Access patterns mining from massive spatio-temporal data in a smart city

Abstract

Facing with massive spatio-temporal data, the traditional pattern mining methods fail to directly reflect the spatio-temporal correlation and transition rules of user access in a smart city. In this paper, we analyze the characteristics of spatio-temporal data, and map the history of user access requests to the spatio-temporal attribute domain. Then, we perform correlation analysis and identify variation rules for access requests by using regional meshing, association rules and ARIMA in the spatio-temporal attribute domain, for the purpose of mining user access patterns and predict the user’s access request. Experimental results show that our pattern mining algorithms is simple yet effective, and it achieves a prediction accuracy of 84.3% for access requests.