Based On Probability Interval Could World Space Co - The Location Pattern Mining Research

With the development of data acquisition technology, a lot of spatial data with resulting, more and more scholars are concerned about spatial data mining. The spatial co-location pattern mining is an important branch of spatial data mining,it is used to discover the association relationships among the spatial features. In many applications, the collected data is often uncertain, we usually using probabilities to describe the uncertainty of the data and this probability is generally expressed in a precise real number. However, in some practical applications, it is difficult to get an precise probability or do not receive the precise value can only get an approximate value of probability. Because of imprecise probability is inherent in practice, the imprecision of the probability is necessary to be considered when using probability to describe the uncertainty of the data. In this paper, we will consider how to mine spatial co-location patterns from uncertain features whose spatial instances are associated with existential probability intervals.First, we introduce the meaning of uncertain data and concepts of imprecise probability.Second, the basic concepts of spatial co-location pattern mining and two important mining algorithms are introduced in this paper.Third, for the uncertain features whose spatial instances are associated with existential probabilities, the concepts of possible world model, prevalence probability of pattern, and probabilistic prevalent co-location were introduced in this section.Fourth, to consider the uncertain features whose spatial instances are associated with existential probability intervals, the possible world model was extended into probability interval. We proved the set of probability intervals of all possible worlds is reasonable and feasible.The probability interval of a possible world was transformed into point probability. The concept of prevalence point probability of a co-location was defined. Then, the basic algorithm of mining probabilistic prevalent co-location patterns in the situation of probability interval was presented. The basic algorithm is time-consuming, with some optimization strategies were proposed, the improved algorithm was studied.Fifth, we make a lot of experiments on synthetic data sets and real data sets; it is showed that our algorithms are effective and significant.Finally, the summary of this paper and future work were presented.