A Multi-density Gradient Grid Clustering Algorithm Based On The Optimal Division

Data mining is a method, which, from a huge number of randomly occurring,objective existence, incomplete and noisy data were extracted and find valuableinformation for business analysis and decision, thus, the theory of natural science, humangeography and provide effective technical support and reliable data basis. Among them,clustering analysis in data mining is a very important and interesting research topics, andbecause of its great application and research value and has attracted more and moreattention.In order to solve the grid clustering algorithm to parameter too sensitive, canautomatically identify the different density gradient and different gradient differentiatesbetween precise enough to wait for a problem, put forward A Multi-density Gradient GridClustering Algorithm based on the Optimal Division (OPMDG). This algorithm onlyneeds to input an approximate density threshold, grid length automatically calculated andcan automatically adjust the adaptation, the algorithm to reduce the sensitivity ofparameter; put forward double partition technology, can be found in different densitygradient class; for in similar on the junction point, introduced charge attraction betweenconcept, can to effectively solve the problem of low accuracy and category clustering.The experimental results show that, OPMDG algorithm can identify a variety ofdifferent shapes, size of the cluster, mining with different density gradient, and can effectively divide the boundary and removal of outliers and noise. In the GCODalgorithm and comparison, can obviously find OPMDG algorithm accuracy is higher, andthe time complexity is a small linear function.