This subsection is concerned with the scheme of computing the similarity between numerical vectors as illustrated in Figure 1. In this research, we call the traditional similarity measures such as cosine similarity and Euclidean distance Int'l Conf. on Advances in Big Data Analytics | ABDA'15 | 65 feature value similarities where consider only feature values for computing it. In this research, we consider the feature similarity as well as the feature value similarity for computing it as the similarity measure which is specialized for text mining tasks. The numerical vectors which represent texts or words tend to be strongly sparse; only feature value similarity becomes easily fragile to the tendency. Therefore, in this subsection, as the solution to the problem, we describe the proposed scheme of computing the similarity between numerical vectors.