Research Of Algorithms For Co-occurrence Patterns Mining

With the rapid development of data acquisition technology, the amount of data is growing dramaticly. We are surrounded by a growing number of smart phones, computers and softwares, which leads to the dramatical increase of application of positioning technology and location-based service. So, whether in real-world applications, or in various research field, it has produced a large number of spatial data. In these vast amounts of spatial data, there are many hidden information which has big benefit to decision-making. However, the study of spatial co-location, spatio-temporal co-occurrence and uncertain data exists in many fields, such as communications industry, volunteered geographic information (VGI) and biological area. At the present stage, the relevant algorithms for spatial data mining have low efficiency and the selection of threshold in different application fields is difficult, and most algorithms can not handle uncertain data effectively. Aim at the low efficiency of existing algorithms for mining spatial co-location patterns under uncertain data, this study proposes the improved algorithm for mining spatial co-location patterns under uncertain data, which is based on Filter-Refine method and join less method. The proposed algorithm improves the efficiency of mining these patterns. On this basis, the definition of spatio-temporal co-occurrence pattern under location uncertainty is proposed in this paper. The integrity and correctness of this model have been proved in this paper. Accordingly, this paper also implement the algorithm for mining the spatio-temporal co-occurrence patterns under location uncertainty. Experimental results by real data sets show this proposed algorithm is effective and reasonable. On the basis of the study of spatio-temporal co-occurrence patterns and partial periodic spatio-temporal co-occurrence patterns, this paper also proposes the study of periodic spatio-temporal co-occurrence pattern, this study proposes the improved algorithm which is based on TOP-K%method, in order to solve the suffering selection of threshold in different application fields. Experimental results by real data sets show this improved algorithm is reasonable and can effectively mining periodic spatio-temporal co-occurrence patterns from spatio-temporal data sets.