Research On Obstructed Space Clustering Algorithm

The research on obstructed space clustering derived by the development of spatial clustering research has become an important topic,focuses on dealing with clustering problems with obstacle constraints and improving spatial clustering results' authenticity,which makes results of analysis are more aligned with distribution status and development trend of real data.Obstructed space clusteing is a kind of process that incorporates data with obstacle constraints into clusters according to a certain similarity.It makes the gathered clusters don't overlap each other,and the similarity of spatial data in the same cluster is higher than that in different clusters.By combining barrier-free spatial clustering technology and obstacles processing method,the clustering results can reflect the obstacle's influence on rule of real spatial sample group distribution more true and reasonably,give expression to characristic of obstructed space structure,also play an important role of data characteristic mining in the process of exploring the real spatial data.In this paper,the research on obstructed space clustering can be divided into the following several parts:First of all,propose minimum obstructed variance heuristic certain data clustering algorithm to aim at the question of obstructed space clustering.By analyzing advantages and disadvantages of traditional spatial partition clustering algorithm,finds out the influence on clustering results' quality made by whether the intial model of partition clustering algorithm reflects overall characteristic of spatial data distribution is big, adopt the initial seeds selection method which is based on minimum obstructed variance to greatly reduce the initial seeds' random influence on obstructed space partition clustering results,help algorithm complete the obstructed space clustering objectively,and the result of algorithm has highertruth;Meanwhile finds out the limitation of existing obstacles can reduce the efficiency of the partition clustering algorithm,utilizing obstructed distance calculation method in the process of initial seed selection and clustering can improve algorithm's efficiency.Secondly,in order to improve the truth of clustering results,analyze the spatial data's common attribute of uncertainty in process of real data collection and think about the situation of obstacles interval,obstructed space uncertain data clustering algorithm is proposed.The algorithm utilizes approximate skeleton theory which can avoid clustering results become local extremum to form the initial seeds of clustering algorithm;and utilizes two kinds of pruning strategy to improve the efficiency of generating local optimal solution and final clustering results.Finally,for achieving obstructed space clustering results with characteristic of data attribute,the algorithm of fuzzy clustering uncertain data with obstacle constraints is proposed.Similarity of the algorithm is the analysis of combination with attribute of uncertainty of data with obstacle constraints and fuzzy,partitioning data to clusters for this reason may improve authenticity of clustering results.In addition,making use of hybrid intelligent optimization method to remedy shortcoming of local convergence of clustering in algorithm and incomprehensive clustering results to ensure validity and accuracy of clustering algorithm.