Study On Co-location Pattern Mining Methods Of POIs From Internet

With the rapid development of Internet technology and geographic information services, Internet-based electronic map has been applied aspects of social life gradually. POIs from Internet have features of massive, heterogeneous, distribute widely, updated frequently etc as an important part of electronic map. Study on distribution characteristics of massive POIs from Internet and excavating the low of knowledge hiding behind the massive geographic information, becoming the focus of attention in daily life gradually. In this case, co-location pattern mining is an effective way to solve this problem. Co-location pattern mining of POIs from Internet can extract the distribution characteristics of massive POI and the relationship of POI, to discover the information and knowledge hiding behind the massive and disorderly POIs. It also can help users find the information clearer and more intuitive, and provide support for the various geographic decisions. Hoever, the most of existing method of co-location pattern mining can not extract the distribution characteristics of POIs and their relationship quickly and efficiently when the POIs are massive. So in this paper, two methods of co-location pattern mining based on MapReduce and multi-threaded parallel are proposed to improve the efficiency of co-location pattern mining with massive POIs.In this paper, firstly, studies of the traditional association rule algorithms and the co-location pattern algorithm. Run the Apriori algorithm which is the classical association rules algorithm in Visual Studio2008platform using C#programming language demonstrates the accuracy of the algorithm through examples, sums up the thoughts of several improved Apriori algorithm. Then some concepts of co-location pattern and join-based, partial Join, join-less co-location pattern mining are introduced by some examples. Secondly, achieves co-location pattern mining based on multi-threaded parallel for massive POIs from Internet. First, accomplishes the storage of massive POIs from Internet via MongoDB storage model. Next, improves efficiency of querying geography markers by the establishment of geospatial indexes. Then, the data sets are divided into a plurality of data blocks which will execute multithreaded parallel Apriori computing through multi-threaded parallel processing programming model. Finally the co-location pattern and co-location rules which are in line with min support and min confidence are got. At last, takes the POIs from Internet of Beijing, Shanghai, Guangzhou and Shengzhen for example, mining the co-location pattern and co-location rules in different conditions by set different parameters to analysis the distribution characteristics of POIs and the degree of association between different types of POIs. The application proves the co-location pattern mining based on multi-threaded parallel can reflect the distribution characteristics of geographical entities within a certain range effectively and has strong application value...