A Fast Density-Based Clustering Algorithm In The Presence Of Obstacles

The similarity judgments between the data objects in the Traditional clustering algorithms are different distance, such as Euclid distance and Manhattan distance. However, in the space containing obstacles, such traditional distances are out of effectiveness, since the connection among the data objects are sniped by obstacles. Hence, some scientists have begun the research in the clustering algorithms in the space containing obstacles. In such background, DBCO algorithm tries to probe the clustering algorithm in the space containing obstacles. In general, DBCO is a density-based clustering algorithm, which focuses on the following aspects:The first aspect is how to modeling the obstacles in the space. The purpose is to form the representation of obstacles so that facilitate the operation between obstacles and data objects. However, the modeling method should not be at the cost of extensive data preprocessing. Secondly, how to judge the similarity among data objects in the space containing obstacles is the core problem of the clustering algorithms with obstacles. Many work focused on how to eliminate the influence that obstacles bring to the similarity judgment. Such work has also been the key problem in the clustering containing obstacles. Thirdly, based on the two aspects mentioned above, how to ensure the efficiency of the clustering algorithms.Simplifying the obstacles model is a good method to reduce the data preprocessing. The new similarity judgment method not only ensures the disconnection between the objects separated by obstacles, but also contains the neighborhood's expanding in the direction facing the connection data and vertex of obstacles. The clustering expanding with chosen represented data point ensures the efficiency of the clustering algorithm in the space with obstacles.