RPCL text clustering based on concept indexing

Text feature space usually has huge dimensionality, and the number of clusters cannot be determined before clustering. Concept indexing (CI) method can reduce text feature space in dimensionality rapidly. CI firstly clusters the texts set into L subsets, then regards the centroid vectors of the L clusters as the axes of the reduced L-dimensional space, and finally projects text vector-space on the reduced L-dimensional space and make it to be L-dimensional text vector-space. RPCL text clustering can work effectively and determine the number of clusters properly. This paper presents a new method using CI to reduce dimensionality for RPCL text clustering. Experimental results show that this algorithm not only improves clustering efficiency greatly, but also makes the averaged accuracy reach a high level.