Moving Objects Frequent Patterns Discovery With Time Stamp

With the rapid development and large-scale applications of mobile devices,wireless sensor network technology, RFID devices and GPS satellite positioningtechnology, it is convenient to collect vast amounts of moving objects trajectory data.Trajectory data includes a lot of potential and meaningful knowledge, how toeffectively extract useful knowledge from trajectory data is a problem need to besolved urgently in trajectory data knowledge discovery research field. As an importantresearch topic of trajectory data knowledge discovery, moving objects frequentpatterns discovery has important applications in location-based services, privacyprotection, location forecasting and many other fields. This thesis focuses on movingobjects frequent patterns discovery with time stamp, the contributions of this thesisare shown as follows.(1) Regions of Interests discovery with relative time constraintFor the problem that uncertain results and low efficiency of traditional clusteringmethods in Regions of Interests discovery, this thesis proposes a method of Regionsof Interests discovery with relative time constraint. This method firstly does even griddivision for moving objects space, then computes even grid density, at the same time,using relative time as a constraint condition, segments trajectories which meet therequirements of grid density, What is more, obtains candidate Regions of Intereststhrough extension dense grid, if a trajectory segment and candidate Regions ofInterests overlap, then the trajectory is regarded as support trajectory for candidateRegions of Interests. And if the number of support trajectory is not less than minimumsupport threshold, then the candidates are Regions of Interests. Experiments show thatthe method can get good results of Regions of Interests, and maintain high miningefficiency.(2) Moving objects frequent patterns discovery with time constraintTraditional moving objects frequent patterns mining results do not contain timemostly, therefore, based on Regions of Interests with relative time constraint, thisthesis puts forward with a method of time-constraint moving objects frequent patternsmining. Firstly, the method translates Regions of Interests with relative time constraint,and obtains translated sequences, then conducts projection of these sequences andcalculates projection, furthermore, extracts candidate frequent patterns. What’s more,adds time constraint. If candidate frequent patterns support is not less than minimum support threshold, and meets time limit at the same time, the candidates are frequentpatterns. Experiments show that this algorithm maintains better efficiency in time andspace performance, compared to traditional frequent patterns mining algorithms.(3) Design and realization of moving objects frequent patterns discovery withtime stamp prototype systemBased on theory in this thesis, we design and realize moving objects frequentpatterns discovery with time stamp prototype system. And then, using real datasetsvalidates accuracy and efficiency of this algorithm. Experiments show that the systemcan get good visualization effects, which show good mining results of the algorithm.