing with the existing one without that information.