A novel algorithm DBCAPSIC for clustering non-numeric data

Data mining techniques are playing an important role in the analysis of mass network information and big data nowadays. The cluster analysis, as a main kind of method in data mining, draws great interest from researchers of various fields who proposed many algorithms such as k-means algorithm and its variants, density-based algorithm and its variants. However, these algorithms all have their own problems. This paper focuses on some of the problems and proposes a novel algorithm DBCAPSIC. The algorithm overcomes the k-means algorithm´s sensitivity to initial conditions and avoids common density-based algorithms´ "clustering failure" in some cases. Also, the algorithm has the linear time complexity of O{n), compared to the quadratic time complexity of common density-based clustering algorithms.