A Personal Important Places Inferring Method Based on Location Entropy Preprocessing and PMM

Important places inferring, the key technology in the intelligence mining from the big data, has important applications in tracking target users. In the paper, a method for important places inferring, aiming at LBS users is studied. The method includes pruning process and parameter estimation of Periodic Mobility Model. In the step of pruning process, to solve the problem that the location of target user in rest days does not match PMM model, a pruning preprocessing method based on position entropy was proposed. In this method, the data with week periodicity was removed using position entropy as reference, ensuring the reliability of data. In the step of parameter estimation of PMM, starting with the pruned data, using EM algorithm to estimate the parameter, to infer the target users´ personal important places successfully. Experimental results show that, compared to the method of largest cluster and the method of parameter estimation without pruning, the improved method has higher accuracy and less estimation error, which is valuable in important places inferring to some extent.