Extracting hierarchical concept with cloud transform

It is essential to extract concepts hierarchy by hierarchy in conceptual ontology. In this paper, a method is proposed to extract hierarchical concept from database with cloud transform. By approaching the original data distribution, cloud transform makes the quantitative data changed into a series of qualitative concepts portrayed by atomic clouds on the bottom level. With the increasing level, the elemental clouds are synthesized level by level, the amount of which become less and less. On the top level, a cloud is generalized. Thus a hierarchical tree on the concepts comes into being by extracting the qualitative concepts from quantitative data level by level. As well as human thinking, it is a pan-concept tree because the boundary between two neighboring cloud-concepts in the same hierarchy is indeterminate for the data randomness and its fuzziness belonging to the concept. In order to get more reasonable hierarchical concepts when the concepts in a lower level are generalized up to the concepts in a higher level, the magnitude coefficient of each cloud droplet is further treated. Finally, to test the effectiveness and efficiency, a case is studied on the dataset of car price. The results show that the proposed method is able to reasonably generate a hierarchical pan-tree for a dataset as human being, and the ontology from these concepts further gives a better specification of the conceptualization.