Research On Clustering Analysis In Spatial Date Mining

With the development of information technology and data acquisition technology,the massive history data which are related to spatial position is collected and maintained by researchers. In view of these huge data, people urgently hope to find out a powerful data analysis tool to seek the knowledge which is hidden in spatial data. Spatial clustering analysis is one of the powerful data analysis tools, it not only can find clustering rule that is hidden in the ocean of data, but also can be combined with other kinds of data mining problems to dig out the deeper level of knowledge.The major research contents of this paper are as follows:1. This paper research the basic theory of clustering analysis, traditional and the new proposed the spatial clustering analysis algorithms.2. Due to the local convergence and stagnation of queue intelligent clustering algorithm, this paper uses hierarchical clustering algorithm to cluster the solution, to improve the quality of population diversity so as to enhance the clustering. Using the UCI machine learning repository of IRIS, CMC and Wine data sets to text, the F value(clustering accuracy) of improved algorithm is better than k-means and k-medoid,and it shows that the improved algorithm has better clustering effect.3. 3. In terms of defect of measurement mean without considering the spatial distribution of the object of degree of membership in fuzzy k-modes clustering, this paper introduces double measurement method of distance and density into fuzzy K-modes clustering algorithm to improve it in a more reasonable way. Using the UCI machine learning repository of vote, mushroom and zoo data sets to test, the F value of improved algorithm is better than before improvement, it shows that the improved algorithm has better clustering effect. Finally, this paper applies the improved algorithmto spatial data which is collected from baidu map to cluster, the clustering result is good.It shows that this algorithm is feasible and validity in application of spatial data clustering.4. Finally, this paper uses the mobile phone GPS data of two Finland cities and the track GPS data of Shanghai buses and taxis, which contain information of the latitude and longitude, using the intelligent queue algorithm based on dynamic clustering to cluster analysis, the clustering results of the mobile phone GPS data of two Finland cities can help the local mobile operators make the position of cellphone tower; the clustering results of the track GPS data of Shanghai buses can help the bus company determine the position of dispatching station; while the clustering results of the taxi GPS data can help the taxi company determine the position of branch location.This paper improves intelligent queue algorithm and fuzzy K-Modes, through the experiment, the clustering results are good. And this paper also takes the improved intelligent queue algorithm applied to the actual problem of spatial data mining, solving the practical problem of spatial data mining.