Design And Implementation On Multi-density Clustering Algorithm Oriented Region Information

With the popularization of mobile terminals and Location Based Service, the location information of Internet users is more easily to be acquired and accumulated. Analysis of regional information hidden vast user data can help enterprises to make better arrangement for resource management, personnel allocation and service site construction, reduce the waste of resources allocation in administrative regions.In this paper, a latitude and longitude data set which extracted from a well-known classified information mobile platform is as the research object to individualize the category sorting based on regional information. Due to the reason that the degree of market coverage and user sharing frequency in the platform varies greatly, data in different regions has different distribution densities. Most of the existing clustering algorithms are devoted to finding arbitrary shapes and size of clusters; it is difficult to deal with those data sets that have different densities with these algorithms. Multi-density clustering algorithm is an effective method to solve this problem. But the existing multi-density clustering algorithms need manual thresholds to distinguish sparse and dense units and they are sensitive to the parameters. In addition, most of them are based on grid as a unit, lack of data observation in the grid, resulting in a low clustering accuracy.Aiming at the problem that most grid clustering algorithms need manual thresholds to identify dense units, a new method for automatically calculating the threshold of sparse units is presented. In this method, the image segmentation algorithm is combined with the grid clustering algorithm, and it can automatically calculate the sparse element threshold according to the data space. Experiments show that this method can effectively remove large area sparse grid. In order to deal with the multi-density data better, a new algorithm of multi-density grid clustering is proposed. The algorithm is based on the detection of windows and the internal data of the grid. Experiments show that the proposed algorithm can deal with multi-density data, and also, it has good time efficiency and can be applied to all kinds of data sets.Based on the research above, region information discovery of classified information platform is designed and developed. According to the business objectives, select the relevant dimension, design the actual application process. After the data quantity is satisfied, clustering analyze is proceeded. Finally, Baidu Maps API is called to display and interpreted the knowledge found in clustering results. Through the analysis of the clustering results, personalized ranking for regional information categories is demonstrated, and the category "crisis" of classified information platform is solved.