Design And Implementation Of Mining Based Uncertain Data Set Movement Pattern

In recent years, with the continuous development of science and technology andthe deepening of knowledge of data mining technology, uncertain data has been widelyattention. Uncertain data mining method has been widely used in lots of fields, forinstances: weather, economy, military, mobile telecommunications and other fields,uncertain data research has become the topic research in the field of data miningproblems.Two major problems are studied in this paper, respectively, uncertain dataclustering algorithm and mining group movement patterns algorithms on uncertain data.Recently, data clustering has been widely studied, but most are to cluster certaindata. In this paper, we studied clustering uncertain data on the base of the improvedtraditional clustering algorithm. We adopted a combination of two kinds of similaritymeasure which are based on the geometric distance and distribution distance, andintroduced the pruning method which is based on Voronoi diagram and R tree index tofurther reduce the costs. We designed the efficient uncertain data clustering algorithm.The experimental results show that our proposed UKLK-means algorithm is fullyconsidered the uncertainty of data distribution and improves the quality of the clusteringresults.In this paper, based on the moving objects tracking applications; we cluster thetrajectory of moving objects. Firstly, it uses markov model to describe the trajectory ofthe moving object. Then, we change the problem of clustering trajectory into theproblem of querying the nearest neighbor. That is querying which representativetrajectory (clustering center) is the nearest neighbor of the trajectories of moving objects.Within a given time period, query from the mobile trajectories, we can get the nearestneighbor of the trajectories which is the cluster center, and then we assign the trajectory to the corresponding cluster. Thus we can cluster the trajectories of moving objects.Thereby we can deal with these trajectories as a group for the unit in the later work. Ifso we can save a large part of energy, particularlly in the energy constrained sensornetworks. Experimental results show that our proposed mobile trajectory clusteringalgorithm can effectively consider the uncertainty of objects; it improves the accuracyof trajectory clustering and has a good scalability.